Amiga.org
Amiga computer related discussion => Amiga Software Issues and Discussion => Topic started by: Acill on December 31, 2015, 10:29:28 PM
-
Ive installed the latest available release on OS3.9 with BB1-4 installed. I am using Picasso96 screens on an A4000T and mediator using a Voodoo 5 at 1440x900 24 bit. Also tried it on 1024x768.
Problem is the window it opens displays the browser pushed to the left with corrupted graphics. I cant select anything in it and its unusable. It does this on all the versions contained in the archive. I'll submit a photo later this evening.
-
NetSurf is not 100% stable on any of the classic platforms. There's a thread here with over 600 responses where many others have already posted screenshots. Take a look at these:
http://www.amiga.org/forums/showpost.php?p=794444&postcount=623
http://www.amiga.org/forums/showpost.php?p=794183&postcount=612
http://www.amiga.org/forums/showpost.php?p=794194&postcount=616
http://www.amiga.org/forums/showpost.php?p=794201&postcount=617
http://www.amiga.org/forums/showpost.php?p=794024&postcount=574
http://www.amiga.org/forums/showpost.php?p=794029&postcount=579
I've only tested up to I believe it was 3.2 or 3.3, don't know if it's gotten any better under the latest 3.4 DEV, I haven't had a chance to test it yet.
In any case, your best bet for a browser on classic systems is still, sadly, Ibrowse. Which actually wouldn't be that danged bad of a browser if it could be updated... *sigh*
Also recommend the WarpDT's for better performance: http://www.warpdt.co.uk/
(you can still register those as the author is still active in the community)
-
Oldsmobile_Mike : Prety much every link you gave is about Chris trying to port Netsurf to OS3.x. It is not even near to be useable, though it had some promising features already. It doesn't lock Computer, it has a real gui, it requires quite few memory, uses standard system libraries, etc. It is not publicly availlable.
Acill has tried most likely Arthur's version wich can be downloaded from aminet it uses SDL and is based to debugging/no gui version of Netsurf.
It is not usefull, you can load maybe one or two page before it crash.
-
Oldsmobile_Mike : Prety much every link you gave is about Chris trying to port Netsurf to OS3.x. It is not even near to be useable, though it had some promising features already. It doesn't lock Computer, it has a real gui, it requires quite few memory, uses standard system libraries, etc. It is not publicly availlable.
Sure it's publicly available. Here's Chris's 3.3 right here on Aminet: http://aminet.net/package/comm/www/netsurf
Not sure where 3.4 went, I recall seeing links to it somewhere a few months ago, but haven't had time to dig into it any further. Or maybe my memory's just playing tricks on me, lol. ;)
Edit - disregard that last, 3.4 is Artur's. Google to the rescue again! ;)
Edit again - Chris's 3.3 on Aminet is for OS4. Remind me again why I hate NetSurf? You can still get his version for classic systems though, otherwise how would I have gotten it? :banana:
-
The 3.4 version is the one I have tried. So its not working for anyone? Strange its even an option to download if that is the case.
-
The 3.4 version is the one I have tried. So its not working for anyone? Strange its even an option to download if that is the case.
It's still very, very, very beta.
The fact that there's like 50 different versions (okay, I'm exaggerating that number by a bit, but not by much) doesn't help ease the confusion any. :(
-
So Ibrowse is still the best usable browser on a 3.x machine?
-
So Ibrowse is still the best usable browser on a 3.x machine?
IMHO, yes. :)
-
Edit again - Chris's 3.3 on Aminet is for OS4. Remind me again why I hate NetSurf? You can still get his version for classic systems though, otherwise how would I have gotten it? :banana:
Try http://homepage.ntlworld.com/cdyoung/tmp/netsurf_os3.lha
btw v3.4 of NetSurf isn't released yet. Anything versioned 3.4 is a post-3.3 development snapshot.
-
Have you created new version to test?
-
Thanks for the version to test. Its VERY SLOW on my A4000T 060/PPC and looks like this when a page finally loads.
Try http://homepage.ntlworld.com/cdyoung/tmp/netsurf_os3.lha
btw v3.4 of NetSurf isn't released yet. Anything versioned 3.4 is a post-3.3 development snapshot.
-
-
-
Not really sure what to advise here, but Arti's port of Netsurf works fine for me. By far and away the best browser for 68k amigas.
Chris's version is thus far next to unusable, but hopefully one day usability will be up to that of Arti's version.
Chances are you'll hear a bunch of crap about "sdl framebuffer, non-gui, debug, etc." while enquiring about 68k Netsurf. Ignore it. It'll come from Chris, who obviously is going to promote his own version, and from other people who don't really understand what theyre talking about parroting him. Ask said people what those terms even mean, or how they influence (or not) things and they wont be able to answer you (sans Chris, who understands, but twists and contorts truths so as to promote his own version), but give it a little time and they'll be parroting again :-)
All in all though it's not a very good browser anyway compared to pretty much everything out there, but if you want a no frills, basic browser that renders partway correctly it's ok. Don't expect to be able to do even simple stuff like uploading images though. A complete lack of javascript makes these things nigh on impossible. Even 68k OWB is more capable.
-
I wouldn't be so quick to dismiss Chris's work. He has made some steady progres
-
I wouldn't be so quick to dismiss Chris's work. He has made some steady progres
Hope to see more of it too, it was nearly functional.
-
Not really sure what to advise here, but Arti's port of Netsurf works fine for me. By far and away the best browser for 68k amigas.
Chris's version is thus far next to unusable, but hopefully one day usability will be up to that of Arti's version.
Chances are you'll hear a bunch of crap about "sdl framebuffer, non-gui, debug, etc." while enquiring about 68k Netsurf. Ignore it. It'll come from Chris, who obviously is going to promote his own version, and from other people who don't really understand what theyre talking about parroting him. Ask said people what those terms even mean, or how they influence (or not) things and they wont be able to answer you (sans Chris, who understands, but twists and contorts truths so as to promote his own version), but give it a little time and they'll be parroting again :-)
All in all though it's not a very good browser anyway compared to pretty much everything out there, but if you want a no frills, basic browser that renders partway correctly it's ok. Don't expect to be able to do even simple stuff like uploading images though. A complete lack of javascript makes these things nigh on impossible. Even 68k OWB is more capable.
I have my doupts that Chris version woun't ever materialize to working version for 68k Amigas.
Arti's version is a hack, he has took a version of Netsurf wich hasn't meant to be used like that. He has made a huge job to make it work like it does now. It will always require two diffrent version for AGA and RTG. AGA version on based to Novacoder's SDL AGA, wich himself is descripted as "quick dirty hack". SDL doesn't support 8 bit screens.
I really hope that Arti get his version to work useable state, but compared to work he has took it doesn't look good. He need to rewrite quite much code, to get it useable.
Another thing is how usefull it will be? For me it just to possibility to download files occasionaly, without need to use anther Computer. Netsurf works well with that. General web browsing woun't be possible ever with real 68k Amigas. Maybe FPGA accelerators will change that some day.
Current version of netsurf has a java script. Saying "Even OWB is more capable", is not fair. OWB is complete web browser. It woun't ever run useable state with real 68k Amigas, just to starting requies 4x ram compared to Netsurf (11mb).
-
Have you created new version to test?
No. I need help to get any further.
-
Thanks for the version to test. Its VERY SLOW on my A4000T 060/PPC and looks like this when a page finally loads.
In Users/{your user, probably Default}/Choices try adding:
friend_bitmap:1
That might fix the colour issue (and maybe speed it up a bit).
These might help with memory/speed:
simple_refresh:1
dither_quality:0
screen_ydpi:72
I can't remember if I ever got use_diskfont:1 working properly, but if it works that will speed it up too.
It is built without compiler optimisations due to an issue with GCC I don't know how to fix, and there's a fair bit of non-optimal code (it uses outline fonts - which are horribly slow on 68k - and the image reduction works via DataTypes)
-
I have my doupts that Chris version woun't ever materialize to working version for 68k Amigas.
Arti's version is a hack, he has took a version of Netsurf wich hasn't meant to be used like that. He has made a huge job to make it work like it does now. It will always require two diffrent version for AGA and RTG. AGA version on based to Novacoder's SDL AGA, wich himself is descripted as "quick dirty hack". SDL doesn't support 8 bit screens.
I really hope that Arti get his version to work useable state, but compared to work he has took it doesn't look good. He need to rewrite quite much code, to get it useable.
Another thing is how usefull it will be? For me it just to possibility to download files occasionaly, without need to use anther Computer. Netsurf works well with that. General web browsing woun't be possible ever with real 68k Amigas. Maybe FPGA accelerators will change that some day.
Current version of netsurf has a java script. Saying "Even OWB is more capable", is not fair. OWB is complete web browser. It woun't ever run useable state with real 68k Amigas, just to starting requires 4x ram compared to Netsurf (11mb).
while i dont agree with fishy about that chris is twisting truths, but rather simply accept his opinion as another point of view i partly agree but partly disagree with. the above post perfectly portrays what fishy is talking about, the mindless parroting.
it would be courious to hear in detail, what that awful "hacks" are, artur has introduced, illustrated with code quotations. what are the parts of code he has to "rewrite" and beyond all why chris version is not satisfactory as is if it is in so much better state, because last time i tried arturs version worked that bad at least under uae.
now. what concerns progress, chris version doesnt seems to get there much faster than arturs one. in fact it seems to be on hold, while artur keeps delivering. would be good if those bragging so much about how bad chris version is and what needs to be done, would help either artur or chris, according to their insight.
what concerns me i might return to this subject as soon as my abilities improve, but being currently working on aros68k under deadwood guidance it may not be soon. on the other hand its possible that by then i can contribute better, so maybe its worth the wait.
also just in case, deadwood has proposed a bounty to port current odyssey to 68k, which has been given up due to no interest.
-
Hack = Add a feature to program wich original source doesn't have and woun't have. SDL AGA is good example, Novacoder himself said it "quick dirty hack". There is no 8bit screen support in SDL and never will be.
Of course coder can take a opensource code and make a changes to it, make it "his own". But future updates are not that easy implement any more. I'm not a coder that true, but I work with coders and seen problems wich are made By making unsupported features to original opensource code.
-
Hack = Add a feature to program wich original source doesn't have and woun't have. SDL AGA is good example, Novacoder himself said it "quick dirty hack". There is no 8bit screen support in SDL and never will be.
so adding features to the source results in this being "hack" by definition? sounds like software development is one big hack altogether.
and adding 8bit paletted support is an example for that? then netsurf is not going to have this anyway, no matter the used backend, just because it would be a hack. what are you expecting then? move on.
Of course coder can take a opensource code and make a changes to it, make it "his own". But future updates are not that easy implement any more. I'm not a coder that true, but I work with coders and seen problems wich are made By making unsupported features to original opensource code.
i admit, i dont know what are these hacks, arti and nova have introduced, it also depends on what one considers to be conform. imho i wouldnt want to manage the browser build with devcpp, this already makes it hard to be maintainable and the opportunity to push changes upstream becomes unlikely if not completely remote.
from this perspective it would be certainly better to have a target platform and developer within the original source tree and not an unsupported fork, as chris proposes. but as it is the fork is working and chris target is not. and there is none to help with it, including you. so go figure.
-
wawrzon: Just wondering, did you misread with purpose, wich would make you just one bully? Or didn't you just pay attention, I did say "don't have and WOUN'T HAVE". Programmer job can be to add features, but changes are evaluated and then added to orginal source.
Point was to point out that IF Chris get his version to useable state, implementing new version would ideally take about 30 min. Arthurs version cannot be build againts original source tree.
What happens if(when) Arti lost his interest to Netsurf? Who want to take over a custom Project? Without a help of original Project? Arthur has already invented his own version numbers to it?
and adding 8bit paletted support is an example for that? then netsurf is not going to have this anyway, no matter the used backend, just because it would be a hack. what are you expecting then? move on.
Was this sarcasm? What was point? Nova's 8bit support is adapted to SDL source tree? Or somebody has took over Project and is maintaing it?
What are hacks what Arti has made to his version Netsurf:
Resizeable screen
Menus
GUI (some sort of)
Chris needs a help from somebody who is experienced 68k coder. Addvertaising that might help him and in the end we could get useable web browser, wich has a future.
-
Hack = Add a feature to program wich original source doesn't have and woun't have. SDL AGA is good example, Novacoder himself said it "quick dirty hack". There is no 8bit screen support in SDL and never will be.
Eh, SDL supports 8-bit screens. Only problem is that many games want hi/true color screens.
Of course coder can take a opensource code and make a changes to it, make it "his own". But future updates are not that easy implement any more. I'm not a coder that true, but I work with coders and seen problems wich are made By making unsupported features to original opensource code.
It depends on what you do, where and how. Graphics stuff is isolated to its own Amiga native backend and if properly implemented it is easy to maintain (I havent seen SDL AGA code). At the moment there is no Amiga code in SDL source repository. It was removed some 10 years ago but Amiga backends are (were) maintained separately.
-
OK then I don't remember correctly why Novacoder calt his SDL AGA a dirty hack
Itix: Could you considering to help Chris? I do know that you don't have experience or interest with Reaction, but problem is elsewhere. Screen is not refreshed propely, etc.
Some years ago you said that you have considered to make 68k version of your netsurf MorphOS port.
-
OK then I don't remember correctly why Novacoder calt his SDL AGA a dirty hack
Itix: Could you considering to help Chris? I do know that you don't have experience or interest with Reaction, but problem is elsewhere. Screen is not refreshed propely, etc.
Some years ago you said that you have considered to make 68k version of your netsurf MorphOS port.
Chris version is using Reaction for GUI, Reaction is closed and not freely available so when you do not want to be dependent of 3.9 or 4.X it is a dead end. Arthurs version is perfectly running and not requiring a certain GUI. For my personal purpose it is perfect, reaction is a no-go.
-
OK then I don't remember correctly why Novacoder calt his SDL AGA a dirty hack
he may have said so, still he considered this hack to be worth to do, and the source is out in the open afair. on the contrary by your own definition, both odyssey and timberwolf are "hacks" because the code is not and cannot be included in the genuine repository upstream, and cannot be easily kept in sync with webkit or modzilla engines as you postulate.
Itix: Could you considering to help Chris? I do know that you don't have experience or interest with Reaction, but problem is elsewhere. Screen is not refreshed propely, etc.
Some years ago you said that you have considered to make 68k version of your netsurf MorphOS port.
itix helped me to get his old mui netsurf frontend to compile o certain extent (with zune) till i gave it up at least temporarily, since chris came up with his reaction port to amiga. maybe he can advise you if you start working on it yourself.
-
Why was reaction the choice? Because the OS4 version is using it? MUI seems like the best option now that both OS4, AOS and MOS all have a version on them.
-
Why was reaction the choice? Because the OS4 version is using it? MUI seems like the best option now that both OS4, AOS and MOS all have a version on them.
because chris frontend was done for os4 and os4 native gui kit is considered to be reaction. itix, according to himself, has once rewritten chris frontend early version to mui, which would be a more portable solution, but this rewrite is outdated and the engine interfaces apparently doesnt fit it anymore. thats what i was intending to attempt to have an universal frontend target for all amiga-like systems.
-
because chris frontend was done for os4 and os4 native gui kit is considered to be reaction. itix, according to himself, has once rewritten chris frontend early version to mui, which would be a more portable solution, but this rewrite is outdated and the engine interfaces apparently doesnt fit it anymore. thats what i was intending to attempt to have an universal frontend target for all amiga-like systems.
MUI would be more adorable than reaction, no doubt for that. MUI would be also more future proof solution, as it is availlable for free and it is availlable for all platforms and also a OS3.0/1. If there were a boynty for Netsurf it should have MUI gui. I would hapily participate.
Problem is if we collect some thousands money, would it help to find coder or or not?
-
MUI would be more adorable than reaction, no doubt for that. MUI would be also more future proof solution, as it is availlable for free and it is availlable for all platforms and also a OS3.0/1. If there were a boynty for Netsurf it should have MUI gui. I would hapily participate.
Problem is if we collect some thousands money, would it help to find coder or or not?
you could have participated in discussion on aros-exec about odyssey port to 68k in order to establish a bounty. i think some word has been spred. currently there is no agreement what to do and also mui, even if available for free in one or another form is not open, so its absolutely not as future proof as you imply. but there is zune.
once im ready with what im doing now ill discuss with deadwood the possibility to include odyssey in aros contribs and e it compile for 68k and maybe work on it some day. but not now. especially i have too limited experience yet.
-
from this perspective it would be certainly better to have a target platform and developer within the original source tree and not an unsupported fork, as chris proposes. but as it is the fork is working and chris target is not. and there is none to help with it, including you. so go figure.
I, for one, would pay good money if everybody could just work together toward a good, 68k/3.9-compatible browser.
Doesn't look like I'll be spending that money any time soon! :(
-
I, for one, would pay good money if everybody could just work together toward a good, 68k/3.9-compatible browser.
Doesn't look like I'll be spending that money any time soon! :(
kay, so there is two of you, you may contact deadwood about the issue.there is a thread about the bounties on eab:
http://eab.abime.net/showthread.php?p=1059700#post1059700
this is about aros68k, but the odyssey bounties were not neccesarily restricted to aros afair. though as there was not much interest its hard to determine, how the bounty aim would have been forumlated.
-
OWB is not a answer, at least I don't belive it would be useable with real Amigas. It would require emulator or FPGA accelerator with 256mb ram.
OWB takes about 45mb ram just to start? Netsurf requires 11mb. Target is not a general web browsing, it just wount be possible with current CPUs. I just want to have possiblity to do some googling and most importan download files from web.
-
OWB is not a answer, at least I don't belive it would be useable with real Amigas. It would require emulator or FPGA accelerator with 256mb ram.
OWB takes about 45mb ram just to start? Netsurf requires 11mb. Target is not a general web browsing, it just wount be possible with current CPUs. I just want to have possiblity to do some googling and most importan download files from web.
odyssey might not be usable for real amigas, thats true, but we dont know exactly until we try. one could fiddle with some constants or so and see if it makes a difference. people who have better insight into the source like deadwood, who as i know is sceptical about this, might tell more. however what i can say from my experience aros owb, which ai have been trying out on an amiga as well, has similar resource demands as netsurf.
-
It would need to be useable with OS3.1, 68040 and 32 mb ram.
Useable means something like this : I know http://www address wich have a file, wich I want to download. I can do that even if web page takes some time to load/display.
Problem is that we already have a proof of consept for those, it is possible with Netsurf. Getting OWB to same level would require time and effort, result is not known?
-
It would need to be useable with OS3.1, 68040 and 32 mb ram.
Useable means something like this : I know http://www address wich have a file, wich I want to download. I can do that even if web page takes some time to load/display.
Problem is that we already have a proof of consept for those, it is possible with Netsurf. Getting OWB to same level would require time and effort, result is not known?
im not sure if you are in a position to make demands. im not even sure the version of netsurf you advocate fulfils your own criteria any better than any other alternative, even treated as a proof of concept. certainly not as far as i have been able to test it.
this said i wish you luck to motivate anybody to do this work the way you want it to be done.
-
Sorry if you feel like I'm making demands, purpose is discussing this. But I'm happy to tell that I'm not a position to make demans, so don't worry, now you can be sure.
I'm not talking about Chris's version, I talk about Netsurf in general. Chris has proven memory usage and CPU requirements. Arti has proven speed.
My example was poor, sorry that. Point was that nobody shouldn't wait that general web browsin would be possible 68k amiga, with currently excisting hardware.
To get participants for boynty, requirements should be considered carefully, so that it would interest as many as possible. CPU 68040 is realistic minium target cpu, so is 16mb ram to at least start (not use), OS3.1 should be possible, if it is done with MUI gui.
030 wich most real Amiga users have, is not just possible to achive.
I use Netsurf a lot, downloading files with it works nicely. It even has same feature as iBrowse, I can close browser and downloading continues in background, until file is downloaded.
If deadwood proofs that those are possible with OWB, it is OK. But I do know that current OWB for 68k requires quite lot memory. 45-60mb? you propably know this better. So just wondering why waist time with it, if there is more likely solution?
-
I didnt start this thread for an argument. I simply wanted to know if getting Netsurf working on a classic Amiga with a PPC card was something I can do.
Any tips on using it, getting it running or anything else positive is appreciated. Anything else is not needed in the thread please. Disrespecting ANY amiga developer is silly and only results in less people wanting to contribute skill to keeping our platform alive.
-
OK then I don't remember correctly why Novacoder calt his SDL AGA a dirty hack
Itix: Could you considering to help Chris? I do know that you don't have experience or interest with Reaction, but problem is elsewhere. Screen is not refreshed propely, etc.
Some years ago you said that you have considered to make 68k version of your netsurf MorphOS port.
I'd like to help but I really lack both time and motivation as I dont have 68k machine here. I have tried to port few of my programs to 68k but they have stagnated...
