Netsurf for 68k amigas, css capable web brower

[MozzerFan]

Still don't know why netsurf 68k runs on some (winuae) systems, but not on mine. I've assigned amitcp as suggested earlier, but netsurf still doesn't see the connection. Pretty strange, as aweb and ibrowse do work perfectly.
Does anyone have some more suggestions
 
x303

If you use AmiKit:
The default assignment for AMITCP: in AmiKit is the Ibrowse drawer, so make sure to replace this assignment.
edit: The assignment is made in user-startup. Either replace the current assignment, or after this assignment add the following:
Assign AMITCP: ADD
Check in a CLI window by entering "assign" to which drawer AMITCP is currently assigned.
 
Also check if you've installed all the fonts that come with Netsurf if your running the TTF version.
Edit: If the nofreetype version runs ok, you haven't installed the fonts correctly. You must install them in Fonts:truetype/ttf-bitstream-vera
 
Also check if you have the latest ixemul lib.

[MozzerFan]

Great work Artur and Bernd. It's pretty fast under WinUAE.
Still have to test it on my a4000.
 
Thank you !!!

[unusedunused]

Here, with OWB 1.3 for MorphOS, it takes about 20s to load fully, and you can already scroll to bottom after 4s (note that on MorphOS, OWB network manager is threaded).

FireFox gave a similar result, loading in about 20s too, and allowing to scroll after about 4-5s.

But i wonder once again why i replied about that kind of benchmark that depends on way too many factors.

You are free to run a cache proxy and bring more accurate results.
But because most are too lazy to do this extra work, then this values are better than nothing to see if amiga browsers work good in compare to firefox, to see if it is usefull to have firefox too

also you can post values when you do a site back to show how long time need to show full page but in cache.

its also intresting to see how the speed is when MOS OWB use no threads.if it is possible to disable by define.

as far i know OS4 OWB use no threads and 68k too not.

Maybe it help users to choose the better OWB.

I think using thread is important, to get same speed as firefox
but if not really know if it have noticable speedup with threa

[chris]

netsurf2.1 on os4 starts to display but locks the system before the images load

Works on the latest dev version - 19.3s reported for t-online.de

[wawrzon]

@x303:

Does anyone have some more suggestions

get out the real hw without w$ underneath:P

@bernd:

If you have your amiga connect to a PC you can run a cache proxy here.

when you set a env var on amiga OS
http_proxy

yeah im lazy, i know. but in this case my amigas and the pc are connected independently to the net via telecom router. i dont now if i can establish a proxy somewhere within my home network but i dont think this speed comparizons are so important except nmaybe for the threading you have spoken of

@chris

Quote:
Originally Posted by wawrzon  
netsurf2.1 on os4 starts to display but locks the system before the images load

Works on the latest dev version - 19.3s reported for t-online.de

i have some probs that might be due to im using elbox voodoo drivers, i have to revert to the hyperion ones if im able to get the network card working that way. again was way too lazy to do it till now. maybe then os4 will get more stable

[wawrzon]

@artur:

Quote:
do this port stays on sdl? i suppose further features like window resizing will not be possible in this case?  

Window resize can be done be editing Resources/Options file.

i mean resize on the fly. but sure it demands an amiga-gui. anyways nice for the full initioal release. i just dl it and everything seems to work nice.

Quote:
running out of memory on big pages is still a problem, regardless the stack. if the reason is unknown maybe it could be asked on ml?  

Can you tell on which page do you get out of memory?

bbc news link on netsurf start page fi. i will find some other sites and post it here. i think generally every complex site runs eventually out of mem. just testing...


edit:
i just noticed a minor issue: alpha channel still doesnt work in overleaping mages like the arrow in the top right corner of netsurf-browser.org. if i recall right this renders ok on the os4 port.

check ns bug tracker page too. there is some color corruption in images. the top bar (sf.net) backgroundis sapphire should be black. the top bar of the list has the same taint, and the list background gradient is rendered i two fields of color. i suppose this is still some overseen minor endiannes issue somwhere.
32bit argb argb on my voodoo the whole sdl window is tainted yellow too, -i think this is adressed to bernd and his sdl port.

edit: wow! ttf text input has been improved! very nice! arrow keys still not working though. any chance window bars can be dragged with the mouse? (mousewheel is working cool)

running out of mem on bild.de (bernds beloved test site, lol)

edit:
oughh!! i just had a hit. and my first netsurf caused complete system lockup under 3.x!! that while reloading t-online.de. the page rendered quite correctly the first time but i had an impression of some images missing. it took 92s btw. now i go take my sunday bike tour. take care!

