Welcome, Guest. Please login or register.
Amiga Kit Amiga Store Iridium Banner AMIStore App Store A1200/A600 4xIDE Interface

AuthorTopic: NetSurf OS3 - testing!  (Read 11335 times)

0 Members and 1 Guest are viewing this topic.

Offline Gulliver

Re: NetSurf OS3 - testing!
« Reply #105 on: April 16, 2019, 07:04:49 PM »
@utri007

That the Apollo accelerator is bugged upto its guts has been known for at least a decade. Specific timing fixes on some A1200 motherboards are required to make it work, then, as Thomas pointed out, memory not being added to the system in a proper manner, then failing on 3.9 BB2 rom update under some circumstances. Yes, they may work ok without any drama if you get exactly the right motherboard, and have a bit of luck running some apps.

That was one of the reasons Jens Schoenfeld critiziced so much its design when he bought the Apollo rights long ago (a adecade ago?). He said something along the lines that he wanted to buy them to bury that bad design forever. You may find his technical arguments against it if you look hard enough on the EAB forum.

Anyway, back to your issue: it seems that you are running a customized kickstart rom. I have been there, done that. It is a nightmare waiting to happen. About 75% of the hacked/patched rom modules out there have a small issue here and there. Not recommended for stability. Bugs are going to surface and you will probably end up blaming it on to something else, like bad psu, old caps, or another patch or system program you use.

The unmodified 3.1.4 kickstart rom is superior to any previous version of AmigaOS rom out there. Let me give you a very quick summary of the changes between AmigaOS 3.9 BB2 and 3.1.4 roms on an A1200:

AmigaOS 3.9 BoingBag 2

bootmenu 44.7 (25.2.2002)
console.device 44.8 (12.6.2001)
exec.library 45.20 (6.1.2002)
FileSystem.resource 45.10 (3.8.2001)
FileSystem 45.13 (3.8.2001)
ram-handler 44.23 (8.4.2001)
icon.library 45.1 (8.2.2001)
scsi.device 43.43 (16.1.2002)
shell 45.7 (13.01.2002)
workbench.library 45.127 (21.2.2001)

TOTAL = 10 modules changed.

Note: Despite these changes, many are widely known to be buggy like ram-handler still displaying 100% ram disk usage while being empty. Or the failed scsi.device that does not support large drives and had to be hacked by users to try to fix it, workbench.library that could have its fuel gauge meter overflowing and also its memory leaking abilities, etc. I could go on and on about each of those modules.

AmigaOS 3.1.4

exec.library            exec 46.45 (18.9.2018)
utility.library         utility 45.2 (15.1.2018)
FileSystem.resource     filesysres 46.1 (13.8.2017)
battclock.resource      battclock 45.1 (30.11.2017)
graphics.library        graphics 45.27 (20.9.2018)
layers.library          layers 45.30 (31.7.2018)
gameport.device         gameport 45.2 (9.7.2018)
timer.device            timer 45.1 (14.7.2018)
card.resource           cardres 45.2 (11.5.2018)
keyboard.device         keyboard 45.2 (9.7.2018)
input.device            input 45.2 (9.7.2018)
ramdrive.device         ramdrive 45.3 (14.7.2018)
trackdisk.device        trackdisk 45.1 (14.7.2018)
scsi.device             IDE_scsidisk 45.7 (16.5.2018)
console.device          console 45.4 (9.7.2018)
mathieeesingbas.library mathieeesingbas 45.9 (16.7.2018)
syscheck                syscheck 45.1 (11.2.2018)
bootmenu                bootmenu 45.6 (5.8.2018)
filesystem              fs 46.13 (23.9.2018)
audio.device            audio 45.15 (20.9.2018)
mathffp.library         mathffp 45.3 (6.1.2017)
icon.library            icon.library 45.22 (24.9.2018)
workbench.library       workbench.library 45.194 (23.9.2018)
shell                   shell 46.10 (12.9.2018)
ram-handler             ram 45.5 (6.5.2018)
expansion.library       expansion 45.4 (12.9.2018)

And then you have the optional module: intuition.library 45.10 (16.9.2018)

TOTAL 24 modules changed + 1 extra optional.

Note: Only one bug uncovered in Shell, that is already in beta testing phase to be released free of charge for 3.1.4 adopters. The fix won't require a new rom.

It is an easy choice regarding kickstarts. I admit there is some nice stuff to have from the disk based components of 3.9, like ReAction programs, media players and internet stuff, but if you are looking for a better core OS, 3.9 is a certainly a well disguised nightmare, and 3.1.4 undoubtly kicks its ass a hundred times.