-
I could loan machine to you? Maybe you could find motivation that way?
-
I could loan machine to you? Maybe you could find motivation that way?
Thats a cool offer!
-
I didnt start this thread for an argument. I simply wanted to know if getting Netsurf working on a classic Amiga with a PPC card was something I can do.
Is installing OS4 an option? Or maybe I could build a WarpOS version? I don't know how to do that though, I suspect it's not as easy as just using different GCC switches.
-
You can only do what you can Chris. Some parts will need help that is out of our control. Hopefully someone that knows that part will step forward and raise their hand!
-
How about something like ELynx?
After roughly an hours work I've got a mostly working 68k binary.
Supports CSS and JavaScript, has a download manager, and is pretty quick. Seems a better match than Netsurf for "real" amigas (and is more capable in many ways too).
Sidebar, but take those RiscOS netsurf requirements with a grain of salt. They're akin to saying Quake will run on an '020+FPU+8meg. Technically true, but far from usable.
If anyone is interested, for a few donations I can polish up and release elynx (after initial test bins are released of course). Failing that it'll just get thrown on the "good enough for my own use" pile.
I'm pretty content with 68k OWB and Arti's Netsurf port myself, but am willing to spend some more time on it for a little bit of money.
-
How about something like ELynx?
After roughly an hours work I've got a mostly working 68k binary.
ELynx is text only? Interesting idea though, if it's more compatible with modern websites than any of the current Amiga offerings...
-
How about something like ELynx?
After roughly an hours work I've got a mostly working 68k binary.
Supports CSS and JavaScript, has a download manager, and is pretty quick. Seems a better match than Netsurf for "real" amigas (and is more capable in many ways too).
Sidebar, but take those RiscOS netsurf requirements with a grain of salt. They're akin to saying Quake will run on an '020+FPU+8meg. Technically true, but far from usable.
If anyone is interested, for a few donations I can polish up and release elynx (after initial test bins are released of course). Failing that it'll just get thrown on the "good enough for my own use" pile.
I'm pretty content with 68k OWB and Arti's Netsurf port myself, but am willing to spend some more time on it for a little bit of money.
Hi there!
Very interesting indeed but do you mean: Lynx (http://lynx.invisible-island.net/) or Elinks (http://elinks.or.cz/)? Would you also consider open-sourcing your current version if no donations come your way?
Cheers!
-
Yeah, I knew I had the spelling wrong, just couldn't recall what it was, so I took a stab in the dark :-)
Anyway, I'd meant Elinks. Which I see now checking the site appears not to have been updated for about 3 years. Not uncommon for a projects website to go unmaintained though while there's activity in its source repository.
As for sharing what I've done, there's really nothing much to share. I just looked through my source code archive on my computer looking for something interesting to port to os3.x/aros. It seems nicely portable and thus far its mostly been getting my dev env up to scratch to get it to build. Having said that though I actually looked at this a few years ago, so there's a chance I'd already started the process back then and forgot (my dev env is full of assorted libs and headers Ive added over the years). Things may not have gone so smoothly on a "clean" (if there's such a thing) os3.x gcc+bin utils setup.
Anyway, Im having fun with this, so I'll release stuff as I go along. If people decide its worth a few dollars in donations here and there then that's just a bonus.
I get distracted easily, so lord knows how far I'll take this, but it's sort of surprising no-one ever attacked the 68k browser problem from the "bare bones start, add stuff as needed/wanted" type angle before.
-
Yeah, I knew I had the spelling wrong, just couldn't recall what it was, so I took a stab in the dark :-)
Anyway, I'd meant Elinks.
:laughing:
I get distracted easily, so lord knows how far I'll take this, but it's sort of surprising no-one ever attacked the 68k browser problem from the "bare bones start, add stuff as needed/wanted" type angle before.
Yes, I agree! I have been trying various versions of NetSurf and got even more interested when Deadwood released the source code of Odyssey Web Browser (http://www.amiga.org/forums/showthread.php?t=70300) but, then, got scared when he told me that it requires 500 MB or more of memory to load some (not so complex) pages!!! :insane:
There seems to be even support for SpiderMonkey in ELinks (http://elinks.or.cz/documentation/manual.html#ecmascript), did you have a chance to try to compile/run it?
Cheers!
-
PS. SpiderMonkey v1.5 CLI for m68k exist here (http://aminet.net/package/dev/lang/SpiderMonkey-js-bin) (haven't tested it yet!).
PPS. v1.5 dates back November 2000 (https://en.wikipedia.org/wiki/SpiderMonkey_%28software%29#Versions)... :o
-
PS. SpiderMonkey v1.5 CLI for m68k exist here (http://aminet.net/package/dev/lang/SpiderMonkey-js-bin) (haven't tested it yet!).
PPS. v1.5 dates back November 2000 (https://en.wikipedia.org/wiki/SpiderMonkey_%28software%29#Versions)... :o
v1.7 works on the Amiga. Anything later... yeah, good luck.
http://git.netsurf-browser.org/toolchains.git/tree/sdk/recipes/patches/spidermonkey/ppc-amigaos-1.7.0 (might need some minor changes for OS3)
-
v1.7 works on the Amiga. Anything later... yeah, good luck.
http://git.netsurf-browser.org/toolchains.git/tree/sdk/recipes/patches/spidermonkey/ppc-amigaos-1.7.0 (might need some minor changes for OS3)
Hi Chris!
Thanks for the info. and the link! Later versions would be a pain to have on AmigaOS?
Cheers!
-
It works great on my A1200 040/PPC OS3.9 128MB Voodoo3.
In less than a minute I got this:
(http://s26.postimg.org/nqn1aicnt/Net_Surf.png)
-
Hi Chris!
Thanks for the info. and the link! Later versions would be a pain to have on AmigaOS?
Cheers!
They need NSPR. Although I was following a bug/enhancement to decouple it from NSPR, I'm not sure that has actually happened yet (and it doesn't necessarily make it any easier).
IIRC v1.80 ought to be easy, but it doesn't work for reasons I don't remember (could be endian related).
NetSurf has moved over to Duktape because sufficiently new versions of Spidermonkey are such a PITA to port, and it's relatively heavy.
It's possibly worth looking at the Timberwolf source to get a working new-ish Spidermonkey.
-
How ELinks work with downloading? I would like to get some information how it performs. Is it so that these days, text only web browsing has a problems, when most of buttons are images, they would have to have a atl= tag, if not they show up only as a empty space?. It doesn't download images at all?
Sounds reasonable.
Yeah, I knew I had the spelling wrong, just couldn't recall what it was, so I took a stab in the dark :-)
Anyway, I'd meant Elinks. Which I see now checking the site appears not to have been updated for about 3 years. Not uncommon for a projects website to go unmaintained though while there's activity in its source repository.
As for sharing what I've done, there's really nothing much to share. I just looked through my source code archive on my computer looking for something interesting to port to os3.x/aros. It seems nicely portable and thus far its mostly been getting my dev env up to scratch to get it to build. Having said that though I actually looked at this a few years ago, so there's a chance I'd already started the process back then and forgot (my dev env is full of assorted libs and headers Ive added over the years). Things may not have gone so smoothly on a "clean" (if there's such a thing) os3.x gcc+bin utils setup.
Anyway, Im having fun with this, so I'll release stuff as I go along. If people decide its worth a few dollars in donations here and there then that's just a bonus.
I get distracted easily, so lord knows how far I'll take this, but it's sort of surprising no-one ever attacked the 68k browser problem from the "bare bones start, add stuff as needed/wanted" type angle before.
-
They need NSPR. Although I was following a bug/enhancement to decouple it from NSPR, I'm not sure that has actually happened yet (and it doesn't necessarily make it any easier).
Okay, understood :)
NetSurf has moved over to Duktape because sufficiently new versions of Spidermonkey are such a PITA to port, and it's relatively heavy.
Thanks for the pointer to Duktape, I didn't know this project but it seems to have very low memory footprint and has some very recent activity! Will try it during the week-end...
It's possibly worth looking at the Timberwolf source to get a working new-ish Spidermonkey.
Ouch :rolleyes: It is soooo big :)
Maybe integrating Duktape into Elinks could give m68k machines a nice, modern Web browser?
BTW, for some reasons, neither versions of Netsurf work well on my A1200 + Workbench 3.1 + Blizzard 68060 (no PPC, no RTG)... Is it just me? :mickeymouse:
Cheers!
-
It works great on my A1200 040/PPC OS3.9 128MB Voodoo3.
In less than a minute I got this:
(http://s26.postimg.org/nqn1aicnt/Net_Surf.png)
Nice! That is better output than I've ever seen on anyone's screenshots. What version is this?
-
Whelp, figured I hadn't tried Netsurf since I upgraded from an '040 to an '060, but still no love.
First I tried the old 3.4 DEV version that I had on my hard drive. Then I tried the version from Chris's post #9. Guess I need to find that text file command again to fix the graphics, but aside from all the font errors on loading being gone (and the crash when not loading from the ram drive), not much difference.
How the heck do you guys get yours to look so good? :confused:
-
The dev one sucks no matter what. I can onlg get Chris version working on 32 bit screen modes. 24, 8 or 16 looks like yours Mike.
-
The dev one sucks no matter what. I can onlg get Chris version working on 32 bit screen modes. 24, 8 or 16 looks like yours Mike.
Dang it! So I guess in addition to an '060, you need a Mediator. Since AFAIK no native Amiga graphics card supports 32-bit screenmodes. Bummer. :(
I normally run my Workbench in 16-bit. Tried it in 24-bit, no significant difference. Added "friend_bitmap:1" back into the choices file, that helped. Note that it has to be without the space, "friend_bitmap: 1" (with a space after the colon) doesn't work. Geeze is this program sensitive, lol. ;)
-
Dang it! So I guess in addition to an '060, you need a Mediator. Since AFAIK no native Amiga graphics card supports 32-bit screenmodes. Bummer. :(
I normally run my Workbench in 16-bit. Tried it in 24-bit, no significant difference. Added "friend_bitmap:1" back into the choices file, that helped. Note that it has to be without the space, "friend_bitmap: 1" (with a space after the colon) doesn't work. Geeze is this program sensitive, lol. ;)
Hi!
Nice to know! Could you please provide a screenshot for comparison?
Do you think that Chris' NetSurf would work with HIGHGFX: Super-High Res Laced in, say, in 64 colours? :rolleyes: Would you be able to try... on my Amiga, it keeps opening on its own screen and it just freezes there... :(
Thanks!
-
Do you think that Chris' NetSurf would work with HIGHGFX: Super-High Res Laced in, say, in 64 colours? :rolleyes:
I assume you're joking, right? :roflmao:
The dev one sucks no matter what. I can only get Chris version working on 32 bit screen modes.
He's not saying "32 colors", he's saying "a 32-bit screenmode". In other words, something you need a very powerful graphics card to use. :(
Edit: For your amusement, here's Netsurf on an 1152x900x256-color Workbench. About the best I can do with my graphics card. It took 53 seconds to load this page with an '060. Scrolling around was painful. ;) Also, why all the chip mem usage?
-
The dev one sucks no matter what. I can onlg get Chris version working on 32 bit screen modes. 24, 8 or 16 looks like yours Mike.
For me seems that you confuse Chris ja Artis's versions.
Chris version "works" with 8bit, 15bit, 16bit and 24bit screens. In other words as a minium it requires AGA Amiga.
-
For me seems that you confuse Chris ja Artis's versions.
Chris version "works" with 8bit, 15bit, 16bit and 24bit screens. In other words as a minium it requires AGA Amiga.
Download link? It's very confusing which version is which. :(
Edit: Added a pic to clarify.
-
Download link? It's very confusing which version is which. :(
Edit: Added a pic to clarify.
Mine is the one on the left. The minimalist interface is intentional, there are lots of options in the menus. It is possible to add a hotlist bar by moving items into the hotlist folder, although I don't know whether that works under OS3.
The graphical problems I could do with some help with. The font spacing issue irritates me, so some pointers on how to fix would be useful. I think I need to query OT_SpaceWidth and advance manually when encountering a space, but I couldn't get much sense out of it when I tried. From a user POV, installing and using a different CG font might also work.
-
Mine is the one on the left. The minimalist interface is intentional, there are lots of options in the menus. It is possible to add a hotlist bar by moving items into the hotlist folder, although I don't know whether that works under OS3.
The graphical problems I could do with some help with. The font spacing issue irritates me, so some pointers on how to fix would be useful. I think I need to query OT_SpaceWidth and advance manually when encountering a space, but I couldn't get much sense out of it when I tried. From a user POV, installing and using a different CG font might also work.
Thanks Chris! I wish I could help with the graphical problems, but aside from Commodore 64 BASIC and a JAVA class in college, I'm afraid I'm useless when it comes to coding. :(
Any idea what the one on the right is? The person who posted that screenshot said they "had it running in less than a minute".
I was thinking the one on the left was Artur's, and the one on the right was yours, also because the one on the left says it's "3.4", and I thought I remember you saying 3.4 "isn't released yet, and utri007 said Arthur's version is "the no gui version".
Bah, I'm so confused! :roflmao:
Edit: On a side note though, when I pulled up www.google.com (http://www.google.com) in yours, it looked damn good! I was kind of taken aback by how good it looked, honestly. Wish every site rendered that well! :) (10:30 minutes in this video (https://www.youtube.com/watch?v=OxQOaDAy2Yk))
-
@mike
I use NetSurf V3.
Older versions had mess palette due 32bit screenmode.
Version 3 works at 16bit on Voodoo3 P96.
Capture done via S-Grab.
-
For me seems that you confuse Chris ja Artis's versions.
Chris version "works" with 8bit, 15bit, 16bit and 24bit screens. In other words as a minium it requires AGA Amiga.
Then its the resolutions I have picked or something. When I switch screen modes the only ones that didn't corrupt were my 32 bit versions. Also yes the one on the right works for me. It seems Chris's version is the one I cant get working.
-
Edit: For your amusement, here's Netsurf on an 1152x900x256-color Workbench. About the best I can do with my graphics card. It took 53 seconds to load this page with an '060. Scrolling around was painful. ;) Also, why all the chip mem usage?
@Oldsmobile_Mike!
Thanks for the screenshot... Now, which version do you use? I am also confused at this point... Could you point me to it or PM me, please? :)
Thanks!
@All
Very interesting discussions!
-
Any idea what the one on the right is? The person who posted that screenshot said they "had it running in less than a minute".
That's Artur's, but I'm not familiar enough with it to say exactly which version.
I was thinking the one on the left was Artur's, and the one on the right was yours, also because the one on the left says it's "3.4", and I thought I remember you saying 3.4 "isn't released yet, and utri007 said Arthur's version is "the no gui version".
Heh. 3.4 isn't released yet, but anything built post-3.3 is versioned 3.4. You can tell the difference because it will say "(Dev)" after the version number. Release versions have the release date.
Artur's is the Framebuffer frontend, which is designed for systems with no GUI. It has a GUI, of course, because it's running windowed under Intuition.
Edit: On a side note though, when I pulled up www.google.com (http://www.google.com) in yours, it looked damn good! I was kind of taken aback by how good it looked, honestly. Wish every site rendered that well! :) (10:30 minutes in this video (https://www.youtube.com/watch?v=OxQOaDAy2Yk))
Yes, people keep posting screenshots of amiga.org, which doesn't render properly. Actually I've not seen a screenshot of anything else from that version!
Interesting video.
-
Artur's is the Framebuffer frontend, which is designed for systems with no GUI. It has a GUI, of course, because it's running windowed under Intuition.
PS. That's one of my confusion because on my '060 AGA (no RTG), I can't get any version (that I tried) to run "windowed"... Could someone points me to a version that would work on my Amiga, as a window on the Workbench screen? :)
-
PS. That's one of my confusion because on my '060 AGA (no RTG), I can't get any version (that I tried) to run "windowed"... Could someone points me to a version that would work on my Amiga, as a window on the Workbench screen? :)
Mine does that. It defaults to running on it's own screen and because the prefs window doesn't work on OS3, it's not obvious how to change it. So... you will need to run it once and quit it. Then open users/{your username, probably Default}/Choices and add the following line:
pubscreen_name:Workbench
That should make it open on WB instead.
Don't forget the following might be needed or helpful too:
friend_bitmap:1
window_simple_refresh:1
-
I can't remember if I ever got use_diskfont:1 working properly, but if it works that will speed it up too.
It appears this does work to some extent. It might be worth checking if this makes NetSurf any quicker:
use_diskfont:1
font_sans:Topaz
font_serif:Topaz
font_mono:Topaz
font_cursive:Topaz
font_fantasy:Topaz
(you can substatute Topaz for a nicer font of your choice, but note the font_mono needs to be fixed-width, like Topaz, and my quick test didn't involve any non-Topaz fonts)
With those I can scroll through the amiga.org homepage at a reasonable speed, on my emulated supposedly real speed '020 :)
I also have:
redraw_tile_size_x:100
redraw_tile_size_y:100
window_simple_refresh:1
-
Here is Artur's NetSurf WIP 24bit version.
I have added autdetect_depth option.
So colors should not be messed now.
Add this to Options and set to 1 to enable.
For now, this feature needs external porgram called getpixel in C:.
It detects current screen pixel format.
It is in archive. I hope to build it into NetSurf.
Close gadget and resizing isn't enabled.
I need to fix NovaCoder's SDL RTG lib.
It was used only for fullscreen games.
This is improved and updated version of SDL library.
Right click on 'Stop' button or ESC to exit.
Let me know how it works!
https://www.dropbox.com/s/jvgnvg7cqkk54a3/netsurf-24bit.7z?dl=1
-
Here is Artur's NetSurf WIP 24bit version.
I have added autdetect_depth option.
So colors should not be messed now.
Add this to Options and set to 1 to enable.
For now, this feature needs external porgram called getpixel in C:.
It detects current screen pixel format.
It is in archive. I hope to build it into NetSurf.
Close gadget and resizing isn't enabled.
I need to fix NovaCoder's SDL RTG lib.
It was used only for fullscreen games.
This is improved and updated version of SDL library.
Right click on 'Stop' button or ESC to exit.
Let me know how it works!
https://www.dropbox.com/s/jvgnvg7cqk...-24bit.7z?dl=1
I get a 404 error on this link.
-
I get a 404 error on this link.
Same. :(
-
Should be fixed now.
My bad, I copied shortened url from eab.
-
OK I made some testing!
First I made those changes to choises file wich Chris sugggested. Performance increase was unbelievable!
friend_bitmap:1
window_simple_refresh:1
redraw_tile_size_x:100
redraw_tile_size_y:100
use_diskfont:1
font_sans:Topaz
font_serif:Topaz
font_mono:Topaz
font_cursive:Topaz
font_fantasy:Topaz
Noticed that both amigaworld.net and wikipedia.com renders correctly? What is happened? I opened wikipedia after amigaworld, no ghost images any more. No Graphics corruption at all?? Would it be that true type fonts made those graphic corruptions?? Or I don't remeber what you said about this friend_bitmap:1 was it originally only for AGA machine? Or is it just a amiga.org wich causes a problems, this time I didn't test amiga.org at all.
If you trust that Netsurf own clock amigaworld.net opens now 22 seconds, compared to arthurs version from aminet.net it took 38 seconds to render amigaworld.net. At least it feels much much more faster now. Arthur's version also corrupted screen when resizing window and eventually crash machine. Locks Computer for tens of seconds, compared to Chris version wich doesn't or does for a second for time to time.
Scrolling work nicely without true type fonts with Chris version :D
Tested Arthurs version was ixemul / no true type fonts version. Clib2 version didn't start at all.
Arthurs version also consumes more than 10mb more memory.
I also made fast test with my AGA / 68040 machine, even with that machine amigaworld.net opens 34 seconds. It even has 8mb ram free from 34mb after that (1.5mb chip and 7.2mb fast). Tried to made tests with Arthur's AGA version, but it crash machine after it renders start page.
So Chris, would you be kind and continue work with this, it is already useable. Considering that it has NONE optimications, wondering how fast it could be?
Notice one little annoying feature. I mistyped address, it took minits to timeout.
-
First I made those changes to choises file wich Chris sugggested. Performance increase was unbelievable!
Good to hear! I've set these as default now and uploaded a new build.
There will be various core changes, but probably nothing significant (most of the work has been on Javascript, which I can't build in without the compiler optimisation bug being fixed).
http://homepage.ntlworld.com/cdyoung/tmp/netsurf_os3.lha
It's suggested to remove all the font lines from the Choices file, I've set some better defaults. If you want to switch back to outline fonts, you'll need to add font lines which specify outline fonts (it won't set the correct defaults)
-
It works, but it is slower than perivous version with Topaz font. Can't say how much, it doesn't display done text anymore.