[unusedunused]

@artur:


bbc news link on netsurf start page fi. i will find some other sites and post it here. i think generally every complex site runs eventually out of mem. just testing...

i get too this out of memory error in netsurf status page that netsurf report when press on BBC Link.

sometimes the error is not show and BBC Page is also not draw.
the last page is show.

I think thats too a netsurf problem or does it work on OS4 ?

btw if you change the window size pref of netsurf to your screen res can cause problems.best is you use 20 pixel less so border can fit.

>32bit argb argb on my voodoo the whole sdl window is tainted >yellow too, -i think this is adressed to bernd and his sdl port.

netsurf SDL have problem to use SDL pixelformats.
it work only when you choose a bgra or bgr screen bgr 16pc

but now this version detect if run on 16 or 32 bit screen and use correct depth.

[wawrzon]

i get too this out of memory error in netsurf status page that netsurf report when press on BBC Link.

since the problem doesnt show up on some other big sites im not sure if this is really out of mem problem. maybe it is wrongly reported as such. i dont know.

bernd,i have reproduced the failure on reloading t-online.de and this timethere was no hit nor crash but the request of ixemul.library: abort trap. after that netsurf closed quietly.

something strange happens on amigans.net (maybe due to java script) and on utilitybase.com. you noticed? i havnt confirmed it using os4 version yet, will try it later.

edit: no, netsurf os4 loads news.bbc.co.uk succesfully (29.1s) reporting one object of a wrong type.
also no probs on utilitybase.com and amigans.net (corruption around the cursor)
well, but this might be due to sdl version limitations.

[unusedunused]

since the problem doesnt show up on some other big sites im not sure if this is really out of mem problem. maybe it is wrongly reported as such. i dont know.

bernd,i have reproduced the failure on reloading t-online.de and this timethere was no hit nor crash but the request of ixemul.library: abort trap. after that netsurf closed quietly.

something strange happens on amigans.net (maybe due to java script) and on utilitybase.com. you noticed? i havnt confirmed it using os4 version yet, will try it later.

edit: no, netsurf os4 loads news.bbc.co.uk succesfully (29.1s) reporting one object of a wrong type.
also no probs on utilitybase.com and amigans.net (corruption around the cursor)
well, but this might be due to sdl version limitations.

can you make screenshot of this page ?
what screen res you have on workbench and what setting you have in netsurf screenwidth option ?

i test the version from aminet, it create a shadow surface and use HWsurface on my system.please look what happen on your system.

normaly output on sashimi should be no shadowsurface

[wawrzon]

can you make screenshot of this page ?
what screen res you have on workbench and what setting you have in netsurf screenwidth option ?

workbench: 1680x1050 netsurf:1200x1000

i test the version from aminet, it create a shadow surface and use HWsurface on my system.please look what happen on your system.

normaly output on sashimi should be no shadowsurface

[/QUOTE]
me too. i have SDL_HWSURFACE set to 1 in the env if this is what you mean. sashimi reports: create shadow surface.

[wawrzon]

screenshot

[wawrzon]

i just had a netsurf crash on loading t-online.de. i have managed to catch the sashimi output:

[unusedunused]

workbench: 1680x1050 netsurf:1200x1000

thats a memtrash error.
I try your resolution and after i press reload i get a abort trap error and netsurf quit.

then i try out 1024*768

this too dont work get after second reload too problem.allis reproducable

I test on 800*600 more than 10 reloads and all work ok.

please try out to use a resolution of 800*600 or below

seem netsurf have a >800 Bug.