It also has a new problem, is shoot down my Network from both amigas. I can download one page, after that it says "dns lookup failed", when I try to browse another web page. Need to reboot amiga, to get it back online. I also tested amia.org, it ;) displayded very fast "dns lookup failed", did you made some changes for timouts?
Amiga.org doesn't display right, but ghost images are gone.
Quiting takes ages and freeing memory also, but it does it.
I will test it with Topaz fonts to see does it make any difference.
-
Some tests more, changing font back to topaz didn't speed up things. However quitting program is fast. Is it even possbile, or was it because I just opened amigaworld.net nothing more? It still much faster than with true type fonts and scrolling is OK.
Network is a problem, it does it every time, I can surf just a one page, after that I need to restart tcp/ip stack. It also does it time to time from start page.
-
It works, but it is slower than perivous version with Topaz font. Can't say how much, it doesn't display done text anymore.
It will be slower because there are more fonts to look through. The data is cached, but lookup is just in an Exec list, so it's not particularly fast (on OS4 I've used a skiplist, but they don't exist in OS3's utility.library). I tried using a splaytree on OS3 but it just crashed, and I couldn't see any perceivable performance benefit on OS4, so I abandoned it.
(edit sorry, I'm talking nonsense, the exec list caching is only for scaleable fonts, bitmap fonts aren't caching at all, which might be part of the problem)
It also has a new problem, is shoot down my Network from both amigas. I can download one page, after that it says "dns lookup failed", when I try to browse another web page. Need to reboot amiga, to get it back online. I also tested amia.org, it ;) displayded very fast "dns lookup failed", did you made some changes for timouts?
ISTR there have been some changes because RISC OS was having timeout problems. If it's knocking out the network, it's probably overwriting memory. However I notice my toolchain is out of date, as at least libcurl has been updated, so I'm rebuilding that and we can see if it helps.
I will test it with Topaz fonts to see does it make any difference.
Topaz is likely to be more reliable, as it's fixed-width, so any bugs in the layout calculations are less likely to show up.
and.... re-built with latest components, archive in usual place. (haven't tested it)
-
I would say that Topaz or any other hand defined font is OK, if speed diffrence is that big. Bad that there is no way to define font like AWeb does?
68k has it limitations, so better just accept them. :) Better have faster than prettier.
-
Network problem is now solved.
Scrolling etc is OK.
It is still much slower than previous version with topaz font. AGA still have some problems with png pictures?? Is it because a datatypes or it can be because of FBlit or both.
-
I would say that Topaz or any other hand defined font is OK, if speed diffrence is that big. Bad that there is no way to define font like AWeb does?
You can define fonts as you've been doing (different fonts for sans, serif, mono). I think the only difference is that AWeb lets you set the sizes for headers etc. This is mostly irrelevant with CSS. NetSurf has a default font size, and the other sizes are calculated from this (AmigaOS scales bitmap fonts if exact size is not available). If the size is defined as an absolute point or pixel height that will be respected. There's a DPI option that can affect the size on screen for those (although I don't think that works for bitmap fonts; probably screen_ydpi:72 is a good option to set to stop things going out of proportion)
Network problem is now solved.
Scrolling etc is OK.
It is still much slower than previous version with topaz font.
Odd. I don't know remember what I've changed since the previous version. I'll need to nose through the git logs.
AGA still have some problems with png pictures?? Is it because a datatypes or it can be because of FBlit or both.
Don't know. Is it just PNGs? What's the problem exactly? Have you tried friend_bitmap:0? (it's on by default now)
If it's just PNGs I can try disabling the internal decoder so it uses the datatype instead.
-
I let netsurf open for a few hous, network dind't work after that. Somebody else should test this also, so that it is not just my network.
I'm prety sure that it just a png pictures, like this http://www.hollywood-mal.com/img/plananarama_small.png
So far best and fastest Netsurf version out. :) Thanks. Though previous version from summer was faster with topaz fonts. Now is only little bit faster than other one, it just works much more reliable.
-
Good to hear! I've set these as default now and uploaded a new build.
There will be various core changes, but probably nothing significant (most of the work has been on Javascript, which I can't build in without the compiler optimisation bug being fixed).
http://homepage.ntlworld.com/cdyoung/tmp/netsurf_os3.lha
It's suggested to remove all the font lines from the Choices file, I've set some better defaults. If you want to switch back to outline fonts, you'll need to add font lines which specify outline fonts (it won't set the correct defaults)
Hi there!
I just realised that NetSurf would not run on Workbench 3.1... Is 3.5 the minimum? Is that a strict requirement? (For example, could I hope to make it run on my 3.1 if I copy some of the 3.5 libraries?)
Cheers!
-
I'm prety sure that it just a png pictures, like this http://www.hollywood-mal.com/img/plananarama_small.png
Interesting. I tried disabling the internal PNG reader so my DataTypes reader would handle PNGs, but for some reason it isn't recognising that file as an image, so it looks like my DataTypes reader is broken under OS3 (I suspect I'm using too new picture.datatype tags).
I don't have time to investigate now.
edit actually, scratch that, it looks like I'm using some datatypes.library v45 features, and I'm on v44 here so they don't work. If you're on OS3.9 it will probably be OK. Try http://homepage.ntlworld.com/cdyoung/tmp/netsurfnopng.lha
-
I just realised that NetSurf would not run on Workbench 3.1... Is 3.5 the minimum? Is that a strict requirement? (For example, could I hope to make it run on my 3.1 if I copy some of the 3.5 libraries?)
3.5 is the minimum - it needs all the Reaction classes and newer icon.library, picture.datatype.
-
edit actually, scratch that, it looks like I'm using some datatypes.library v45 features, and I'm on v44 here so they don't work. If you're on OS3.9 it will probably be OK. Try http://homepage.ntlworld.com/cdyoung/tmp/netsurfnopng.lha
Tested, it seems to disable png images? Somebody else should test this also. Even in startup page has a png image, if it is trashed problem is with all AGA users. I tested it without FBLit didn't help.
Download works now :D thanks. This is starting to be usefull.
Bad that speed is degraded so much from summer version. :(
Am I right but time calculator is now in bottom of Windows and it is "displayed" outside of window, so that it is not visible?
Could it be possible / easy to make so that quitting is fast any time? If Netsurf is doing somethin, it doesn't allow quiting, it can take minits before it quits.
There a still some display / rendering problems, like amigaworld menus.
-
Tested, it seems to disable png images? Somebody else should test this also. Even in startup page has a png image, if it is trashed problem is with all AGA users. I tested it without FBLit didn't help.
Which version of datatypes.library do you have? And picture.datatype whilst you're checking version numbers?
Bad that speed is degraded so much from summer version. :(
I doubt I can do much about this unless you can point to some glaring error I've introduced several months ago. I have no idea what state the bitmap code was in in that build, I only tried it here in a new build but I know I had been changing the bitmap code inbetween. Probably there was some quick hacky code I'd put in for testing and replacing it with a working implementation has slowed it down. Also note that fonts aren't being cached, there are so many different parameters for each open request that I'm not sure how to keep track of them... or if it is even neccessary.
Am I right but time calculator is now in bottom of Windows and it is "displayed" outside of window, so that it is not visible?
It should be in the window border but border gadgets appear to be completely broken in OS3. I've added a non-border status bar in the new build (usual place).
Could it be possible / easy to make so that quitting is fast any time? If Netsurf is doing somethin, it doesn't allow quiting, it can take minits before it quits.
It's single tasking, so if busy you have to wait.
There a still some display / rendering problems, like amigaworld menus.
Yes there are, exactly the same amount as in the early builds, as I've made no attempt to fix them.
The bitmap font layouting is slightly broken too.
-
OK, I think the image problem (and the aw.net menu bar corruption) is related to the tiled rendering, and caused by the same bug which causes the previous page to show through on the amiga.org homepage.
I don't know how to fix this though.
You can turn off tiled rendering by setting the tile sizes to the same as your screen size (or 0, possibly).
btw, I'm pissing about with this on an (emulated at real speed) 020/AGA and it's perfectly usable. It's only the fetching/processing speed that is annoyingly slow.
-
What ever you did for it last summer, you did it right. It was twice as fast than it is now.
Downloading amigaworld.net takes 45.4 seconds (22 seconds) and surpricingly my 040 amiga does it same time 45.9 seconds.
I can log on to the amigaworld.net, but it takes about a 5 minits. Also after that there is starange screen corruption. See attached picture. Nothing like this hasn't happened before with this version.
Is it possible to make it stop, at least downloading page?? After that Quit would be fast?
My datatypes.library is 44.47
PS. May I suggest that you fix what you can? Not spend too much time to those problems, wich takes too much energy. It could be possible that you found solution one day, it is perfectly useable now. Maybe you could fix preferences? Note also, that you are best to say what to put users/choises file
-
Downloading amigaworld.net takes 45.4 seconds (22 seconds) and surpricingly my 040 amiga does it same time 45.9 seconds.
I can log on to the amigaworld.net, but it takes about a 5 minits.
Shows up in 15s here (using NetSurf's timer). It logged in quick, and then infylooped during layout (bitmap font layout bug). 5 mins would probably be because of a timeout.
Also after that there is starange screen corruption. See attached picture. Nothing like this hasn't happened before with this version.
That's the same plotter bug again.
Is it possible to make it stop, at least downloading page?? After that Quit would be fast?
Acting on the close gadget click already happens as fast as it can. If it is doing heavy layout work it doesn't necessarily yield to the GUI very quickly. All the layouting, downloading and GUI events happen on the same process. Most people don't decide to quit a web browser whilst it is still downloading the last page they requested!
My datatypes.library is 44.47
And picture.datatype? You're using OS3.9, right? That at least explains why the datatypes loader doesn't work.
Maybe you could fix preferences?
Yes, that should be relatively easy but time-consuming as I'll need to go through and figure out what it doesn't like by trial and error.
Note also, that you are best to say what to put users/choises file
The defaults should be sensible now.
-
Utri007 - Thanks for the tips, and for the link. I've been following along on the thread unfortunately have been too busy to respond. In any case, with the new download and those options in-place, it is a big improvement in performance on my system. 25-35 seconds for amiga.org, 3-6 seconds for google.com (I reloaded these pages a couple times), etc. Scrolling is also significantly faster.
Unfortunately more complex pages still seem to take an age ("fetching, processing"). cnn.com took almost three minutes to load. Wikipedia took 673 seconds (and boy was it ugly in all Topaz font). But hey, we're getting there. Also we're now in a realm of pages that IBrowse can't even load - I tried bankofamerica.com, for example - IBrowse just throws an error about "outdated encryption" or something along those lines. NetSurf at least loaded the first page of it. ;)
Aminet also works and I was able to download files (yay, new version of Peter K. icon library! ;) )
I concur with the bug about it freezing up now, however. As a test I left it sitting at one page for about an hour while I went off and did other things, and when I came back to it it was completely frozen. Wouldn't exit, wouldn't refresh, etc. The system was still operational, I could load other programs, but no sign of life from NetSurf. Had to give it the three-finger-salute to shut it down.
More testing required...
Chris - any possibility of reaching out to any of the original IBrowse dev team? I know at least Oliver Roberts was still active in the Amiga community, with his WarpDT's (which are great, btw). I vaguely recall hearing something about him having an update for SSL, as well. Perhaps he could help you with your datatype issues, at least?
Just a thought.
My .02 cents, it is a real shame what happened there. They had a great browser, and then just :( Doesn't make sense to not even be accepting registrations anymore, since I'm sure at least a few people would like to pay for it rather than download a pirate copy... :( :(
Another probably obvious question for Chris - since this is built off NetSurf source, there's no possibility of you ever going commercial with it, is there?
-
Chris as a Oldsmobile_Mike said, I would also hapily pay for a working solution.
Oldsmobile_Mike : I don't belive that there is no way that we could never use our 68k amiga for general web browsin, not even if there where CSS capable iBrowse. :( What would make me happy would be possibility to download files, if needed without going to another machine.
Did you try new build or old build with those swicthes?
-
You can turn off tiled rendering by setting the tile sizes to the same as your screen size (or 0, possibly).
How I do that, it should be tested. If porblem is only with tiled rendering, maybe it could be disable as default?
Is it possible to make it stop, at least downloading page?? After that Quit would be fast?
Acting on the close gadget click already happens as fast as it can. If it is doing heavy layout work it doesn't necessarily yield to the GUI very quickly. All the layouting, downloading and GUI events happen on the same process. Most people don't decide to quit a web browser whilst it is still downloading the last page they requested!
Think that this would be useable now, when Netsurf doesn't work 100% right. Like those time outs with loging amigaworld.net. If I try to close browse, it quits after 5 minits, when it gets time outed. Also with such a limite raw CPU power, it is easy get a situation where browser is occupied several minits. Of course if this is hard or imposible to do, then situation is what it is.
I also doubt that Oldsmobile_Mike problem is related to time outs?. For me it just terminated lan/web and didn't react anything. Could it be that Netsurd tries to do something and there is no time out for that?
-
I concur with the bug about it freezing up now, however. As a test I left it sitting at one page for about an hour while I went off and did other things, and when I came back to it it was completely frozen. Wouldn't exit, wouldn't refresh, etc. The system was still operational, I could load other programs, but no sign of life from NetSurf. Had to give it the three-finger-salute to shut it down.
Yeah, that's the infinite loop layout bug. I hate that code so much I'm not keen on fiddling with it.
Another probably obvious question for Chris - since this is built off NetSurf source, there's no possibility of you ever going commercial with it, is there?
No.
-
Picture.datatype is 45.17
Reinstalled Netsurf and noticed that it is slow, amigaworld render took 80 seconds. I made changes to choises file, it was fast again.
My choises file is :
theme:PROGDIR:Resources/Themes/Default
pubscreen_name:Workbench
window_simple_refresh:1
redraw_tile_size_x:100
redraw_tile_size_y:100
use_diskfont:1
font_sans:Topaz
font_serif:Topaz
font_mono:Topaz
font_cursive:Topaz
font_fantasy:Topaz
Should friend_bitmap be ON or OFF? 0/1 or totally removed from choises file?
How do I do this : You can turn off tiled rendering by setting the tile sizes to the same as your screen size (or 0, possibly).
-
Dear all!
I still have to install OS 3.5 (or 3.9) to test the new versions of NetSurf following Chris' advice but I wanted to support the proposal below:
Chris as a Oldsmobile_Mike said, I would also hapily pay for a working solution.
+1
I would happily donate to Chris or a bounty or... :)
Oldsmobile_Mike : I don't belive that there is no way that we could never use our 68k amiga for general web browsin, not even if there where CSS capable iBrowse. :( What would make me happy would be possibility to download files, if needed without going to another machine.
Same here, I cannot believe that, in 2016, we could not have a decent Web browser or high-end 68k Amiga. At least something that renders okay most pages and allow "normal" uses...
Of course, your "normal" is that my "normal" :laughing: but you get my meaning and, besides, Oldsmobile_Mike's screenshots look quite "normal": Wikipedia, Google, Aminet... all these sites are probably regularly used by most of us...
Cheers!
-
I checked Chris' home page, there is no donate button. :(
-
Picture.datatype is 45.17
Ta. So the bitmap scaling should be working correctly (not that I've noticed any problems with v44), it's just the datatypes linking up that won't work. I should be able to work around that, but it's not a priority as I don't think it'll solve anything.
Reinstalled Netsurf and noticed that it is slow, amigaworld render took 80 seconds. I made changes to choises file, it was fast again.
My choises file is :
theme:PROGDIR:Resources/Themes/Default
pubscreen_name:Workbench
window_simple_refresh:1
redraw_tile_size_x:100
redraw_tile_size_y:100
use_diskfont:1
font_sans:Topaz
font_serif:Topaz
font_mono:Topaz
font_cursive:Topaz
font_fantasy:Topaz
That can only be down to your choice of fonts; Topaz is always going to be faster for a number of reasons. The rest there are defaults (actually the top two are added by the install script).
You can see your running config by visiting about:config
Should friend_bitmap be ON or OFF? 0/1 or totally removed from choises file?
Whatever works best. 1 is the default. If that works, I'd leave it like that as it could makes blits faster.
How do I do this : You can turn off tiled rendering by setting the tile sizes to the same as your screen size (or 0, possibly).
redraw_tile_size_x:0
redraw_tile_size_y:0
0 will either turn it off or make NetSurf crash. Please let me know which!
I checked Chris' home page, there is no donate button. :(
Actually there is, although it took me a while to find it.
-
Tile size y/x 0/0 doesn't work it has miniums and maximum size.
But I noticed that from menu Browser - > Images and disable background images, solves prety much every display problem. Even Amiga.org looks about right. Menu text are readable etc. nothing is in wrong place etc. Even logging to amigaworld.net doeasn't take more than 87 seconds, or is this just a coinside?. Hope this helps.
There is no config under the about, there is a credits and license.
You have hidden that donate button very well in your home page.
-
Tile size y/x 0/0 doesn't work it has miniums and maximum size.
Set them to your screen size then. There's definitely no maximum (or at least, the maximum isn't any smaller than the screen)
But I noticed that from menu Browser - > Images and disable background images, solves prety much every display problem. Even Amiga.org looks about right. Menu text are readable etc. nothing is in wrong place etc. Even logging to amigaworld.net doeasn't take more than 87 seconds, or is this just a coinside?. Hope this helps.
No, probably not. A lot of the background images are repeated tiles - on OS4 I use layers.library's backfill functions to tile plot very efficiently. On OS3 I can't do that, so it's probably my old slow (and seemingly broken) code. That would certainly explain some of the performance problems. I don't know why it fixes the rendering, but that's interesting useful information, thanks.
There is no config under the about, there is a credits and license.
I mean, browse directly to "about:config"
You have hidden that donate button very well in your home page.
It's only there because some people were asking, and it's on the most relevant page!
If you donate then it's for work already done, not future progress. But, actually, the feedback is more motivating and useful than donations.
-
I tested tile size 1024/768 it didn't work OR was so slow that my patience ended. ;) It didn't gave any errors like 0/0 did.
No, probably not. A lot of the background images are repeated tiles - on OS4 I use layers.library's backfill functions to tile plot very efficiently. On OS3 I can't do that, so it's probably my old slow (and seemingly broken) code. That would certainly explain some of the performance problems. I don't know why it fixes the rendering, but that's interesting useful information, thanks.
Would use of Thomas Richter's back port of OS4 layers.library help and make things significantly easier?
-
Would use of Thomas Richter's back port of OS4 layers.library help and make things significantly easier?
Possibly - I didn't know such a thing existed.
I've just uplaoded a new version with some of the text layout problems fixed. It's still broken (missing text on the welcome page!) but shouldn't go into an infinite loop any more.
-
Possibly - I didn't know such a thing existed.
I've just uplaoded a new version with some of the text layout problems fixed. It's still broken (missing text on the welcome page!) but shouldn't go into an infinite loop any more.
I think this is the one he's referring to. Long thread about it here: http://www.amiga.org/forums/showthread.php?t=67649
And download here: http://aminet.net/package/util/sys/layers
Is http://www.unsatisfactorysoftware.co.uk/index.php?pg=netsurf is the best place to download your latest builds? Thanks!
-
Tested, I think that loop is still there? But it indeed looks now some way different?
Second pic is without background. Looks good. If I disable background, I can't turn it ON again from menu/gui. Need to edit user/choises file.
-
I think this is the one he's referring to. Long thread about it here: http://www.amiga.org/forums/showthread.php?t=67649
And download here: http://aminet.net/package/util/sys/layers
Interesting. I've had a look at it, and at my code, and... it appears I only use one OS4 function, which is avoidable. I've ported it over to OS3, and the annoying layout problems have gone! Hurray!
Is http://www.unsatisfactorysoftware.co.uk/index.php?pg=netsurf is the best place to download your latest builds? Thanks!
Yes, although it only points to the same place as the other link I've been giving out.
New version there now with the layers-based tile plotter.
-
Nice, I will do some more testing and report on how its working on my system. I still dont understand why it takes as long to load pages as it does though! I have 408MB free and an 060, lol.
-
Working nice now but still slow. I cant adjust any of the prefs if I go into settings. Also I noticed if I play with the scale setting and reduce it the screen gets very angry and doesnt redraw properly. It starts to mix up previous screens with new. If I return it back to normal it fixes itself.
Looking forward to the day this is optimised and works as well as Ibrowse for general browsing if that happens!
-
Hi fishy_fiz and all!
Anyway, I'd meant Elinks. Which I see now checking the site appears not to have been updated for about 3 years. Not uncommon for a projects website to go unmaintained though while there's activity in its source repository.
Sorry to change the topic of this thread back and forth (maybe this thread should be split between a thread on NetSurf and another on "alternative" Web browsers?) but following your idea, fishy_fiz, I took Links v2.12 (http://links.twibright.com/download.php) and managed to compile and run it :) I had to make a small change to handle correctly the arrow keys. (Why would the arrow key codes be different on Amiga and other OSes? :rolleyes:)
(http://www.chingu.asia/wiki/userfiles/image/Programming/Links/Links%20v2.12%201.png)
I also managed to get Duktape (http://duktape.org/) to compile and to run with Sami Vaarala's help... It is quite easy to use and fast actually!!! :banana:
So, I will now try to compile ELinks, which seems in colour (http://elinks.or.cz/screenshots/google-news--2004-09-29.png) :laughing:
And, maybe, one day, put ELinks and Duktape together?
Cheers!
-
Congratulations Chris. :) well done. Need to test new build this evening.
Working nice now but still slow. I cant adjust any of the prefs if I go into settings. Also I noticed if I play with the scale setting and reduce it the screen gets very angry and doesnt redraw properly. It starts to mix up previous screens with new. If I return it back to normal it fixes itself.
Looking forward to the day this is optimised and works as well as Ibrowse for general browsing if that happens!
If you but something like this to user/choises file it is much faster.
use_diskfont:1
font_sans:Topaz
font_serif:Topaz
font_mono:Topaz
font_cursive:Topaz
font_fantasy:Topaz
Matthey suggested these fonts http://www.amiga.org.ru/websurf/
I also managed to get Duktape to compile and to run with Sami Vaarala's help... It is quite easy to use and fast actually!!!
This is interesting, would this be usefull with netsurf? How modular it is, would it allow easily external java script engine? What are term of use with Ducktape?
-
Hi fishy_fiz and all!
Sorry to change the topic of this thread back and forth (maybe this thread should be split between a thread on NetSurf and another on "alternative" Web browsers?) but following your idea, fishy_fiz, I took Links v2.12 (http://links.twibright.com/download.php) and managed to compile and run it :) I had to make a small change to handle correctly the arrow keys. (Why would the arrow key codes be different on Amiga and other OSes? :rolleyes:)
I also managed to get Duktape (http://duktape.org/) to compile and to run with Sami Vaarala's help... It is quite easy to use and fast actually!!! :banana:
So, I will now try to compile ELinks, which seems in colour (http://elinks.or.cz/screenshots/google-news--2004-09-29.png) :laughing:
And, maybe, one day, put ELinks and Duktape together?
Cheers!
Sounds great!
Any chance you could give us a download? :)
-
...Links v2.12...I also managed to get Duktape (http://duktape.org/) to compile and to run...So, I will now try to compile ELinks
How cool's that? :D
-
If you check about:config, there are a couple of lines which don't add up:
PROGDIR:Users//FontGlyphCache
PROGDIR:Users//Cookies
PROGDIR:Users//URLdb
PROGDIR:Users//Hotlist
This would mean these files are in progdir: instead of 'users'.
-
If you but something like this to user/choises file it is much faster.
use_diskfont:1
font_sans:Topaz
font_serif:Topaz
font_mono:Topaz
font_cursive:Topaz
font_fantasy:Topaz
Matthey suggested these fonts http://www.amiga.org.ru/websurf/
Does speed up some, but it sure is ugly at 1024x768!
-
If you check about:config, there are a couple of lines which don't add up:
PROGDIR:Users//FontGlyphCache
PROGDIR:Users//Cookies
PROGDIR:Users//URLdb
PROGDIR:Users//Hotlist
This would mean these files are in progdir: instead of 'users'.
There should be a "Default" or some other value between those two slashes. Does GETENV USER return anything?
-
There should be a "Default" or some other value between those two slashes. Does GETENV USER return anything?
Setting user to 'default' works now.
Still some (most ?) of the options aren't working/loaded yet I guess.
For example: homepage_url works, but enable_javascript doesn't.
-
Setting user to 'default' works now.
Still some (most ?) of the options aren't working/loaded yet I guess.
For example: homepage_url works, but enable_javascript doesn't.
JS isn't compiled in, because GCC optimisation issues I can't fix.
-
JS isn't compiled in, because GCC optimisation issues I can't fix.
That was just an example. Netsurf should still be able to read the option and show it in about:config. Same with block_advertisements & do_not_track (and others ?).
Still have to find out what caused the crash when reading fonts the first time though (changed all the fonts with another set).
-
That was just an example. Netsurf should still be able to read the option and show it in about:config. Same with block_advertisements & do_not_track (and others ?).
Yes, it should, but the Javascript option is forced off if it isn't compiled in.
block_ads and dnt should be read, are you sure they're not?
Still have to find out what caused the crash when reading fonts the first time though (changed all the fonts with another set).
I know the outline font scanner was crashing on OS3 at one point, but that isn't even called when bitmap (diskfont) fonts are set to be used.
-
Works :D Both AGA and RTG, no Graphics corruption.
Speed decreace was huge, just wondering is simple refresh still on? Or is it possible that load times vary? 55 - 90 seconds
[edit: it is still on] and only amiga.org takes longer time to render, amigaworld.net taks still 42 seconds with 060 and 47 seconds with 040
This is anyway really good start, now it is more motivating to make it more speedy? I ques :)
-
Found AGA only problem. No problems with text with previous build and no problems with amiga.org at least with fast test. PNG problem is still on with AGA.
[edit] Seems that AGA has problems with also a gif pictures
[edit] [edit] There seems to be problems when browsing from site to another site, time outs? Or when I left it open for a 15 minits and trying to surf to another site. That happenet same time with both amigas.
-
Found AGA only problem. No problems with text with previous build and no problems with amiga.org at least with fast test. PNG problem is still on with AGA.
[edit] Seems that AGA has problems with also a gif pictures
There are no problems with AGA here. Try switching friend_bitmap off.
Note the black corruption nonsense on aw.net is because that site insists on overlaying everything with a graphic that has alpha transparency. On OS3 I convert this to a mask at the 50% point. You can change this with mask_alpha in Choices (takes a value between 0 and 100)
-
There are no problems with AGA here. Try switching friend_bitmap off.
Note the black corruption nonsense on aw.net is because that site insists on overlaying everything with a graphic that has alpha transparency. On OS3 I convert this to a mask at the 50% point. You can change this with mask_alpha in Choices (takes a value between 0 and 100)
Tested both mask_alpha:0 and 100 no visible diffrence
Tested friend_bitmap:0 no visible diffrence
Both checked from about:config so that they are user defined.
Installed warpdatatype for png pictures, didn't do anything.
Random TCP/IP problems has started again.
-
Found the cause(s) why some options didn't work as expected.
1) I used true instead of 1.
2) I used a space before the option.
And the font handling, hmmmm. Seems to only like/put 1 font at the time in the cache file.
0x0020 "CGTriumvirate"
0x0021 "CGTriumvirate"
0x0022 "CGTriumvirate"
or
0x0020 "LetterGothic"
0x0021 "LetterGothic"
0x0022 "LetterGothic"
0x0023 "LetterGothic"
-
And the font handling, hmmmm. Seems to only like/put 1 font at the time in the cache file.
0x0020 "CGTriumvirate"
0x0021 "CGTriumvirate"
0x0022 "CGTriumvirate"
or
0x0020 "LetterGothic"
0x0021 "LetterGothic"
0x0022 "LetterGothic"
0x0023 "LetterGothic"
That's correct. This isn't really a cache file per se, but a fallback lookup table. The CG fonts don't have a very diverse range of glyphs, so the first font scanned is pretty much all you'll see.
You will only get this file if you have switched off using diskfont - use_diskfont:0
The default for OS3 is use_diskfont:1 (if you're using the latest version).
If that isn't setting itself for some reason please tell me.
-
You will only get this file if you have switched off using diskfont - use_diskfont:0
The default for OS3 is use_diskfont:1 (if you're using the latest version).
If that isn't setting itself for some reason please tell me.
Well use_diskfont:0 locks up the machine completely.
-
Sounds great!
Any chance you could give us a download? :)
Hi!
Yes, absolutely! I will package the files and will upload Links on Aminet before the end of the week0end :)
Stay tuned!
-
Chris:
Downloaded latest version, there seems to be problems again with layout? Amiga.org looks like it looked some versions ago.
Before that I spend some time surfing web with iBrowse, AWeb and Netsurf.
Seems that it does kill my Network, need to quit and start genesis to get online again. It doesn't happen so often with 060/RTG, but lefting netsurf open for a some time kills a Network with both Computers. Goining offline and back online doesn't help, need to quit and start again.
Also noticed that surfing from site to site doubles time to display site?
Disabled background and foreground my 040/AGA it gives nice speed up and maybe? that way Network doesn't die so often? Does it have something to do with lack of cpu power?
Tested some fonts and noticed that using Xen font keeps all tekst readable. Topaz has some problem when scaled to small.
-
Sounds great!
Any chance you could give us a download? :)
Here you go! (http://www.amiga.org/forums/showthread.php?p=802267) :)
Enjoy!
-
Downloaded latest version, there seems to be problems again with layout? Amiga.org looks like it looked some versions ago.
I'm getting some weirdness with the bitmap font layout here, it isn't doing what it's supposed to and I don't know why. It is at least consistent on OS4, but on OS3 how it lays out the welcome page changes every time I run it. I suspect an uninitialised variable is getting used, but GCC warns about those and it hasn't picked anything up.
Seems that it does kill my Network, need to quit and start genesis to get online again. It doesn't happen so often with 060/RTG, but lefting netsurf open for a some time kills a Network with both Computers. Goining offline and back online doesn't help, need to quit and start again.
Almost certainly memory trashing. If you can find it (with Enforcer or something) let me know where it is. Might need Matthey's disassebling skills.
Tested some fonts and noticed that using Xen font keeps all tekst readable. Topaz has some problem when scaled to small.
No bitmap fonts scale particularly well.
I've put a new version up, but there's not much change. I've changed the way it calls the font code which might speed it up a tiny bit, but really I want to get to the bottom of those layout problems. It's something in http://git.netsurf-browser.org/netsurf.git/tree/amiga/font_diskfont.c if anybody wants to look at it.
-
Here you go! (http://www.amiga.org/forums/showthread.php?p=802267) :)
Enjoy!
Thank you!
-
I've put a new version up, but there's not much change. I've changed the way it calls the font code which might speed it up a tiny bit, but really I want to get to the bottom of those layout problems. It's something in http://git.netsurf-browser.org/netsurf.git/tree/amiga/font_diskfont.c if anybody wants to look at it.
Download link doesn't work, get http 404 error
http://homepage.ntlworld.com/cdyoung/tmp/netsurf_os3.lha
-
I'm getting some weirdness with the bitmap font layout here, it isn't doing what it's supposed to and I don't know why. It is at least consistent on OS4, but on OS3 how it lays out the welcome page changes every time I run it. I suspect an uninitialised variable is getting used, but GCC warns about those and it hasn't picked anything up..
Current version doesn't display fonts at all, but you propably know that already. :)
Weir thing was that it displayed them on my aga machine until I enabled background and foreground. I have those disable as a default, makes it faster. Couldn't get fonts displayed again when I disabled backgrounds.
Could it be simple as problems with choises file?
-
Current version doesn't display fonts at all, but you propably know that already. :)
edit 3 (deleted previous versions of this post!)
Should be fixed now, found the bug that was causing inconsistent layout and font sizes (and in some cases no text at all).
-
Well done :) nice speed up. 45 - > 35 downloading amiga.org
Why it takes 3x more time when moving amigaworld.net to amiga.org 104 seconds? It is not site depending, just surfing from site to another site takes 3x time.
With 040/AGA amigaworld.net alpha_mask is problem, changing value doesn't seems to affect anything? 1 or 100
-
Why it takes 3x more time when moving amigaworld.net to amiga.org 104 seconds? It is not site depending, just surfing from site to another site takes 3x time.
I would hazard a guess that it's related to NetSurf clearing out the cache. Try disabling the disk cache (disc_cache_size:0)
With 040/AGA amigaworld.net alpha_mask is problem, changing value doesn't seems to affect anything? 1 or 100
Probably another bug. It certainly was working on OS4 back before I broke the 8-bit modes on OS4.
-
I would hazard a guess that it's related to NetSurf clearing out the cache. Try disabling the disk cache (disc_cache_size:0)
Disabling disc cache makes it generally faster, even first page loaded loads/render faster.
Difference is not that big, 1.5 seconds and 10 seconds. (amigaworld.net - > amiga.org) Considering that disc cache purpose is make thigs faster that is weird. :) Is this something wich can be fixed or is this just something because slow disc Access?
-
Disabling disc cache makes it generally faster, even first page loaded loads/render faster.
Difference is not that big, 1.5 seconds and 10 seconds. (amigaworld.net - > amiga.org) Considering that disc cache purpose is make thigs faster that is weird. :) Is this something wich can be fixed or is this just something because slow disc Access?
Slow disk access. Because it single tasks you get a pause when it writes out. I did make some changes to push writes to a separate process, but it caused problems as NetSurf isn't designed for that.
-
Would it be possible to make it not to clear disk cache? Or does it matter? I can live with that information that my hard disk contains images from these sites, if it makes future openings faster. Any way still bit a mystery why it takes so long, deleting images shouldn't take that much longer.
-
Would it be possible to make it not to clear disk cache? Or does it matter? I can live with that information that my hard disk contains images from these sites, if it makes future openings faster. Any way still bit a mystery why it takes so long, deleting images shouldn't take that much longer.
It's not clearing the disk cache, it's clearing the memory cache - which involves writing out everything to disk (which is what takes time). If it's slowing down initial page access then you're better off leaving it disabled.
-
OK
Downloading and rendering amiga.org after amigaworld.net takes 94 seconds with disk cache OFF
Amiga.org downloads and render takes about 35 seconds, that mean that Netsurf takes 59 seconds to clear it memory cache, wich sounds a lot? Generally deleting from memory should be much faster than writing to memory???
-
OK
Downloading and rendering amiga.org after amigaworld.net takes 94 seconds with disk cache OFF
Amiga.org downloads and render takes about 35 seconds, that mean that Netsurf takes 59 seconds to clear it memory cache, wich sounds a lot? Generally deleting from memory should be much faster than writing to memory???
Hmm, that doesn't sound right, freeing items from memory should be quick.
I've just tried it here and I'm not seeing this issue.
-
I've put a new build up. Potentially a bit quicker again, but quite likely also completely broken.
-
Eep! Sorry I've been a bit AWOL from this thread. Life got in the way, including now working on replacing the front end on a friend's car that they totalled. Haven't had much time for tinkering. :(
In any case, downloaded this most recent version today. I tried changing the font to XEN, then I tried changing it to XHelvetica (I was getting really tired of looking at Topaz). Didn't seem to make much difference in performance so I took those lines out completely. Looks much nicer with the default fonts, anyway. :)
I also tried changing the redraw_tile_size settings. Tried 10x10 and 200x200, but it seems like the 100x100 values still yield the best balance.
I'm not noticing the alpha issues on amigaworld.net I saw in utri007's screenshots. But then again I don't have AGA, either.
One thing new I did notice was an output window popped up that said "Timeout was reached" when I tried to load Wikipedia. I think this was one of the sites that took like 600+ seconds to load during some previous testing. Is this something new you've added? Is the timeout value adjustable?
Edit: I just tried Wikipedia again. 749 seconds, ouch. Could it be because it redirected my http:// to https:// ?
Is there any chance to get "iconify" working? I'd love to be able to minimize this thing to quickly get to other icons on my Workbench.
Overall performance seems better each time (other than on Wikipedia). Here's a few screenshots; thanks, Chris!
-
I tried loading Facebook since I saw somebody asking about it on the amigaworld.net thread. It let me log in, but wouldn't let me get very far without Javascript. Then it gave me this error on the "mobile" site:
Edit: Dang it. It did actually load the mobile Facebook site, eventually, after I clicked through a dozen error messages or so. But by that point I'd already clicked on the "close" box for Netsurf, too. So wasn't able to get a screenshot before it closed out. :(
-
You don't have any problems when you surf from site to site, amiga.org - > amigaworld.net? Rendering / downloading amiga.org takes 35 seconds and amigaworld.net takes 29 second, but when going first amiga.org after that amigaworld.net downloads / render 104 seconds.
Is netsurf still killing your Network if left open?
-
Tested 21st day version.
A1200 040/AGA
Very unstable. Locks Computer in random point, two times it didn't display anything, some times it loaded start page, once I got part of amiga.org displayed. One time it said please insert volume workbench in any drive and amiga freezed.
No gurus or erros, just freezed amiga.
A1200 060/RTG
Works better no lock ups. But lots of random weirdness. Once I typed amiga.org address after amigaworld.net address, it started loading and fetching, said it ready but page didnt change. One it displayed amiga.org after amigaworld.net quite fast, but it was still loading and fetching some 60 seconds.
Installed Miami to see does it make diffrence to AmiTCP/Genesis, no diffrence. Network dies some point.
This version has some weird issues, something wich seems to be very random.
-
Edit: I just tried Wikipedia again. 749 seconds, ouch. Could it be because it redirected my http:// to https:// ?
I don't know why Wikipedia takes so long. I assume it's the CSS processing but it seems insanely long.
Is there any chance to get "iconify" working? I'd love to be able to minimize this thing to quickly get to other icons on my Workbench.
Umm.. possibly. Are you definitely running NetSurf on the Workbench screen? As the iconify gadget options don't say they are new, so it should work... but only on the WB screen, if NetSurf is on another screen iconify isn't available (for boring reasons).
Very unstable. Locks Computer in random point, two times it didn't display anything, some times it loaded start page, once I got part of amiga.org displayed. One time it said please insert volume workbench in any drive and amiga freezed.
No gurus or erros, just freezed amiga.
When you find the code that is randomly overwriting memory (or re-using freed pointers?) please let me know. I'll give a hint: it's somewhere in the OS3-only code, as there are no such problems on OS4.
Unfortunately debugging such things on OS3 is virtually impossible. Enforcer might give some clues, but as I have no way of figuring out where the offset is pointing to in the source the only way I'm likely to fix this is by accident (or another pair of eyes looking over my source code for silly errors).
btw, it has got worse because I'm allocating some memory in pools now, so the chances of NetSurf overwriting memory NetSurf has allocated is increased.
-
Have to be some where in AGA used only code, as such a thing doesn't happend with RTG.
-
Have to be some where in AGA used only code, as such a thing doesn't happend with RTG.
Try friend_bitmap:0
-
Try friend_bitmap:0
Tested, didn't help.
But I checked it with snoop dos and it reading a helvetica font some hundered times "helvetica read OK". Crash happen when it does those.
Please insert volume workbench also happened once again, when it was reading fonts.
Note: This is only with AGA
-
Tested, didn't help.
But I checked it with snoop dos and it reading a helvetica font some hundered times "helvetica read OK". Crash happen when it does those.
Please insert volume workbench also happened once again, when it was reading fonts.
Note: This is only with AGA
I was getting weirdness until I switched friend bitmaps back off.
I have fixed some other bugs in the meantime though.
-
Been doing some testing on the latest down on my A4000T 060. Its still the same as previous versions. It looks great but is just slow.
What I dont understand is on the version that comes in Amikit under OSX FS-UAE and even on an older E-UAE its lightning fast. Thats even setting my emulation up as an 040. What is different in the build that comes with Amikit and yours? I recorded a quick video and am uploading it to youtube now. Will post the link when it completes.
EDIT:
https://www.youtube.com/watch?v=4ijjYDolchQ
-
I was getting weirdness until I switched friend bitmaps back off.
I have fixed some other bugs in the meantime though.
I turned OFF backgroung and foreground images and it seems to work. So it is images wich make those problems with AGA.
Acill: I have tested arthurs version only with real amigas, there is no speed difference. I didn't know that there is a diffrence under the emulation.
Arthur's RTG version is running top of SDL and AGA version is running Novacoder's AGA SDL hack. It is based to framebuffer version of Netsurf wich is meant to be used debugging and with "no gui" systems.
Chris version is based to ordinary desktop version. He has some problems with GCC version he is using, he can't compile it optimications turned ON.
-
I was getting weirdness until I switched friend bitmaps back off.
I have fixed some other bugs in the meantime though.
I think that I found a reason for some problems.
AmiTCP/Genesis has maximum ram usage set to 256kb and buffers to 8092
I increased ram limit to 1024kb and buffers to 20000. After that I couldn't "kill" my Network. Also surfing from amigaworld.net to amiga.org took only 45 seconds. But surfing back to amigaworld.net took some 100 seconds and repeating it took more than 200 seconds.
Same thing with miami and I'm prety sure that Roadshow has those limits also. Because Oldsmobile_mike uses it and has also suffering Network lockups.
I have bought Roadshow and I spend some time installing it, but I couldn't get DHCP working, so I decided forget it. I'm familiar with Genesis and DHCP would have been one advantage over Genesis.
-
I think that I found a reason for some problems.
AmiTCP/Genesis has maximum ram usage set to 256kb and buffers to 8092
I increased ram limit to 1024kb and buffers to 20000. After that I couldn't "kill" my Network. Also surfing from amigaworld.net to amiga.org took only 45 seconds. But surfing back to amigaworld.net took some 100 seconds and repeating it took more than 200 seconds.