I try out resolutions 761*571 to see maybe sdl have a allign memtrash problem.but this resolution too work ok with more than 10 reloads no crash

maybe you can try out the OS4 version and set window to below 800*600.

maybe on OS4 its same the large window cause the crash.

[wawrzon]

using 1024x768 on another system i had a hangup and now repeating hit without a crash just like that:

Code: [Select]


BUFFER WAS OVERRUN - Data may have been lost
BUFFER WRAPPED - This is the most recent captured data

 movea.l a1,a0
08dc5408 : *10c0                       move.b d0,(a0)+
08dc540a :  5381                       subq.l #$1,d1
08dc540c :  66fa                       bne.s $8dc5408
08dc540e :  2009                       move.l a1,d0
08dc5410 :  4e75                       rts
08dc5412 :  0000 226f                  ori.b #$6f,d0
08dc5416 :  0004 122f                  ori.b #$2f,d4
08dc541a :  000b 7000                  ori.b #$0,a3 ;illegal opcode
08dc541e :  b211                       cmp.b (a1),d1
08dc5420 :  6602                       bne.s $8dc5424
08dc5422 :  2009                       move.l a1,d0
08dc5424 :  4a19                       tst.b (a1)+
08dc5426 :  66f6                       bne.s $8dc541e
Name: "Netsurf"  


20-Apr-08   21:43:16
BYTE WRITE to  00000458        data=13         PC: 08DC5408
USP : 08BE986C SR: 0000  (U0)(-)(-)  TCB: 08B45D30
Data: 00000013 FFFFFBA6 00008536 00000001 00000000 00000000 00000000 00000000
Addr: 00000458 FFFFFFFF 08D88488 092511B8 08F6308E 08BE9880 08037B1C 0800232C
Stck: 09198FF0 FFFFFFFF 00000013 FFFFFFFF FFFFFFFB 08BE9898 0918C10C FFFFFFFF
Stck: 000002B7 0918BAAC FFFFFFFF 08BE98AC 091934F8 FFFFFFFF 095AE838 08C3F598
Stck: 08BE98C0 091A41C4 08C3F598 095AE838 FFFFFFFF 08BE98E8 09193AA8 08C3F610
Stck: 095AF9B8 08C3F598 095AE838 00000001 08C3F610 09AAA338 095AF9B8 08BE9904
Stck: 0918B9A4 08C3F598 08BE98FC 0918B938 0000003C 480BB8B2 08BE991C 0918BA0A
Stck: 08C3F598 0000003C 480BB8B2 480BB8B2 08BE9938 0918F482 095F9480 FFFFFFFF
Stck: 00000000 08BE9960 095F9480 08BE9960 0918FB6A 09AD8918 08BE9960 09191D50
Stck: 08BE999C 00010000 00000000 09AD8918 00000000 08BE9BAC 09192A4E 095F9480
----> 08DC5408 - "Work:Libs/ixemul.library"  Hunk 0000 Offset 00017F28
----> 09198FF0 - "Work:Internet/NetSurf/Netsurf"  Hunk 0000 Offset 002A8358
----> 0918C10C - "Work:Internet/NetSurf/Netsurf"  Hunk 0000 Offset 0029B474
----> 0918BAAC - "Work:Internet/NetSurf/Netsurf"  Hunk 0000 Offset 0029AE14
----> 091934F8 - "Work:Internet/NetSurf/Netsurf"  Hunk 0000 Offset 002A2860
----> 091A41C4 - "Work:Internet/NetSurf/Netsurf"  Hunk 0000 Offset 002B352C
----> 09193AA8 - "Work:Internet/NetSurf/Netsurf"  Hunk 0000 Offset 002A2E10
----> 0918B9A4 - "Work:Internet/NetSurf/Netsurf"  Hunk 0000 Offset 0029AD0C
----> 0918B938 - "Work:Internet/NetSurf/Netsurf"  Hunk 0000 Offset 0029ACA0
----> 0918BA0A - "Work:Internet/NetSurf/Netsurf"  Hunk 0000 Offset 0029AD72
----> 0918F482 - "Work:Internet/NetSurf/Netsurf"  Hunk 0000 Offset 0029E7EA
----> 0918FB6A - "Work:Internet/NetSurf/Netsurf"  Hunk 0000 Offset 0029EED2
----> 09191D50 - "Work:Internet/NetSurf/Netsurf"  Hunk 0000 Offset 002A10B8
----> 09192A4E - "Work:Internet/NetSurf/Netsurf"  Hunk 0000 Offset 002A1DB6
PC-8: 61FFFFFF FE0A4FEF 000C2002 241F4E75 226F0004 202F0008 222F000C 67082049
PC *: 10C05381 66FA2009 4E750000 226F0004 122F000B 7000B211 66022009 4A1966F6
08dc53e8 :  61ff ffff fe0a             bsr.l $8dc51f4 ;extended opcode
08dc53ee :  4fef 000c                  lea.l $c(a7),a7
08dc53f2 :  2002                       move.l d2,d0
08dc53f4 :  241f                       move.l (a7)+,d2
08dc53f6 :  4e75                       rts
08dc53f8 :  226f 0004                  movea.l $4(a7),a1
08dc53fc :  202f 0008                  move.l $8(a7),d0
08dc5400 :  222f 000c                  move.l $c(a7),d1
08dc5404 :  6708                       beq.s $8dc540e
08dc5406 :  2049                       movea.l a1,a0
08dc5408 : *10c0                       move.b d0,(a0)+
08dc540a :  5381                       subq.l #$1,d1
08dc540c :  66fa                       bne.s $8dc5408
08dc540e :  2009                       move.l a1,d0
08dc5410 :  4e75                       rts
08dc5412 :  0000 226f                  ori.b #$6f,d0
08dc5416 :  0004 122f                  ori.b #$2f,d4
08dc541a :  000b 7000                  ori.b #$0,a3 ;illegal opcode
08dc541e :  b211                       cmp.b (a1),d1
08dc5420 :  6602                       bne.s $8dc5424
08dc5422 :  2009                       move.l a1,d0
08dc5424 :  4a19                       tst.b (a1)+
08dc5426 :  66f6                       bne.s $8dc541e
Name: "Netsurf"