I don't know how those limits relate to anything NetSurf is doing, but you could try changing the following options:
/** Maximum simultaneous active fetchers */
NSOPTION_INTEGER(max_fetchers, 24)
/** Maximum simultaneous active fetchers per host.
* (<=option_max_fetchers else it makes no sense) Note that rfc2616
* section 8.1.4 says that there should be no more than two keepalive
* connections per host. None of the main browsers follow this as it
* slows page fetches down considerably. See
* https://bugzilla.mozilla.org/show_bug.cgi?id=423377#c4
*/
NSOPTION_INTEGER(max_fetchers_per_host, 5)
/** Maximum number of inactive fetchers cached. The total number of
* handles netsurf will therefore have open is this plus
* option_max_fetchers.
*/
NSOPTION_INTEGER(max_cached_fetch_handles, 6)
eg. try:
max_fetchers:12
max_fetchers_per_host:2
max_cached_fetch_handles:3
There are timeout and retry options too which might help. http://git.netsurf-browser.org/netsurf.git/tree/desktop/options.h
-
I run Roadshow with the mediator fast ethernet included setup and have zero issues, other than the same slowdown you describe going from one site and back.
-
AmiTCP log has lots of memory and buffer errors. This problem doesn't exist with AWeb or IBrowse.
Acil: Seems that A4000 users has some difference to A1200 and A2000 users.
Edit :
max_fetchers:12
max_fetchers_per_host:2
max_cached_fetch_handles:3
Didn't help, it does little slowdown, nothing more.
-
I tested this under emulation, most noticeable thing is that there is no Network slow down when surfing. Even login to amiga.org was fast. Got this error after login :
Assertion of condition "perror == parseutils_ok" failed in file "utils/utf8.c", line 135
Problems with amiga tcp/ip software would be logical explanation for that. There is no Network problems even with cycle exact emultion, UAE uses it own Network.
Also there is a big difference compared to Arthur's port, it is indeed much faster under emulation, but not with real amigas.
-
I run Roadshow with the mediator fast ethernet included setup and have zero issues, other than the same slowdown you describe going from one site and back.
I've managed to reproduce this finally, and I think it's because of the GIF anim that sometimes appears on the aw.net header.
Try:
animate_images:0
-
I've managed to reproduce this finally, and I think it's because of the GIF anim that sometimes appears on the aw.net header.
Try:
animate_images:0
Didn't help. Problem has nothong to do with amigaworld.net. It just basicly first page loads fast, second one loads about 2x longer than it should, third about 4x longer etc.
It even happens if I go to amiga.org forum or any other page.
I don't know is this some sort of TCP/IP stack problem or not, but I loaded some pages from local hard drive and there was no slow down when changin page. I add 4x memory to genesis 1024kb and increased send / receive buffres to 16184, that didn't help either. Though no memory / buffer erros anymore and it doesn't die if I left Netsurf open for some time.
-
My NetSurf version compiled with clib2 also gets slower after couple of pages.
Ixemul and libnix versions works normal.
-
I made some testing more. Yesterday I browsed some pages from CU-Amiga CD, no slowdown at all. Yesterday just before I went to sleep, I found this site from another thread here http://www.amigahistory.co.uk
That is hard to find site, it pure HTML, build with tables. No CSS or Java. Prety much like those www sites from CU-Amiga CD. It is displayed right even with iBrowse.
No slow down at all, every page loaded same time 10-34 seconds. Could it be possible make a conclusion that TCP/IP is not a problem?
-
...
Chris - any possibility of reaching out to any of the original IBrowse dev team? I know at least Oliver Roberts was still active in the Amiga community, with his WarpDT's (which are great, btw). I vaguely recall hearing something about him having an update for SSL, as well. Perhaps he could help you with your datatype issues, at least?
...
Hmmm - don't know about Oliver Roberts's work - but I know Norbert Kett used new Versions of OpenSSL in his "Google Drive Handler (http://aminet.net/package/comm/tcp/GoogleDrive_handler_68k)":
The "Google Drive Handler"-readme says:
“... GD handler is based on … OpenSSL v0.9.8zh. This version of OpenSSL contains all security patches until 2015.12.03, but supports only TLS 1.0. Use this handler at Your own risk. If You have stronger CPU (68040/68060/Apollo FPGA), You can try the handler built with the latest, full featured,but slower OpenSSL v1.0.2e. ...”
From Norbert's E-Mail (I had mailed him regarding a possible use of this latest OpenSSL in YAM):
"Openssl 1.0.2e is the latest, full featured SSL library, it requires a fast 68k machine. (68040+)
AmiSSL is a good project, but not updated years ago, and it uses very old openssl. YAM should be recompiled with the latest openssl.
The openssl what i use is integrated into the handler. This is unusable by other apps like Yam.
AmiSSL is a shared lib. As you wrote, Yam uses OpenSSL through AmiSSL. So AmiSSL need to be updated with the latest openssl, but i see small chance for it in the near future. I contacted them to get their openssl, but i did not get reply. So BSzili made fresh openssl ports."
-
...
Oldsmobile_Mike : I don't belive that there is no way that we could never use our 68k amiga for general web browsin, not even if there where CSS capable iBrowse. :( What would make me happy would be possibility to download files, if needed without going to another machine.
...
Hmmm - my registered copy of IBrowse 2.4 can download files perfectly - even without being CSS capable...
:confused:
-
My NetSurf version compiled with clib2 also gets slower after couple of pages.
Ixemul and libnix versions works normal.
sounds like bernd has been right after all, that ixemul has the most effective memory allocator (afair for applications that frequently allocate and release small pieces of memory, probably anything more complex belongs in this category). good that at least libnix works fine. simply dont link against clib2, or what are the crucial advantages of this library?
-
As crazy at it sounds I went and bought an X-Surf 100 card and a Rapid Road module for it. I was using the Mediator with Spider 2 USB and a PCI Fast Ethernet card prior. Since I sold a couple non Amiga related items and had the money in paypal I figured what the heck and want to see if speed tests were any different.
Meditor Fast Ethernet Driver with Roadshow is slightly slower than the X-surf 100 card. Its hardly noticable at all. So its not the hardware slowing things down.
I also have an 060 with my network device optimised for it, so again not really software as far as TCP/IP stack goes. MEdiator driver and the x-surf driver are both optimised to take advantage of Thor's MMU lib package as well, booted with and without it and didnt notice any significant speed differences.
My system is an A4000T with a Radeon 256MB PCI card, csppc accelerator with an 060, I have over 600MB of fastram free when launching browsers.
Aweb, Voyager2, Ibrowse are all extremely fast in comparison when visiting simple HTML based sites. NetSurf still takes between 25 seconds at best to as long as 2 minutes to load a page.
Just for kicks I booted from a partition that still has MiamiDX and it was even slower than Roadshow, but I knew that. I also installed AmigaKits EasyNET that came with my x-surf 100 and it was slightly faster than MiamiDX, but not faster than Roadshow.
On an emulated system that I have setup as close as I can to my real amiga, using a backup of my current drive and WinUAE bsdsocket emulation its WAY faster, so I am at a total loss here.
The only thing I havent tried is to install 4.1FE classic on my A4000T and see how it runs, but thats a totally different version of NetSurf.
-
sounds like bernd has been right after all, that ixemul has the most effective memory allocator (afair for applications that frequently allocate and release small pieces of memory, probably anything more complex belongs in this category). good that at least libnix works fine. simply dont link against clib2, or what are the crucial advantages of this library?
Clib2 version uses less memory than ixemul and libnix,
loads pages faster at the begining. It is stable and debug uses amiga interface.
No external library needed.
-
Sounds like its worth trying to use the clib2 version. Whats preventing that?
-
Clib2 version uses less memory than ixemul and libnix,
loads pages faster at the begining. It is stable and debug uses amiga interface.
No external library needed.
what are actual memory requirements differences between these versions?
if the memory gets used faster with libnix version, then probably the advantage of clib comes at the cost of memory fragmentation, which may be much more a burden for a slow cpu while not noticeable on a faster one.
sounds as if the options are:
1. profile and eventually fix clib (i doubt there is anyone who could immediately do that)
2. use libnix (whats wrong with it, isnt it stable?)
-
I don't know about memory requirements, because clib version doesn't start any of my systems. Other versions requires at least 24 mb ram, they doesn't start my 040/32mb system becuase lack of memory.
Another thing is that Arthur version is very stable and fast with emulated amigas. There is no real speed diffrence with real amigas, execpt Chris version is much more stable.
Memory fragmention would be one logical explanation, but I'm usually running this on my startup-sequence. http://aminet.net/package/util/sys/PoolMem
It doesn't have any effect to Chris' Netsurf decreasing speed, is it running or not. Memory framention can be monitored any way.
-
My 060 has plenty of free ram when surfing amiga.org/amigaworld.net, so memory allocator should be broken, if it fragments memory with them. Another thing is that at least amitcp and miami has limis for ram usage, could it be that that memory is fragemted? AmiTCP has 256kb limit as a default for ram usage. Setting limit 4x higher 1024kb doesn't have any effect to speed decrease.
-
sounds as if the options are:
1. profile and eventually fix clib (i doubt there is anyone who could immediately do that)
If there really is a problem with clib2 then somebody needs to write a test case which shows the problem (and, crucially, build it also with libnix to prove the problem is in clib2), and raise a bug report.
NB: A web browser is not a test case.
-
@chris
correct, but this is unfortunate, since none will write a test case nor submit a bug report. all we can do is to bring it to olsen attention and ask his opinion, while in the meantime staying with libnix.
-
New build up, same place.
The big change is that the prefs window now (mostly) works.
-
Preferences do work. :D One note, it is not actually preferences, but I can disable images, but I can't enable them back on.
You have spend some time with aga? It works/displays right again, but 4/5 times it goes unresbonsive before it it displays anything. This with my 68040/AGA
My RTG amiga has wrong colours, usually it is fixed with friend bitmap, but this time it doesn't have any effect.
About increasing download/fetching/rendering time. Downloading amiga.org takes now 27 seconds, going to forums takes 55 seconds, downloading amiga.org frontpage again tooks 129 seconds.
-
About increasing download/fetching/rendering time. Downloading amiga.org takes now 27 seconds, going to forums takes 55 seconds, downloading amiga.org frontpage again tooks 129 seconds.
Create a log file. This should do it:
NetSurf -V ram:ns.log
With that I have half a chance of figuring out what is causing it. Without I have no way of ever tracking it down.
-
Log file is too big, even in zipped to this forum. Send to your email.
-
Log file is too big, even in zipped to this forum. Send to your email.
Got it, thanks.
The slow bits look like fetching, rendering and (possibly) clearing out the cache.
Try increasing the memory cache size, and setting the disc cache to 0.
I say "possibly", because I'm not entirely sure what NetSurf is doing between these two lines:
(512.243855) content/content.c:690 content_remove_user: content http://amiga.org/forums/images/headers/header_ao_logo.jpg (0x696c1440), user 0x68c15a1a 0x6882b800
(566.575433) content/content.c:386 content_destroy: content 0x6993bc40 http://amiga.org/forums/images/headers/bar_bg.gif
-
Preferences do work. :D One note, it is not actually preferences, but I can disable images, but I can't enable them back on.
I can't see anything wrong with that code. All it does is check the state of the menus and then set the images options so they match.
You have spend some time with aga? It works/displays right again, but 4/5 times it goes unresbonsive before it it displays anything. This with my 68040/AGA
My RTG amiga has wrong colours, usually it is fixed with friend bitmap, but this time it doesn't have any effect.
I've only been testing with AGA, can't help with the wrong colours thing, I'm not set up anywhere to test it. I'd hazard a guess it's to do with the endianness of the screenmode, I've not changed anything related to colours or friend bitmaps AFAIR.
-
@utri007
What about TLSFMem, does it improve speed a little?
-
I can't see anything wrong with that code. All it does is check the state of the menus and then set the images options so they match.
I've only been testing with AGA, can't help with the wrong colours thing, I'm not set up anywhere to test it. I'd hazard a guess it's to do with the endianness of the screenmode, I've not changed anything related to colours or friend bitmaps AFAIR.
Bit off topic :
I reinstalled myt amiga some days ago. I can't remember did colors work after that or not. But I have a 15bit PC, 16BIT PC and 24bit PC screenmodes availlable. I'm using Picasso 96. Do I have setup problem or not, sorry my ignorance.
What about TLSFMem, does it improve speed a little?
Netsurf starts and loads wellcome screen, but locks Computer after 10-15 seconds when trying to surf any other web address, if tlsfmem or tlsfmempool is running.
-
Bit off topic :
I reinstalled myt amiga some days ago. I can't remember did colors work after that or not. But I have a 15bit PC, 16BIT PC and 24bit PC screenmodes availlable. I'm using Picasso 96. Do I have setup problem or not, sorry my ignorance.
The PC modes are little-endian, so that may be the problem. It's likely I have some code which is assuming the screen is big-endian. Does text show up the correct colour? I know for sure that uses pen allocation so I'd expect it to. Bitmap plotting is somewhat more involved.
-
The PC modes are little-endian, so that may be the problem. It's likely I have some code which is assuming the screen is big-endian. Does text show up the correct colour? I know for sure that uses pen allocation so I'd expect it to. Bitmap plotting is somewhat more involved.
I had excatly same hardware and picasso96 earlier also. I wonder did I had 15bit RGB and 16bit RGB modes earlier. Need to figure out why they have disappeared. Any hints for that?
Really can't say anything about text color, at least colors are not obiviously wrong.
-
Tried the latest version. Seems the program calls the picture.datatype quite a few times. But it opens the font even more (hundreds and hundreds of times). Normally these should be loaded in memory just one time, or am i wrong !?!
-
But it opens the font even more (hundreds and hundreds of times). Normally these should be loaded in memory just one time, or am i wrong !?!
It opens the font many times, but the OS will only load it from disk once (unless you're running out of memory and it has flushed it)
-
Checked this with snoopdos, it says Fonts:helvetica.font. Should it be just helvetica.font, if it is preloaded? Any way
On thing wich took a lots of time is Ramlib open ENV:Gadgets/select.prefs
Need to spend some more time with this. :)
-
Checked this with snoopdos, it says Fonts:helvetica.font. Should it be just helvetica.font, if it is preloaded?
Possibly. Are you looking at Open, OpenDiskFont or OpenFont?
-
Possibly. Are you looking at Open, OpenDiskFont or OpenFont?
All font things on snoopdos are Open FONTS:helvetica.font Read
I thought that I found some kind of pattern, but I'm not sure. First time when loading amiga.org there are only some ChangeDir, seems that number of them increases when going to forums and even more when going back to FrontPage.? ChangeDir can take several tens of sceonds.
Weird thing was also that ramlib Load Libs:Gadgets/select.gadget, it should be Classes:Gadgets/select.gadget? Would it be because of netsurf?
-
All font things on snoopdos are Open FONTS:helvetica.font Read
I'll check if I'm calling the full path... Nope, just the filename, so diskfont isn't caching. It shouldn't be opening from disk all the time, only if the size is different (that will still be quite a lot of reads though).
Weird thing was also that ramlib Load Libs:Gadgets/select.gadget, it should be Classes:Gadgets/select.gadget? Would it be because of netsurf?
Classes: is part of the Libs: assign.
I don't know what select.gadget is though!
-
I'm a curious what is changed this new version?
Is it so that aminet release link is wrong, it goes here, it says OS4 version : http://www.netsurf-browser.org/downloads/amiga/
-
I'm a curious what is changed this new version?
Is it so that aminet release link is wrong, it goes here, it says OS4 version : http://www.netsurf-browser.org/downloads/amiga/
Typo, have fixed, thanks.
ChangeLog (http://download.netsurf-browser.org/netsurf/releases/ChangeLog.txt)
Only the last few entries are relevant, so probably something like this:
* Don't try to look up mailto URLs in database, as they aren't stored.
* Optimised HTML layout of floats.
* Fixed HTML layout bug relating to position of floats.
* Fixed bug when user toggled javascript_enabled setting during page load.
* Slight optimisation of DOM node inserted handling.
* Introduced 10s script execution timeout.
* Used new LibCSS API to optimise handling of HTML presentational hints.
* Removed logging during browser window resize.
* Improved cache logging.
* Fixed to avoid generating pseudo elements for HTML replaced elements.
* Simplified status bar updating while fetching HTML contents.
* Fixed bug in accounting of objects being used by HTML contents.
* Stopped GIF animations for when GIF has no users.
Note that page processing is significantly quicker now.
The only thing I did on the frontend was re-enable tabs. It isn't advisable to try to close them though.
Thee's a bug which stops NetSurf quitting properly under certain circumstances. 3.4r2 will fix that (grab it from the ntlworld link if you don't want to wait)
-
New typo :D All the downloads are gone in netsurf page.
-
New typo :D All the downloads are gone in netsurf page.
That file re-uploaded along with something else but never completed. It's there now.
-
Tested.
You really have put some effort to AGA. It is possible that with AGA slow down is gone.
060/RTG is faster when going to amiga.org frontpage, but going to forums takes 119 seconds. With 040/AGA going to forums takes 59 seconds..!
Problem is, that after that AmiTCP is dead both of my amigas (tested several times). GUI says that it is online, but I can't ping gateway, etc. So question is, does Genesis GUI really know state of AMITCP if so it must be just a prism2.device?
Would be really nice to know does it kill anyothers TCP stack, with other real amiga users. There is no slow down or dead TCP in UAE.
-
NetSurf 3.4 Release
* added autodetect_depth option.
* added auto playback of modules from modarchive.org and modules.pl
Supported players Hipoplayer, Eagleplayer, DeliTracker2.
If Hip is slow than switch to AHI.
Set module_player option to player path and
module_autoplay:1 to enable.
* enabled javascript. Added new button to turn on/off
* changed youtube api
* added stop button
* added go back/forward mouse buttons support
* fixed restart action
* fixed 100% CPU usage bug
http://ami-soft.blogspot.com/
-
Im assuming the new "defacto" performance standard in 2016 will be Vampire and 128+ megabytes of ram.
Have you tested Netsurf with Vampire? Ibrowse is quite decent with my 030, but since its not updated, I can see myself switching to the "in development" Netsurf once I get Vampire with my A1200.
-
Vampire doesn't have FPU.
I'm working on no fpu version ATM.
-
NetSurf 3.4 Release
* added autodetect_depth option.
* added auto playback of modules from modarchive.org and modules.pl
Supported players Hipoplayer, Eagleplayer, DeliTracker2.
If Hip is slow than switch to AHI.
Set module_player option to player path and
module_autoplay:1 to enable.
* enabled javascript. Added new button to turn on/off
* changed youtube api
* added stop button
* added go back/forward mouse buttons support
* fixed restart action
* fixed 100% CPU usage bug
http://ami-soft.blogspot.com/
It is important that we do not confuse each other:
This version of Netsurf for OS3 you mentioned was provided by Artur Jarosik and unfortunately is named the same as the version provided by Chris Young. As I understand it, Chris' version has a Reaction-based GUI and uses a native Amiga toolkit, while Jarosik's port prtly requires ixemul - but he also has released a non-ixemul version.
-
It is important that we do not confuse each other:
This version of Netsurf for OS3 you mentioned was provided by Artur Jarosik and unfortunately is named the same as the version provided by Chris Young. As I understand it, Chris' version has a Reaction-based GUI and uses a native Amiga toolkit, while Jarosik's port prtly requires ixemul - but he also has released a non-ixemul version.
Chris' version is based desktop version of Netsurf, it's builds against official source tree. It requires just a 11mb free ram to start, so it is useable with 32mb ram. Chris' version also supports Netsurf themes, like Firefox theme, just to mention one. :)
Arthur's version is based SDL/ Framebuffer version of Netsurf.
-
It is important that we do not confuse each other:
This version of Netsurf for OS3 you mentioned was provided by Artur Jarosik and unfortunately is named the same as the version provided by Chris Young.
Artur's fork should not really be called NetSurf: https://listmaster.pepperfish.net/pipermail/netsurf-dev-netsurf-browser.org/2010-February/001739.html
-
Artur's fork should not really be called NetSurf: https://listmaster.pepperfish.net/pipermail/netsurf-dev-netsurf-browser.org/2010-February/001739.html
chris, few things need to be mentioned. the post you have linked to is very old (six years by now) and i remember that artur was ready and has tried to sastisfy the demands of the developers, as you can judge from the content of this followup post:
https://listmaster.pepperfish.net/pipermail/netsurf-dev-netsurf-browser.org/2010-February/001742.html
i have not been following it but im certain, that if artur was categorically demanded to rename his fork he would do that. meanwhile he was supplying amiga community with a functional browser based on netsurf sdl frontend for years, while you only decided to put some effort into the 68k frontend (as official it might be) just few months ago and i doubt it is in a state to compete in terms of usability with what artur has delivered even by now.
i think its a little unfair to publicly demand of him to rename his port in this situation, as everybody currently associates netsurf on amiga with his work. artur has admired your work, and afair proposed cooperation. i dont know if it is practical or only words, even if from technical point of view it would be better to have unified effort, but the attitude towards him puts me off a little.
-
chris, few things need to be mentioned. the post you have linked to is very old (six years by now) and i remember that artur was ready and has tried to sastisfy the demands of the developers, as you can judge from the content of this followup post:
https://listmaster.pepperfish.net/pipermail/netsurf-dev-netsurf-browser.org/2010-February/001742.html
i have not been following it but im certain, that if artur was categorically demanded to rename his fork he would do that. meanwhile he was supplying amiga community with a functional browser based on netsurf sdl frontend for years, while you only decided to put some effort into the 68k frontend (as official it might be) just few months ago and i doubt it is in a state to compete in terms of usability with what artur has delivered even by now.
i think its a little unfair to publicly demand of him to rename his port in this situation, as everybody currently associates netsurf on amiga with his work. artur has admired your work, and afair proposed cooperation. i dont know if it is practical or only words, even if from technical point of view it would be better to have unified effort, but the attitude towards him puts me off a little.
I apologise, I wasn't demanding, just pointing out that the confusion is justified and shared (and it's not just Artur's version, but there's also a build of the GTK frontend for AmiCygnix on OS4). As you say, that thread did not come to any conclusion.
My OS3 work has been on and off for a year or so, not just a few months. Actually John-Mark did the initial work to get it to build (maybe five years ago? Can't remember now) although that obviously wasn't a working binary.
-
thx, chris. need to check out your version one of these days;)
-
Chris version has some benefits. It uses much less memory, so it is useable with 32mb ram. Surprisingly it is more stable than Arthur version with real Amiga. Even if it freeses, it doesn't lock OS.
I'm not a big fan of emulators, so I was quite surpriced how well Arthur's fork performs under emulation. Considering that with real harware diffrence is not a big.
Main problem is slow down when surfing, every new page takes at lest 2x time than previous. Some builds kills TCP/IP stack at some point, tested with Genesis(AmiTCP), Miami and Roadshow. AmiTCP things that it is online all the time, so it can be a device driver wich dies?
It doesn't seem to be memory fragmention, as I have all the time lots of free ram. biggest continous blog is about 30mb all the time.
It also made some AmiTCP errors about memery usage and buffers. There is preferences for them max memory usage of AmiTCP is just 256kb.
-
Chris should try compiling with libnix . There are no slowdowns in my netsurf-libnix version.
Another example: Running ixemul version with ixemul_poolmem.library results with same slowdowns as clib2.
It is because ixemul uses cache buddy memallocator and clib2 uses amiga mem alloc.
-
as artur says it must be the problem with the library the core gets linked against, i really doubt its the fault with the choosen frontend. on aminet there is a simple tool to compare time spant in a particular library function, very easy to use. its called librarytimer. you only need a valid fd files for the library you want to observe. and of course rtfm.
-
Another example: Running ixemul version with ixemul_poolmem.library results with same slowdowns as clib2.
This is interesting.
I've just put up a test build, can somebody try it and see if the problems have gone away?
-
It doesn't fix problem. It also makes it slower, amiga.org FrontPage 89 seconds, forums 248 seconds. After that timeout and AmiTCP is dead.
-
that doesnt sound like a proper testing. first of all a number of factors need to be eliminated that may be related to other parts of the system, like network stack. how about storing some webpages locally on hd drive and trying to switch between them. does slowdown also occures? because it might be an amitcp bug, or a big in interaction with amitcp, but not particularly netsurf bug.
i still have an impression it is a memory allocation issue, and tha is what i would be obseving, but im a noob.
-
I'd like to see more testing too. The TCP/IP issues could easily be causing slowness not related to this problem.
The memory allocation in this version is just using AllocMem - almost identical to libnix.
I'm not seeing the problem here but then I could never reproduce it consistently anyway.
-
OK I can't test this with 3 different Network stacks. Maybe I could rip some sites, so that I could test without TCP stack. I have already tested this with plain html pages, from CU Amiga cds, no slow down. By the way, this site doesn't have slow down at all http://www.amigahistory.co.uk/ it is also plain html, no css etc.
-
OK I can't test this with 3 different Network stacks. Maybe I could rip some sites, so that I could test without TCP stack. I have already tested this with plain html pages, from CU Amiga cds, no slow down. By the way, this site doesn't have slow down at all http://www.amigahistory.co.uk/ it is also plain html, no css etc.
simply test with pages saved locally. complex or not.
we ae entering the territory where amiga software has been tested very little in praxis. huge files, frequents allocations, it might be and probably is, kernel libs implementation issue, more general tha just a browser. it might be also an issue when compiling on native 68k..
-
Hello guys. I am reading this thread with great interest but I was not able to post earlier as my account didnt allow that until recently.
I dont have Amiga but on Atari platform we have usable 68k binaries for several years now. As I read you have exactly the same probles as we have. At first web pages are rendered fast but later they renders slower until we get "Connection timeut" message and you have to restart NetSurf to be able to browse on. TCP/IP stack on Ataris doesnt hang. What I want to say it seems those problems arent platform specific but as I see it is common problem for most of the platform. Also seems that faster the computer is the problem is harder to spot. From descriptions I read on this forum I see this problem is even more critical on Amiga platform. To me it seems like memory alocation problems = handling the memory. But I am not a coder so my assumptions can be totaly wrong.
I dont know in which state your port of the browser is ... but you can compare it to the Atari 68k version from a year ago I shoot a short video of browsing with NetSrf on my FireBee:
https://www.youtube.com/watch?v=4q6t_Jerv7c
Good luck to your efforts bringing it to Amiga platform!
-
Do you know who is solving those problems? Maybe Chris could save some time to asking what he/she has found.
There are problems to log on to this site, some times it works, some times it doesn't.
-
Hello guys. I am reading this thread with great interest but I was not able to post earlier as my account didnt allow that until recently.
I dont have Amiga but on Atari platform we have usable 68k binaries for several years now. As I read you have exactly the same probles as we have. At first web pages are rendered fast but later they renders slower until we get "Connection timeut" message and you have to restart NetSurf to be able to browse on. TCP/IP stack on Ataris doesnt hang. What I want to say it seems those problems arent platform specific but as I see it is common problem for most of the platform. Also seems that faster the computer is the problem is harder to spot. From descriptions I read on this forum I see this problem is even more critical on Amiga platform. To me it seems like memory alocation problems = handling the memory. But I am not a coder so my assumptions can be totaly wrong.
Weird.
Mintlib doesn't apepar to do anything special wrt memory allocations.
https://github.com/ArmstrongJ/MiNTLib/blob/master/mintlib/malloc.c
The commonality between clib2 and the ixemul_poolmem that Artur mentioned is that both use memory pools.
My newer build uses plain memory allocation (AllocMem) and apparently still has the same problem.
I'm sceptical this is purely a memory alloc/dealloc issue unless it's a fundamental 68k problem.
Maybe people can try a few different CPU models to see if they all have the same problem?
-
it may well be a fundamental 68k problem, or an exec issue. amiga and atari machines werent built with huge multiple megabytes binaries in mind, frequently allocating and deallocating small bits of memeory. i mean to have observed similar issues when for instance trying to compile something larger with cubic ide. with ixemul the problem likely has been masked by using its own linux memory allocation mechanism. on faster machines or uae the inefficiency might be masked by the pure computing power, or jit, even though os4, aros and morphos exec may differ from the genuine one in this respect.
some general profiling would be neccessary, based on a simple testcase, which may result in need to rewrite parts of the system itself to get rid of this bottleneck.
-
Also: utri007, are you still running PoolMem? If so, can you please re-test without it as that would counteract the change I made. (in fact really you shouldn't be running any patches *at all* when testing things)
-
Poolmem has been removed from startup-sequence when we first talk abou it.
-
Spend some time testing.
I started to configuring Roadshow.
Good news is that RoadShow doesn't die, even if Netsurf gives time out error. I can't still ping gw and mount a samba share. I could test longer period of time, so this was more meaning full than testing with Genesis/AmiTCP
My 040/AGA started to give time out errors earlier than my 060/RTG
About slow down effect. This version is slow as snail, but I could surf more with RoadShow than with AmiTCP. Memory allocator migh have a reason for slow down, downloading amiga.org forums/frontpage took about same amount of time every time. I this version renders page about 3x longer than prevoius version. So "normal" time to render amiga.org front page is about 90 seconds, forums renders about 140 seconds and I could test that about 6 times before I got time out error.
This version has some random weirdness. Suddenly it just started to display pictures with wrong colors. Opening own screen, like it would not read settings, no matter what put choises file or even if it is empty. Or what you have put to defaults??
AGA seems to be faster than RTG? Everything tooks about same amout of time wit 040 / AGA than 060 /RTG
Difficult to trying explain this english. Hope you understand.
-
About slow down effect. This version is slow as snail, but I could surf more with RoadShow than with AmiTCP. Memory allocator migh have a reason for slow down, downloading amiga.org forums/frontpage took about same amount of time every time. I this version renders page about 3x longer than prevoius version. So "normal" time to render amiga.org front page is about 90 seconds, forums renders about 140 seconds and I could test that about 6 times before I got time out error.
OK, this is more like I would expect - everything running consistently at a slower speed.
I know exactly what is causing the slowdown now. clib2 uses memory pools with a puddle size and expected allocation size of 4K. I modified that in the newer build to use normal memory allocations instead.
What is happening, is that early memory allocations are fast and efficiently allocated in 4K chunks. Then, when bits of memory is de-allocated it leaves holes. When new memory blocks are allocated it - and this is where I'm not sure of the implementation details in the OS - is trying to fill in the gaps in the already-allocated pools? With a lot of pools it may be taking some time to search through and find a gap of the correct size, which is similar to how normal memory allocations work when searching through all of RAM (and thus a similar speed).
Quite simply, we are allocating and de-allocating so much memory that we quickly lose any advantage of memory pools.
To fix it... well, that's tricky. The correct way would be to pool together elements of the same size to avoid fragmentation, but I can't do that in the core and all libraries without re-writing all the memory allocations (which would definitely not be popular). Note I already do this in the frontend everywhere it is practical (this was one of my earlier OS3 optimisation attempts!)
It may simply be a case of making the memory pools bigger, and I will try that first.
This version has some random weirdness. Suddenly it just started to display pictures with wrong colors. Opening own screen, like it would not read settings, no matter what put choises file or even if it is empty. Or what you have put to defaults??
The default is to open on its own screen.
use_screen:Workbench will open it on the Workbench screen, or play around with the unnamed radio buttons on the Display tab in prefs until the two fields are ghosted.
AGA seems to be faster than RTG? Everything tooks about same amout of time wit 040 / AGA than 060 /RTG
I don't do anything special for RTG so it is unlikely to be faster.
-
OK, try the new build.
I'm interested in whether it is any faster, whether any slowdown occurs (and if so, how soon and how much), and whether it uses more memory.
-
I'll spend this evening with beer and friends, so tomorrow evening is next test time for me.
-
Am I correct in assuming this version requires a FPU?
Edit: I don't have one and simply get the error "8000000B" when trying to start it. (Amiga 1200 / ACA1231 / 3.1 ROMs / OS3.9BB4)
-
OK, this is more like I would expect - everything running consistently at a slower speed.
I know exactly what is causing the slowdown now. clib2 uses memory pools with a puddle size and expected allocation size of 4K. I modified that in the newer build to use normal memory allocations instead.
What is happening, is that early memory allocations are fast and efficiently allocated in 4K chunks. Then, when bits of memory is de-allocated it leaves holes. When new memory blocks are allocated it - and this is where I'm not sure of the implementation details in the OS - is trying to fill in the gaps in the already-allocated pools? With a lot of pools it may be taking some time to search through and find a gap of the correct size, which is similar to how normal memory allocations work when searching through all of RAM (and thus a similar speed).
Quite simply, we are allocating and de-allocating so much memory that we quickly lose any advantage of memory pools.
To fix it... well, that's tricky. The correct way would be to pool together elements of the same size to avoid fragmentation, but I can't do that in the core and all libraries without re-writing all the memory allocations (which would definitely not be popular). Note I already do this in the frontend everywhere it is practical (this was one of my earlier OS3 optimisation attempts!)
It may simply be a case of making the memory pools bigger, and I will try that first.
its pretty much as what i imagine to happen. okay, i would expect the system should take care of it, but perhaps its woth to try if it can be tweaked in the application code.
-
Am I correct in assuming this version requires a FPU?
Edit: I don't have one and simply get the error "8000000B" when trying to start it. (Amiga 1200 / ACA1231 / 3.1 ROMs / OS3.9BB4)
It *shouldn't*, certainly early versions were working without FPU but recent ones I can't run either (although I've not tried for a while).
It's built with soft float, but maybe not everywhere.
-
Chris version should work every amiga
with aga and enoug memory.
-
Fair, oddly if I launch from the CLI it hangs rather than crashes. Specifically it hangs here (at my 64 color AGA screen being detected):
global_history_init: Loaded global history
(193.284900) content/llcache.c:1355 llcache_process_metadata: Retriving metadata
(193.413112) content/fs_backing_store.c:889 get_store_entry: url:http://www.google.com/favicon.ico
(193.570131) content/fs_backing_store.c:897 get_store_entry: Failed to find ident 0x634ae90c in index
(193.727449) content/fs_backing_store.c:1923 fetch: entry not found
(193.839826) content/fetchers/curl.c:280 fetch_curl_setup: fetch 0x8e6d558, url 'http://www.google.com/favicon.ico'
(194.015665) amiga/font.c:54 ami_font_setdevicedpi: WARNING: Using diskfont.library for text. Forcing DPI to 72.
(194.182763) amiga/plotters.c:102 ami_init_layers: Screen depth = 6
Also, for the installer script to run I have to rename the readme to get rid of the "_os3" line. Thank you so much for all your efforts!
-
@Chris
I think I've found why my libnix version is not getting clib2 slowdown.
I've based it on libnix 3.0 by Diego Casorran who made some improvements in memory operations code.
Changes between libnix 2.1 and 3.0:
* stdlib/realloc.c: Changed CopyMem() to bcopy()
* string/memchr.c: Speed Improvement
* string/memcmp.c: Speed Improvement
* string/memcpy.c: Replaced CopyMem() by internal bcopy()
* string/memset.c: Optimized small operations
-
Fair, oddly if I launch from the CLI it hangs rather than crashes. Specifically it hangs here (at my 64 color AGA screen being detected):
global_history_init: Loaded global history
(193.284900) content/llcache.c:1355 llcache_process_metadata: Retriving metadata
(193.413112) content/fs_backing_store.c:889 get_store_entry: url:http://www.google.com/favicon.ico
(193.570131) content/fs_backing_store.c:897 get_store_entry: Failed to find ident 0x634ae90c in index
(193.727449) content/fs_backing_store.c:1923 fetch: entry not found
(193.839826) content/fetchers/curl.c:280 fetch_curl_setup: fetch 0x8e6d558, url 'http://www.google.com/favicon.ico'
(194.015665) amiga/font.c:54 ami_font_setdevicedpi: WARNING: Using diskfont.library for text. Forcing DPI to 72.
(194.182763) amiga/plotters.c:102 ami_init_layers: Screen depth = 6
I might have to add some extra debug in to track this down, but firstly are you using the version from my test link (ntlworld) or Aminet? If the latter, please try the test version instead.
Secondly, try opening your Choices file (in Users/your-username) and adding:
friend_bitmap:1
Or change the value to 0 if it is already present.
edit is this the 8000000b error?
Also, for the installer script to run I have to rename the readme to get rid of the "_os3" line. Thank you so much for all your efforts!
Thanks, fixed.
-
OK, this is more like I would expect - everything running consistently at a slower speed.
I know exactly what is causing the slowdown now. clib2 uses memory pools with a puddle size and expected allocation size of 4K. I modified that in the newer build to use normal memory allocations instead.
What is happening, is that early memory allocations are fast and efficiently allocated in 4K chunks. Then, when bits of memory is de-allocated it leaves holes. When new memory blocks are allocated it - and this is where I'm not sure of the implementation details in the OS - is trying to fill in the gaps in the already-allocated pools? With a lot of pools it may be taking some time to search through and find a gap of the correct size, which is similar to how normal memory allocations work when searching through all of RAM (and thus a similar speed).
Quite simply, we are allocating and de-allocating so much memory that we quickly lose any advantage of memory pools.
To fix it... well, that's tricky. The correct way would be to pool together elements of the same size to avoid fragmentation, but I can't do that in the core and all libraries without re-writing all the memory allocations (which would definitely not be popular). Note I already do this in the frontend everywhere it is practical (this was one of my earlier OS3 optimisation attempts!)
It may simply be a case of making the memory pools bigger, and I will try that first.
I suspect that this may not make much of a difference. The memory pools, which is what the malloc()/alloca()/realloc()/free() functions in clib2 are built upon, were intended to avoid fragmenting main memory. This is accomplished by having all allocations smaller than the preset puddle size draw from a puddle that still has enough room left for it to fit. Fragmentation happens inside that puddle.
The problems begin when the degree of fragmentation inside these puddles becomes so high that the only recourse is to allocate more puddles and allocate memory from that. The number of puddles in use increases over time, and when you try to allocate more memory, the operating system has to first find a puddle that still has room and then try to make the allocation work. Both these operations take more time the more puddles are in play, and the higher the fragmentation within these puddles is. Allocating memory will scale poorly, and what goes for allocations also goes for deallocations.
The other problem is with memory allocations whose length exceeds the puddle size. These allocations will be drawn from main memory rather than from the puddles. This will likely increase main memory fragmentation somewhat, but the same problems that exist with the puddles apply to main memory, too: searching for a chunk to draw the allocation from takes time, and the same goes when deallocating that chunk. There's an additional burden on this procedure because the memory pool has to keep track of that "larger than puddle size" allocation, too.
Because all the memory chunk/puddle, etc. allocations and deallocations use the humble doubly-linked Exec list as its fundamental data structure, the amount of time spent finding the right memory chunk, and putting the fragments back together, scales poorly. Does this sound familiar?
From the clib2 side I'm afraid that the library can only leverage what the operating system provides, and that is not well-suited for applications which have to juggle large number of allocated memory fragments.
Question is what size of memory chunk is common for NetSurf, how many chunks are in play, how large they are. If you have not yet implemented it, you might want to add a memory allocation debugging layer and collect statistics for it over time.
It may be worth investigating how the NetSurf memory allocations could be handled by an application-specific, custom memory allocator that sits on top of what malloc()/alloca()/realloc()/free() can provide and which should offer better scalability.
-
Question is what size of memory chunk is common for NetSurf, how many chunks are in play, how large they are. If you have not yet implemented it, you might want to add a memory allocation debugging layer and collect statistics for it over time.
It may be worth investigating how the NetSurf memory allocations could be handled by an application-specific, custom memory allocator that sits on top of what malloc()/alloca()/realloc()/free() can provide and which should offer better scalability.
I would just take a shortcut and install TSLFmem: http://dump.platon42.de/files/
Other than that, there is no really solution. Designing good memory allocator is an art of its own where one has to consider memory fragmentation, allocation performance and deallocation performance.
Yeah, I read from previous page that with TLSFmem Netsurf is crashing but this is very likely due to internal memory trash somewhere in Netsurf... with good old memory lists and standard memory pools buffer under/overflows often go unnoticed but with TLSF you are likely going to crash right away.
Of course, Wipeout session could reveal this albeit it is going to be painfully slow experience with such a complex application.
-
This time when launching from the CLI it gave me "8100 0005" which is a corrupted list in Freemem and then my machine hard rebooted. I am using the latest & greatest from your site. At this point I'm willing to cede there's maybe an odd library incompatibility? Something making my memory management not work quite right. My particular 1230 card, while lacking a FPU, does have a MMU.
I'd be happy to run a debug mode, but don't have a cross-compiler set up. Any programming I do is in Python or PHP, which says exactly how much I like compilers.
Edit: this is what I mean about the corrupted list http://eab.abime.net/showthread.php?t=64159
Edit Edit: Okay, I'm getting inconsistent and different freezes/crashes depending on how I launch this, so I'm going to probably try on a clean OS install next. Still get the original 8000000B error when trying the regular double-click launch instead of launching from the CLI, which makes no sense as I get different errors or a hang there.
-
@Olsen
Lists, this remind me of this : https://isocpp.org/blog/2014/06/stroustrup-lists
Kamelito
-
Edit Edit: Okay, I'm getting inconsistent and different freezes/crashes depending on how I launch this, so I'm going to probably try on a clean OS install next. Still get the original 8000000B error when trying the regular double-click launch instead of launching from the CLI, which makes no sense as I get different errors or a hang there.