while using the 800x600 did work on both. even if reload was sometimes rediculously slow (obout 5 min). might be due to my home network with 2 miggys and one pc connected to the router.

edit: bernd, were you able to reproduce these strange pointer effects on utilitybase and amigans? i had something similar while editing image on arteffect with the netsurf window in the background. the image disaperared and i was able to get it back by resizing window or just "painting" with cursor over it. i think there is still some sdl bug.

edit: under os4 i get a crash while rendering t-online no matter what resolution

[Fab]

You are free to run a cache proxy and bring more accurate results.
But because most are too lazy to do this extra work, then this values are better than nothing to see if amiga browsers work good in compare to firefox, to see if it is usefull to have firefox too

also you can post values when you do a site back to show how long time need to show full page but in cache.

its also intresting to see how the speed is when MOS OWB use no threads.if it is possible to disable by define.

as far i know OS4 OWB use no threads and 68k too not.

Maybe it help users to choose the better OWB.

I think using thread is important, to get same speed as firefox
but if not really know if it have noticable speedup with threa

Actually, the way it's currently done, threads don't speed up the process, since all the connections run in a single (and separated from main context) thread.

But like i already said previously, all CURL handles are handled asynchronously (non blocking sockets, process stuff when data arrives,...). The problem is some parts of the process are still a bit long (data reception) or very long (DNS resolution which is a major blocking call, unless c-ares or an additional lookup thread is spawned).

So, even tho CURL is mostly asynchronous, it still makes sense to run it in a dedicated thread to avoid blocking the gui during DNS lookups or long data reception phase.

[unusedunused]

edit: bernd, were you able to reproduce these strange pointer effects on utilitybase and amigans? i had something similar while editing

Edit:
this i see too.when scroll with cursor then a width rectangle is see.

But when wait until the bitmaps(OS4 AROS classic images) are draw before mouse move or scroll, then all is ok.but sometimes i get no images show only when scroll full down and then up

strange what SDL can do with that