I'm confident that there is some memory trashing happening in NetSurf, I just wish I knew how to figure out where.
There is a slightly different code path depending on whether it is started from WB or CLI which might account for the differences in how it crashes. The trashing must happen before the branch which narrows it down a bit, but not much.
I can add some more logging which might help see which function is crashing, and running something like Enforcer might help (although I have no way of interpreting any output).
-
I suspect that this may not make much of a difference. The memory pools, which is what the malloc()/alloca()/realloc()/free() functions in clib2 are built upon, were intended to avoid fragmenting main memory. This is accomplished by having all allocations smaller than the preset puddle size draw from a puddle that still has enough room left for it to fit. Fragmentation happens inside that puddle.
Hi Olaf, thanks for commenting. I increased the puddle size to 16K from the default 4K and it seems to have helped in my limited testing (no feedback yet from anybody else). I figure this reduces the number of puddles in the list which need to be searched through, as well as allowing larger allocations into the pool.
From the clib2 side I'm afraid that the library can only leverage what the operating system provides, and that is not well-suited for applications which have to juggle large number of allocated memory fragments.
Which is exactly what you get in a web browser, with lots of memory being allocated for one page, and then for the next, and some of the old memory being deallocated...
Question is what size of memory chunk is common for NetSurf, how many chunks are in play, how large they are. If you have not yet implemented it, you might want to add a memory allocation debugging layer and collect statistics for it over time.
I haven't. I think there are cache statistics in the log which might offer some clues though. Beyond the cache data everything else is structures which should all be very small.
-
I suspect that this may not make much of a difference. The memory pools, which is what the malloc()/alloca()/realloc()/free() functions in clib2 are built upon, were intended to avoid fragmenting main memory. This is accomplished by having all allocations smaller than the preset puddle size draw from a puddle that still has enough room left for it to fit. Fragmentation happens inside that puddle.
The problems begin when the degree of fragmentation inside these puddles becomes so high that the only recourse is to allocate more puddles and allocate memory from that. The number of puddles in use increases over time, and when you try to allocate more memory, the operating system has to first find a puddle that still has room and then try to make the allocation work. Both these operations take more time the more puddles are in play, and the higher the fragmentation within these puddles is. Allocating memory will scale poorly, and what goes for allocations also goes for deallocations.
The other problem is with memory allocations whose length exceeds the puddle size. These allocations will be drawn from main memory rather than from the puddles. This will likely increase main memory fragmentation somewhat, but the same problems that exist with the puddles apply to main memory, too: searching for a chunk to draw the allocation from takes time, and the same goes when deallocating that chunk. There's an additional burden on this procedure because the memory pool has to keep track of that "larger than puddle size" allocation, too.
Because all the memory chunk/puddle, etc. allocations and deallocations use the humble doubly-linked Exec list as its fundamental data structure, the amount of time spent finding the right memory chunk, and putting the fragments back together, scales poorly. Does this sound familiar?
From the clib2 side I'm afraid that the library can only leverage what the operating system provides, and that is not well-suited for applications which have to juggle large number of allocated memory fragments.
Question is what size of memory chunk is common for NetSurf, how many chunks are in play, how large they are. If you have not yet implemented it, you might want to add a memory allocation debugging layer and collect statistics for it over time.
It may be worth investigating how the NetSurf memory allocations could be handled by an application-specific, custom memory allocator that sits on top of what malloc()/alloca()/realloc()/free() can provide and which should offer better scalability.
thanks for confirmation olaf. this is ecxactly what i had in mind. however designing application specific memory allocator oer allocation method is probably not the right way..
-
Leap day treat: I've managed to get NetSurf compiling with optimisations enabled!
It doesn't provide much noticeable speed-up (most of the heavy processing happens in the libraries which were already built with optimisations on) and Javascript still doesn't work (fatal error on launch).
Usual place, built with optimisations but without Javascript, for testing.
-
It doesn't provide much noticeable speed-up (most of the heavy processing happens in the libraries which were already built with optimisations on)
as mentined also, the compiler optimizations usually are a bit overrated, there isnt that much to expect of it, maybe 20-30% at most.
-
I would just take a shortcut and install TSLFmem: http://dump.platon42.de/files/
Other than that, there is no really solution. Designing good memory allocator is an art of its own where one has to consider memory fragmentation, allocation performance and deallocation performance.
All true... but given the limited scalability of the built-in AmigaOS 68k memory management system, it's worth a try looking for alternatives. What's the worst that could happen?
Yeah, I read from previous page that with TLSFmem Netsurf is crashing but this is very likely due to internal memory trash somewhere in Netsurf... with good old memory lists and standard memory pools buffer under/overflows often go unnoticed but with TLSF you are likely going to crash right away.
Of course, Wipeout session could reveal this albeit it is going to be painfully slow experience with such a complex application.
One could do what Wipeout does directly within the application itself. There exist wrappers for the entire malloc()/alloca()/realloc()/free() family which hook into a tracking system which verifies consistency and also keeps score of where the allocations were made from (with file name and line number), etc. You could even build your own system if you wanted to (I did with Wipeout - this isn't so hard, really).
Wipeout is several layers removed from the application performing the memory allocations and will provide little information to go on. It's a system-wide debugging aid, whereas here you would be better served by an application-specific solution.
Update: I looked around a bit and found the Boehm garbage collector (http://hboehm.info/gc/) which can be used as a drop-in for the malloc(), etc. family. Might be worth a shot. Debug wrappers for the malloc(), etc. family exist in several flavours, e.g. http://dmalloc.com or http://www.hexco.de/rmdebug/. Some assembly may be required.
-
Tested
Install script doesn't work, it says it requires OS4 system.
First this one seems to be fast, with my 060 it loads amiga.org front page 18sec and with 040 takes 25 sec.
Sounds very useable at this point, but ...
Second page (amiga.org forums) loads also fast, but third page (amiga.org front page) loads 88 seconds and fourth page (forums) gives time out and this time RoadShow dies.
Just to note, when I installed / configured roadshow I put 8x more buffres than default. Default is 32, now there is 256 buffers. I did that test AmiTCP also, but I ques current install has only 32 buffers. This made noticeable diffrence with AmiTCP.
There is something wrong in my users/choises file on my 060 amiga. Netsurf doesn't load it and displays wrong colors, need to test that more.
-
@Olsen
Lists, this remind me of this : https://isocpp.org/blog/2014/06/stroustrup-lists
Kamelito
As always, he has a point :)
The memory management system in AmigaOS 68k is, in my opinion, adequate for the type of machine it was intended for when the system was designed. It's fast and managing free space is efficient, too, for small amounts of memory (say, 256 KBytes to tens of MBytes). Because it is built around doubly-linked lists it's easy to understand how it works, which is a big bonus in this field. These properties seem to be a good match for Stroustrup's view on the matter.
Unfortunately, a simple and robust memory management system is still simple and while it may have held up well for more than 30 years in its domain, the design could not anticipate the needs of a web browser, which is likely one of the most complex systems ever designed ;)
-
Hi Olaf, thanks for commenting. I increased the puddle size to 16K from the default 4K and it seems to have helped in my limited testing (no feedback yet from anybody else). I figure this reduces the number of puddles in the list which need to be searched through, as well as allowing larger allocations into the pool.
It might help, but you will have to put up with the limitations of the list-based memory management system if you stick with the pools.
As I mentioned before, you might want to cast a bigger net for alternative memory management solutions. No matter how much you tweak the puddle size, there will be side-effects which are just as strange as the ones you hope to get under better control (fragmentation, number of items in each list that need to be checked, etc.). Hence, I would suggest you look into finding a memory management system which can be tuned to the needs of the application.
I haven't. I think there are cache statistics in the log which might offer some clues though. Beyond the cache data everything else is structures which should all be very small.
Well, you can't get a good handle on the memory problem until you have enough data to permit you to make a decision on how to proceed.
For example, you could collect information on the average sizes of allocations made and then group them. Let's say you have one group of 128 bytes or less, one of 256 bytes or less, up to 4KBytes or less and then everything else. You could set up different pools from which only these specific allocation sizes would be drawn. For best effect match the puddle sizes to the most frequently used allocation sizes.
Another idea would be to recycle memory allocations once they are freed. Again, you would group allocations by size (128 bytes or less, 256 bytes or less, etc.) and once an allocation is freed, you'd stick it into a list of chunks of the same size available for reuse. If an allocation is made which matches the chunk size of an entry in the list, you'd pick up the first entry and reuse it. That saves effort because you don't have to merge allocations back into bigger chunks upon freeing them, and because you don't have to search for allocations inside the fragmented puddles: you just pick up the first entry in the respective list, remove it from the list and the use it. Mind you, you will have to watch how much memory is tied up in these lists and prune them to avoid running out of memory over time.
-
thanks for confirmation olaf. this is ecxactly what i had in mind. however designing application specific memory allocator oer allocation method is probably not the right way..
Sometimes it's the only option to regain control over how the application uses the memory if how it is used is a poor match for the abilities of the operating system.
An extreme example may be the BSD-Unix derived TCP/IP stacks for the Amiga which all retain the source kernel's memory management system, which is fused with the TCP segment reassembly logic. This code sits atop the Amiga memory management system, where it emulates the 2K memory pages of the Unix system it was originally designed for.
I have used simpler custom memory management code in my own server applications, e.g. by directly recycling freed memory chunks instead of delegating the work to the operating system and the 'C' link library that sits on top of it. I helped to keep long term fragmentation of memory at bay.
All of these are perfectly legitimate solutions :)
-
Made testing with WinUAE, results were quite impressive.
First, it starts but freeses if CPU is not set to 68040, no matter is FPU enabled or not.
It is very fast and no slow down at least noticeable. All the pages loads less than 5 seconds.
I got one error after I log on to this site and cliked quick reply button :
Assertion of condition
"perror == PARSEUTIL_OK"
failed in line "utils/utf8.c", line 135.
-
More testing with real amiga.
user/choises file is not read anymore. So I can't enable friend_bitmap.
Amigahistory.co.uk doesn't cause any slow down, Spend quite many minits surfing that site, every page was displayed at lest 3 times. It is plain html page.
-
It might help, but you will have to put up with the limitations of the list-based memory management system if you stick with the pools.
Yes... I don't really want to write my own memory management though!
-
Install script doesn't work, it says it requires OS4 system.
Make sure you don't have a file called netsurf.readme in the install source (or netsurf.readme~, which may well confuse the pattern matching)
Made testing with WinUAE, results were quite impressive.
First, it starts but freeses if CPU is not set to 68040, no matter is FPU enabled or not.
Works on an '030 here. The freeze is probably that damn memory corruption bug.
I got one error after I log on to this site and cliked quick reply button :
Assertion of condition
"perror == PARSEUTIL_OK"
failed in line "utils/utf8.c", line 135.
I've added some code to try to protect against that, however I saw a similar assert in some random code when I tried to launch NetSurf from the Shell. It went away if I increased the stack first, so you might want to check the stack is high enough (the stack cookie might be getting optimised away; didn't check). Try at least 100K.
user/choises file is not read anymore. So I can't enable friend_bitmap.
It definitely does read the file. Note it's called "Choices" and there's an extra level under Users.
-
ShK got these problems.
http://s7.postimg.org/3p3mrss3v/003.png
Installer doesn't work
http://s7.postimg.org/434yrec7f/004.png
Requires usergroup.library
His install is clean OS3.9 with Genesis
-
Yes... I don't really want to write my own memory management though!
Need a hand? :)
-
Spend some time to figuring out why my choices file is not readed. Location has changed, now it is users/default/choices
I have just reinstalled it to top of previous versions.
I also tried to run TLSFmem and it doesn't crash Netsurf anymore.
It has some effect to browsing with Netsurf, but it is not a solution. It seems to make Netsurf faster generally and slow down effect comes latter.
-
Installer doesn't work
Installer reads a different version number to what I thought it did.. fixed.
Requires usergroup.library
Provided by the TCP/IP stack. You can just nick the one from the AmiTCP archive though.
-
Need a hand? :)
If you want to help, then yes please!
-
Does this still require OS3.5+ or would Cloanto's "3.X" plus a a TCP/IP stack like Roadshow be enough to launch? Still curious to figure out if I was having a software or hardware issue and debating between a clean install 3.X + Roadshow or OS3.9 + BB4 + Genesis.
-
Does this still require OS3.5+ or would Cloanto's "3.X" plus a a TCP/IP stack like Roadshow be enough to launch? Still curious to figure out if I was having a software or hardware issue and debating between a clean install 3.X + Roadshow or OS3.9 + BB4 + Genesis.
3.5 or 3.9. I don't know what 3.X contains but I suspect it doesn't have all the v44+ Reaction classes that NetSurf needs.
-
@chris:
if reaction/classsact classes is all your frontend needs i can test it under aros. for instance aweb woked there. but im not sure if this is helpful, since we get additional unknown varibles. i dont have a clean 3.5 or 3.9 setup at hand and have no time to set one up, unfortunatelly. however testing without being able to build locally may be not helpful again. and im not sure when i have time to ste up netsurf build system. hmm. maybe ill see if it runs at all.
-
I tested this in Amiga Forever's 3.X release and, other than having to grab a copy of usergroup.library it seemed to run just great. Going to see about testing that on my 1200.
-
I tested this in Amiga Forever's 3.X release and, other than having to grab a copy of usergroup.library it seemed to run just great. Going to see about testing that on my 1200.
Good to know :)
-
I've uploaded a new version with the buttons fixed. I tried rebuilding gcc to default to no FPU. I'm not convinced it worked, but NetSurf seems a lot more stable now on open/close. Maybe the 8000000B error is fixed?
-
Tested with WinUAE. Got whole bunch of weird errors. First I tested with 030 witout FPU. Goint to amiga.org made a guru, tested three times. Changed to cpu to 040, whole UAE crashed. Think that maybe I should document these and couldn't get any of them preproduced. Going to eab.abime.net made one recoverable alert and after that OS freezes. Restarted UAE. Tried to reproduce, didn't happen, but starting Netsurf after that made some weird errors more. It said Please instert Volume System and ENV?
Seems very random?
I will test with real amigas tomorrow or latter today.
-
I did libc comparison tests :
bbc.co.uk/news reload 10 times
libnix clib2 ixemul
3.5 1.5 2.4
2.3 1.1 1.5
2.7 0.7 1.4
2.4 0.7 1.4
3.0 0.8 1.5
3.1 14 1.4
2.6 9.8 1.4
2.9 13 1.4
3.1 10 1.4
2.7 3.3 1.5
(in seconds)
Libnix is slower than ixemul but doesn't get 10x slowdown as clib2.
Ixemul is the fastest.
Got working no-fpu version too.
-
Tested with real amigas.
040/AGA is some 20-30% slower than it used to be. Slowdown problem is still there.
060/RTG is only slightly slower than it used to be. Nothing when compared to 040.
Forget to say, buttons does work. :)
Tested with TLSFmem and noticed that when slowdown happen, might be just luck. Amiga.org "fetching" time was five first times about 18.5 - 43 seconds and not that order. Untill it decided to spend some 230 seconds to render amiga.org's home page. Next time is displayed page quite fast, but it was still "fetching" some hundred seconds, nothing visible was happening, untill I press stop button.
Nothing weird didn't happen, no erros etc.
-
tested with locally saved page?
if not: all kinds of things can happen online at random to make testing like this of no value. such additional factors need to be eliminated
-
Tested with real amigas.
040/AGA is some 20-30% slower than it used to be. Slowdown problem is still there.
060/RTG is only slightly slower than it used to be. Nothing when compared to 040.
Forget to say, buttons does work. :)
Tested with TLSFmem and noticed that when slowdown happen, might be just luck. Amiga.org "fetching" time was five first times about 18.5 - 43 seconds and not that order. Untill it decided to spend some 230 seconds to render amiga.org's home page. Next time is displayed page quite fast, but it was still "fetching" some hundred seconds, nothing visible was happening, untill I press stop button.
Nothing weird didn't happen, no erros etc.
tlsf memory is too a bad memory allocator, because it is best fit. this give lots of cache misses, so all slow down. tlsf memory is more for realtime dsp systems which use no caches and it is important to have constant allocate time and no mem fragmentation
i know years ago, that ixemul is the fastest lots faster libnix and clib2. so i furtherdevelop ixemul. many bash against ixemul. thanks to the work of Arti we all know the confirm now. On my system too ixemul is lots faster
-
@Arti. You write in upper post got working non fpu version too. can you post time results, to see how much slower it is ?
-
tlsf memory is too a bad memory allocator, because it is best fit. this give lots of cache misses, so all slow down. tlsf memory is more for realtime dsp systems which use no caches and it is important to have constant allocate time and no mem fragmentation
i know years ago, that ixemul is the fastest lots faster libnix and clib2. so i furtherdevelop ixemul. many bash against ixemul. thanks to the work of Arti we all know the confirm now. On my system too ixemul is lots faster
Ixemul has better memory allocator but ixemul.library is not safe from application using native Amiga API.
-
Ixemul has better memory allocator but ixemul.library is not safe from application using native Amiga API.
i have never understood whats not safe about it..
howener libnix results are nit that bad in comparison. one simply needs to stay away from clib2.
@bernd
you still here?
-
@Arti. You write in upper post got working non fpu version too. can you post time results, to see how much slower it is ?
Welcome back Bernd! :)
-
Ixemul has better memory allocator but ixemul.library is not safe from application using native Amiga API.
can you more precise explain what go wrong ? sdl lib use native AOS functions too. so if it is not safe then it is not possible that sdl open a amiga window or amiga screen or use amiga os semaphores or other aos functions.
since version 47(i do not) ixemul use aos semaphores. so its possible to use aos and ixemul together for semaphore protect too
"Added low-level mutex (semaphore) functions: ix_mutex_lock(),
ix_mutex_attempt_lock() and ix_mutex_unlock(). These functions are
wrappers for the AmigaOS or p.OS kernel semaphore functions. See ix.h for
more info. "
i fix all problems, mainly thread safe problems, when more tasks of 1 program use ixemul functions.
it work now over years ok. netsurf is mix of aos function and unix functions.
only small limit is, that it is not possible to create a new task with the aos function, because for ixemul tasks need additional structure. but it crash if not do correct. as far i remember i add function IX_createtask or something that make it easy. have same parameter as aos function but create the structure need for unix functions
here is sdl code that create a new task using AOS functions
int SDL_SYS_CreateThread(SDL_Thread *thread, void *args)
{
/* Create the thread and go! */
char buffer[20];
#ifdef WARPOS
struct TagItem tags[6];
#endif
#ifdef SHARED_LIB
extern void *myLibPtr;
((thread_args *)args)->LibBase= myLibPtr;
D(bug("CreateThread: librarybase %lx\n",myLibPtr));
#endif
D(bug("Sending %lx to the new thread...\n",args));
sprintf(buffer,"%ld",args);
#ifdef WARPOS
tags[0].ti_Tag = TASKATTR_CODE; tags[0].ti_Data = (ULONG)RunThread;
tags[1].ti_Tag = TASKATTR_NAME; tags[1].ti_Data = (ULONG)"SDL subtask";
tags[2].ti_Tag = TASKATTR_STACKSIZE; tags[2].ti_Data = 100000;
tags[3].ti_Tag = (args ? TASKATTR_R3 : TAG_IGNORE); tags[3].ti_Data = (ULONG)args;
tags[4].ti_Tag = TASKATTR_INHERITR2; tags[4].ti_Data = TRUE;
tags[5].ti_Tag = TAG_DONE; tags[5].ti_Data = 0;
thread->handle=CreateTaskPPC(tags);
#else
unsigned int stack;
struct Task *t;
long *pid;
thread->handle=(struct Task *)CreateNewProcTags(
/*NP_Output, Output(),
NP_Input, Input(),*/
NP_Name, (ULONG)"SDL subtask",
/*NP_CloseOutput, FALSE,
NP_CloseInput, FALSE,*/
NP_StackSize, 100000,
NP_Entry, (ULONG)RunThread,
NP_Cli, TRUE,
NP_Arguments, (ULONG)buffer,
TAG_DONE);
/*Delay(1); // its a testhack dont activate
Disable();
struct Task *t;
struct Task *t2;
APTR *data;
t = FindTask(0);
data = t->tc_UserData;
t2 = thread->handle;
t2->tc_UserData = data;
Enable();*/
#endif
if(!thread->handle)
{
SDL_SetError("Not enough resources to create thread");
return(-1);
}
return(0);
}
-
I look in history and there are add many functions that increase compatibility to amiga OS. It is possible to switch from a program to amiga path mode.
* Add ix_UseAmigaPaths(mode) func.
When mode = 1 then amiga Path Mode is used.
For example when current dir is sys:wbstartup
ix_UseAmigaPaths(1);
h = fopen("/tools.info","r");
work
That it is possible to create amiga tasks with the amiga function, this function add the needed structure so all is 100% compatible
* ix_CreateChildData function add in libc.a.This create correct Userdata structure ixemul need
to store filehandles and used memlist of current task when a task is create with amiga OS create Task.
-
i have never understood whats not safe about it..
To my understanding it is use of tc_SigExcept and Exec exception handlers. I dont havei nterest to fully analyze what ixemul is doing there but it seems it is used to catch CTRL-C and abort program execution immediately. If you are not prepared to that memory leaks can occur. If you are not prepared to that and you are executing critical OS call you crash and burn. It can be fixed by disabling signals before making an OS call, at least ixemul is internally disabling signals when making such OS calls.
-
To my understanding it is use of tc_SigExcept and Exec exception handlers. I dont havei nterest to fully analyze what ixemul is doing there but it seems it is used to catch CTRL-C and abort program execution immediately. If you are not prepared to that memory leaks can occur. If you are not prepared to that and you are executing critical OS call you crash and burn. It can be fixed by disabling signals before making an OS call, at least ixemul is internally disabling signals when making such OS calls.
the ctrl c exception handler is only used on programs start from shell. a program that is start from wb (netsurf)do not do anything when press CTRL C, i have check it now, press and hold ctrl+c when netsurf run, do nothing. maybe other can too press ctrl+c if they can bring to crash netsurf. if so of course ixemul can fix. most users do not use ctrl+c. so best is for netsurf, switch it off
-
I test netsurf start on shell window. many single ctrl+c presses do not do anything, but press and hold ctrl+c down more than 10 sec close the netsurf window and let freeze amiga. so better is CTRL+c signal deactivate for netsurf. in ixemul docu stand how this can do
-
ixemul programs are in many cases not very responsive to keyboard input. especially on real hardware.
-
thats because desktop Linux programs that were later 1995 are never test at a 50 MHZ CPU. I have a cheap tablet 2* 1GHZ, this is slow as hell. often when press a key it need 2 or more sec until key is show. i disable name completion, but still slow. I only use for banking app. seem unix interrupt on general is slow.
-
Chris :
Tabs do work but they have some layout problems?
Sometimes when browser window is moved buttons doesn't refresh/show up again.
Prefs works just fine, but there is layout problem in screenmode page.
-
Tabs do work but they have some layout problems?
.
Have you tried closing them?
-
Closing tabs has random problems, sometimes it works, sometimes it gives Netsurf error, sometimes it just crash.
Do you have any use those error codes?
-
Closing tabs has random problems, sometimes it works, sometimes it gives Netsurf error, sometimes it just crash.
Do you have any use those error codes?
Not really, I know it's broken and can reproduce it easily enough.
-
NetSurf 3.5 BETA for AmigaOS 3.5+ is now available here: http://aminet.net/package/comm/www/netsurf_os3
Don't ask me what's changed, I can't remember :)
-
Great !
It runs very well and fast. A couple of things though:
- Couldn't it use the FPPrefs?
- Can't it be done without FPU needed?
-
Great !
It runs very well and fast.
Cool!
- Couldn't it use the FPPrefs?
I have no idea what that is.
- Can't it be done without FPU needed?
Probably, if I could figure out how to force gcc to use certain compiler switches by default.
-
About using the FPPrefs I was wondering if the images couldn't have some treatment and use the palette you are using, or at least some other. IBrowse for example looks much better -compared- when you use few colors. I believe that's because a option of MUI that IBrowse uses though.
-
Apologies if this has been asked before but any chance of a WarpOS version?
-
FFPrefs = Full palette http://aminet.net/package/util/wb/FullPalette22
I quest that Netsurf uses them if you open it to workbench screen.
FPU makes big difference to performance at least with 68040 CPU.
It seems that this version uses more memory than previous version used?
Tabs has problems closing and one new problem. Only half of second tab/page will be displayed. Is it because of Graphics card memory is ending?? Previous version could dispaly both amiga.org / amigaworld.net same time.
Memory fragmention is still there. Two first pages are rendered fast, third one uses 3x more time than should.
-
Still a long way to go as far as optimising it. It would be nice to see a team put in place to help with this. Also WarpOS is a great idea now that UAE can be configured with a CSPPC it should be easy to do as long as the developer knows how to code for it.
-
About using the FPPrefs I was wondering if the images couldn't have some treatment and use the palette you are using, or at least some other. IBrowse for example looks much better -compared- when you use few colors. I believe that's because a option of MUI that IBrowse uses though.
The palette mapping/dithering is done wrt the screen NetSurf is running on. If Multiview displays OK, then NetSurf should too. picture.datatype handles all this, I don't have any dedicated code for it as yet (that's an optimisation which really needs doing but I have no idea where to start).
Apologies if this has been asked before but any chance of a WarpOS version?
Presumably it will require yet another build of GCC. So, in theory, yes. In practice, probably not.
It seems that this version uses more memory than previous version used?
Shouldn't do, there are very few changes.
Only half of second tab/page will be displayed. Is it because of Graphics card memory is ending?? Previous version could dispaly both amiga.org / amigaworld.net same time.
Probably. Again, I've not knowingly done anything to screw that up.
The only thing I remember changing for rendering involved reducing the number of pen changes - as apparently that's a slow operation.
-
Is netsurf usable on 040@25hmz?
-
Why don't you test it yourself? Is it useable depends how you define it. Amiga.org renders 37 seconds with my 040/40mhz.
-
Is netsurf usable on 040@25hmz?
Usable yes, fun to use? Not so much. I have been testing it since he released it on my 060. Its getting better with every update, but still needs some more optimising. Check it out and see what you think is all I can say.
-
To Do List :
Fix memory fragemtnion / slow down effect
Implement faster image loading and also dithering for AGA
Best / fasted solution would be if there are some patches wich could be used for this already. At least image loader has been done hundres of times already, so some of those should be possible to adapt this Project.
-
Image loading isn't a problem. It's resampling it for a particular screen which is the tricky bit.
-
Now it is possible to log on amiga.org/amigaworld.net :) Previous versions gave errors.
-
OK, I've put a new version up - this is an experimental guigfx.library one.
Rendering=>Cache Native Versions should be set to "All" for fastest image display.
Play with the dithering too: "Low" is really "None", "Medium" is EDD and "High" is Floyd-Steinberg.
Here it looks like it has converted to a palette of about four colours, which is horrible, but might just be me.
Anyway, I'm more interested in the performance, I'll fix the display later, as it isn't working properly on OS4 either.
Now it is possible to log on amiga.org/amigaworld.net :) Previous versions gave errors.
ITYF you've reverted to an old version.
-
ITYF????
Also amigaworld doesn't have that annoying Graphics problem any more, can remember what caused it? Alpha mask??
Interesteing unfortunately I can't test it today, need to get some people from harbour late this night. So I need to start driving soon.
-
seems netsurf expects v44 version of reaction classes. is this a necessary requirement, or could this check be lowered in order to check with v41 class act ones?
-
seems netsurf expects v44 version of reaction classes. is this a necessary requirement, or could this check be lowered in order to check with v41 class act ones?
I'd have to check for sure which features I use, but I suspect not. It needs v44 of other components anyway.
-
OK, I've put a new version up - this is an experimental guigfx.library one.
Rendering=>Cache Native Versions should be set to "All" for fastest image display.
Play with the dithering too: "Low" is really "None", "Medium" is EDD and "High" is Floyd-Steinberg.
Here it looks like it has converted to a palette of about four colours, which is horrible, but might just be me.
Anyway, I'm more interested in the performance, I'll fix the display later, as it isn't working properly on OS4 either.
Your site still offers 16th april version?
-
I'd have to check for sure which features I use, but I suspect not. It needs v44 of other components anyway.
mmm. ok.. simply tried to run it on aros with the freely available classes. but it doesnt load there even with the proper reaction components, which probably expect 3.9 functionality. on my 3.9(+) (i dont have a clean install anymore) it loads properly (have not installed it) and runs for a time being loading simpler sites (images rendering still seems not to work well even with 32 bit) but crashes rather soon. im not able to make any pattern of it at this time, however.
edit: (i gave it a lot of stack of course, which is necessary for it to load from command line anyway, so this is not an issue)
might be a good idea to have a version with full debug compiled in for testing.
edit2: btw. im using the version from aminet, which probably is not the current one.
-
edit2: btw. im using the version from aminet, which probably is not the current one.
Other than the guigfx experiment it's the same.
Your site still offers 16th april version?
Yes, on Aminet.
-
It must be in upload queue, as I can't found it. Aminet version is excat same than you site offres.
-
It seems to be fast.
040/AGA loads amiga.org 29 sec
060/RTG loads amiga.org 19 sec
Cache native version seems to slow down 060/RTG, takes almost 10 seconds more to display amiga.org. It doesn't have any effect to scrolling either.
With 040/AGA it has huge difference fo scrolling and load time.
This one is most buggy and easy to get crash version so far :D resizing Windows etc is dangerous.
This version has still "slow down" effect. It slows down but load times doesn't grow exponentially any more. First two pages loads fast, but after that pages loads about 80-100 seconds, but load time doesn't seems to grow after that.
After quit, reboot is mandatory. Netsurf is slow after (program) restart without reboot.
I used that guigfx library from aminet, is it latest one (2005)? I could found newer one but it is for MUI.
-
It seems to be fast.
After quit, reboot is mandatory. Netsurf is slow after (program) restart without reboot.
It has always crashed on exit here, probably memory trashing.
I used that guigfx library from aminet, is it latest one (2005)? I could found newer one but it is for MUI.
That's the latest AFAIK. It's probably better to use the old 68k asm version of render.library with it, rather than the newer one.
-
netsurf must be heavily trashing the memory all around. it takes down the whole uae with it when it fails. i might try to recall how ro run muforce or uae enforcer in order to catch these hits. but in order to localize the offending function the binary would have to be compiled with debug symbols anyway. the question is if gccfindhit would work with the utilized compiler, whichever it is. however probably the most sensible would be to enable all debug first. it doesnt make sense to try to speedup and optimize a buggy binary.
-
netsurf must be heavily trashing the memory all around. it takes down the whole uae with it when it fails.
Really? I've not seen that here.
i might try to recall how ro run muforce or uae enforcer in order to catch these hits. but in order to localize the offending function the binary would have to be compiled with debug symbols anyway. the question is if gccfindhit would work with the utilized compiler, whichever it is. however probably the most sensible would be to enable all debug first. it doesnt make sense to try to speedup and optimize a buggy binary.
gccfindhit doesn't work, so I have no way of tracking down Enforcer hits beyond somebody who knows what they are doing disassembling the code.
I'm basically doing what I can with my limited resources, and that doesn't include spending hours tracking down a memory bug I have no hope of finding except by chance. I suspect it's something like the MinList bug which I would never have found without matthey's help.
-
OK, new version up which should now be showing the correct colours. guigfx doesn't seem to do anywhere near as good a job as picture.datatype at colour reduction, but I think the speed up is worth it. If it's broadly OK then the DataTypes-based quantizer will be removed.
-
Really? I've not seen that here.
have you browsed from site to site or simply loaded a site just once to verify the browser works? a reproducable way to crash netsurf and uae was here to navigate to google.com, enter amiga as search phrase and click a link to amiga wikipedia article, afair.
gccfindhit doesn't work, so I have no way of tracking down Enforcer hits beyond somebody who knows what they are doing disassembling the code.
I'm basically doing what I can with my limited resources, and that doesn't include spending hours tracking down a memory bug I have no hope of finding except by chance. I suspect it's something like the MinList bug which I would never have found without matthey's help.
maybe its due to my setup, not being clean 3.9, but it isnt crashy otherwise,so i must assume its netsurf this time. what concerns the debug techniques thats unfortunate, since it makes it almost impossible to somehow contribute useful feedback to this project. dont you have at least debug statements spread in the code you could simply turn on in order to find the crash location at least roughly? i know i could look at the sources myself;..
how about others? dont they experience stability issues? because if so, it would be better to go for stability first and then for performance.
-
Netsurf has been very stable here, no crashes.
Fast test today. Nothing about speed this time, as I just installed it and loaded few pages. Lets see if I have any energy left latter today.
Amiga.org's logo text amiga.org is red with AGA. No color problems with RTG.
Made some tests with iBrowse, it starts to display site faster, but over all loading time is quite near same.
-
OK tested more. I thing that speed is little bit degraded, not much, it is still faster than version wich uses picture.datatype.
Memory usage is about same, no noticeable diffrence. Speed is still a problem after few pages.
How about this : I noticed that amigaworld.net uses much less memory than amiga.org. Netsurf freed memory nicely. Wich made me thing IF Netsurf would load empty page always before new "real page", it would terminate memory fragmetnion? Loading empty page wouldn't take too much time, but memory would have big contionous free block? This would be very stupid question and if it would work it would be very unsophisticated solution
-
just tried the last development version. images seem to display all right on rtg 32bit. the crash no matter uae version and cpu setting remains the same. but so far not confirmed by others i assume its my setup fault.
-
have you browsed from site to site or simply loaded a site just once to verify the browser works? a reproducable way to crash netsurf and uae was here to navigate to google.com, enter amiga as search phrase and click a link to amiga wikipedia article, afair.
I'll try that when I have some more time. I do browse around a few pages and try different pages each time, but I've not been using it extensively. I'm surprised there haven't been more crashes reported.
maybe its due to my setup, not being clean 3.9, but it isnt crashy otherwise,so i must assume its netsurf this time. what concerns the debug techniques thats unfortunate, since it makes it almost impossible to somehow contribute useful feedback to this project. dont you have at least debug statements spread in the code you could simply turn on in order to find the crash location at least roughly? i know i could look at the sources myself;..
how about others? dont they experience stability issues? because if so, it would be better to go for stability first and then for performance.
NetSurf -v will produce a log. That's useful for indicating where it crashes but doesn't necessarily pinpoint the exact location.
just tried the last development version. images seem to display all right on rtg 32bit. the crash no matter uae version and cpu setting remains the same. but so far not confirmed by others i assume its my setup fault.
I suspect your crash is genuine. I know for sure there's some memory corruption happening which is having weird effects. But as it's only happening on OS3 (which I can't easily debug) and not OS4 it isn't easy to track down - especially as the time the crash happens doesn't relate to when the memory was overwritten.
Amiga.org's logo text amiga.org is red with AGA. No color problems with RTG.
I noticed white text showed up in an odd colour in another image. Is everything else the right colour on AGA?
-
NetSurf -v will produce a log. That's useful for indicating where it crashes but doesn't necessarily pinpoint the exact location.
ah, i recall. i have forgotten that. thats good, might give some hint;)
-
i need to redirect the outputu to serial to catch the log orderly.
so far after a lot of ami_font_bm_open calls, which seem to process all fine there is browser_window_update_favicon in desktop/browser.c 1182
and then amiga/schedule.c 247amiga_scheduler_run callback with at the first sight possibly valid arguments. callback like that look like being executed alright previously
and thats it.
-
Would it be possible add friend_bitmap to settings GUI? or is it there already?
Amigaworld.net logo has also red color. I think that everything is OK otherwise, dithering colors are acceptable.
-
i need to redirect the outputu to serial to catch the log orderly.
so far after a lot of ami_font_bm_open calls, which seem to process all fine there is browser_window_update_favicon in desktop/browser.c 1182
and then amiga/schedule.c 247amiga_scheduler_run callback with at the first sight possibly valid arguments. callback like that look like being executed alright previously
and thats it.
That sounds like something I spent quite a while looking into previously, and decided it was only crashing there because something previously had overwritten memory. I was still none the wiser as to what.
-
maybe running enforcer wouldnt be that bad idea after all, with the -v option we could roughly see where the write hits are coming from. then we could inspect suspicious segment of code closer, even if with deadwoods method inserting debug statements by hand. or maybe someone familiar with uae debugger could help single step it. it woul probably be the content of one function, wouldnt it?
-
Would it be possible add friend_bitmap to settings GUI? or is it there already?
No, I'd rather get rid of it completely.
Amigaworld.net logo has also red color. I think that everything is OK otherwise, dithering colors are acceptable.
It's looks like red and white are getting swapped... weird.
-
I made some testing.
I created empty html document, nothing on it, just
Named it to test.html
It loads 0,5 seconds.
Loading amiga.org and after that test.html, swapping between amiga.org/test.html, I could keep amiga.org loading times less than 36 seconds. I also tested to load amiga.org -> forums -> test.html even after that amiga.org loaded less than 36 seconds. Tested with netsurf local home page, it also works.
Noticed that largest free block is still getting smaller every time, and memory is slowly eaten. Loading those pages about a dozen times reserves about 7mb ram. Does it free memory at all??? Amiga.org loaded I have 35mb free ram, test.html loaded I have 35mb free ram. Going amiga.org again eats memory less than 1mb. First time loaded amiga.org uses about 5-7mb ram.
So what it keeps on ram?
-
I made some testing.
Noticed that largest free block is still getting smaller every time, and memory is slowly eaten. Loading those pages about a dozen times reserves about 7mb ram. Does it free memory at all??? Amiga.org loaded I have 35mb free ram, test.html loaded I have 35mb free ram. Going amiga.org again eats memory less than 1mb. First time loaded amiga.org uses about 5-7mb ram.
So what it keeps on ram?
Are you using gcc to build this beastie Chris?
When I was porting Quake 2 to AGA I discovered that the standard memory allocation functions 'malloc/free' are buggy and would not always release allocated memory (I was compiling with -noixemul BTW).
The solution (which came to me after sacrificing many virgins), was to wrap my memory allocation functions and redirect to the Amiga OS standard AllocVec/FreeVec functions.
If that doesn't work you'll need to write your own hunk allocation memory manager.
http://gameprogrammingpatterns.com/object-pool.html (http://gameprogrammingpatterns.com/object-pool.html)
-
What you see is quite normal.
The C standard library contains its own memory allocation manager.
It allocates big chunks from the OS and manages them internally, which is why you dont see OS memory releases as often as you might expect.
-
now with winuaeenforcer i cannot trace the issue, it actually opened the wikipedia page. however i must correct myself. the images are not rendered correctly.
-
Tested with WinUAE (Amiga Forever 2013)
It doesn't start, it kills UAE. Also previous install was tested with it without problems.
-
Tested with WinUAE (Amiga Forever 2013)
It doesn't start, it kills UAE. Also previous install was tested with it without problems.
By "previous one" I assume you are comparing the previous guigfx version with the new guigfx version?
There's very little change between them but I could have screwed something up.
-
now with winuaeenforcer i cannot trace the issue, it actually opened the wikipedia page. however i must correct myself. the images are not rendered correctly.
Not rendered correctly how?
-
What you see is quite normal.
The C standard library contains its own memory allocation manager.
It allocates big chunks from the OS and manages them internally, which is why you dont see OS memory releases as often as you might expect.
Also switching between a blank page and amiga.org isn't going to free up much memory, because most of amiga.org will remain in the memory cache.
Reduce the size of the memory cache to free memory up more frequently (won't stop it fragmenting though)
-
By "previous one" I assume you are comparing the previous guigfx version with the new guigfx version?
I mean version wich used datatypes.
Also switching between a blank page and amiga.org isn't going to free up much memory, because most of amiga.org will remain in the memory cache.
It keeps Netsurf speed acceptable level. It is easy to reproduce By someone else, but it would require real hardware.
Reloading amiga.org front page makes slow down. Two times it is about same time, third is 60-80 seconds, fourth is something like 120 seconds and fith is over three minits.
So loading amiga.org front page / forums about a twelve times less than 36 seconds, is something.
If pages are keeped in memory, fragmention happen because??? What is does when it loads page from memory???
Did you notice Novacoder's reply?
When I was porting Quake 2 to AGA I discovered that the standard memory allocation functions 'malloc/free' are buggy and would not always release allocated memory (I was compiling with -noixemul BTW).
The solution (which came to me after sacrificing many virgins), was to wrap my memory allocation functions and redirect to the Amiga OS standard AllocVec/FreeVec functions.
-
OK, new version: http://cy2.uk/netsurfos3
-
Thank you.
But: Please tell us, which kind of lha-compression or version do you use?
I cannot decompress the archive under OS3.9. It says a newer lha version is required.
I have installed version 2.15 from January 2011.
-
No Problems opening the archive here, using default os3.9 unarc.
-
Thank you.
But: Please tell us, which kind of lha-compression or version do you use?
I cannot decompress the archive under OS3.9. It says a newer lha version is required.
I have installed version 2.15 from January 2011.
It's a Linux version, you need to use xpk as the command line LhA doesn't support it.
-
Slow down effect seems to be gone. Congratulations :)
Browser is generally slower now and first page downloads and render quite long.
Amiga.org took 48 seconds in first load, after that amiga.org page download took 25-30 seconds. home - > Forums - > home - > Forums - > etc.
Cache native versions helps quite lot. With AGA is good idee to open browser it own screen.