So stay safe and use an unmodified rom. We have had many success reports of 3.1.4 users
with Apollo accelerators. You should not be an exception.   :)
« Last Edit: April 16, 2019, 07:08:46 PM by Gulliver »
 

Offline utri007

Re: NetSurf OS3 - testing!
« Reply #106 on: April 16, 2019, 07:33:47 PM »
But I do have a 3.1.4 kickstart. It has all the rom modules from A1200 modules disk. Others are from original 3.1 kickstart.

There is nothing patched or exotic inside of it. It has a workbench and intuition libraries. Some file systems and cf device.

I'm happy to demonstrate problem with with Blizzard 68060 accelerator.

I hate the idee that I need a special boot floppy if I need a reinstall them. Now I can start it (in theory) from Deluxe Paint III floppy so that I can start installing it from CD or CF card.

Thing is, I don't have any problems. Except that Chris' Netsurf is bog slow, but I think that issue is Netsurf, not a kickstart (or 3.1.4 modules) just thinking that something wich has changed is causing problems and that could help to solve Netsurf problems.

With Kickstart using KS31 and OS3.9 modules it worked much faster and it didn't had that "crash on exit" bug that others had. With current kickstart it crash when closed.
« Last Edit: April 16, 2019, 07:48:16 PM by utri007 »
ACube Sam 440ep Flex 800mhz, 1gb ram and 240gb hd and OS4.1FE
A1200 Micronic tower, OS3.9, Apollo 060 66mhz, xPert Merlin, Delfina Lite and Micronic Scandy, 500Gb hd, 66mb ram, DVD-burner and WLAN.
A1200 desktop, OS3.9, Blizzard 060 66mhz, 66mb ram, Ide Fix Express with 160Gb HD and WLAN
A500 OS2.1, GVP+HD8 with 4mb ram, 1mb chip ram and 4gb HD
Commodore CDTV KS3.1, 1mb chip, 4mb fast ram and IDE HD
 

Offline Gulliver

Re: NetSurf OS3 - testing!
« Reply #107 on: April 16, 2019, 08:24:55 PM »
Blizzards also work, we have testers and many users working succesfully with them.

Anyway, I see your point, but the problem is that you have a lot of variables in your system, and pointing the finger to something is just speculation at this point (it could just be NetSurf, 3.1.4, a 3.9 component, a 3.1 compomentt, another filesystem, etc.). Something that you might feel is safe/harmless might not be that way, so do not exclude components when you still aren't 100% sure what is the cause.

Bottomline: more reasearch is definately needed.

Offline utri007

Re: NetSurf OS3 - testing!
« Reply #108 on: April 16, 2019, 08:43:04 PM »
Quote
Blizzards also work, we have testers and many users working succesfully with them.

Anyway, I see your point, but the problem is that you have a lot of variables in your system, and pointing the finger to something is just speculation at this point (it could just be NetSurf, 3.1.4, a 3.9 component, a 3.1 compomentt, another filesystem, etc.). Something that you might feel is safe/harmless might not be that way, so do not exclude components when you still aren't 100% sure what is the cause.

Bottomline: more reasearch is definately needed.

All amigas has a lots of variables :D

My "main" amiga is a towerized A1200 with Zorro extender. It has a notorius eXpert Merlin RTG card and Delfina Lite. But also a DCE scan magic and Ide fix express.

So "custon" is a standard here, lots of variables.

My Blizzard machine is much more clear A1200 with Blizzard 1260 accelerator.

My point is NOT a say that there is a problem because of 3.1.4 rom modules, I just say that Netsurf has a problems with them. And that way maybe it will help to find bugs. Netsurf window is refreshed "totally" 2 or 3 times before it displays a Netsurf's start screen. I thing that wasn't there before either.


« Last Edit: April 16, 2019, 08:47:29 PM by utri007 »
ACube Sam 440ep Flex 800mhz, 1gb ram and 240gb hd and OS4.1FE
A1200 Micronic tower, OS3.9, Apollo 060 66mhz, xPert Merlin, Delfina Lite and Micronic Scandy, 500Gb hd, 66mb ram, DVD-burner and WLAN.
A1200 desktop, OS3.9, Blizzard 060 66mhz, 66mb ram, Ide Fix Express with 160Gb HD and WLAN
A500 OS2.1, GVP+HD8 with 4mb ram, 1mb chip ram and 4gb HD
Commodore CDTV KS3.1, 1mb chip, 4mb fast ram and IDE HD
 

Offline Gulliver

Re: NetSurf OS3 - testing!
« Reply #109 on: April 16, 2019, 10:31:57 PM »
Yes, it is certainly true that all Amigas have lots of variables, and that is why testing is so difficult.

What I was trying to say, is that allways when testing, try to reduce all uneeded variables to narrow down the missbehaving component. A clean system, no matter what flavor/version, will make things easier, as you will eliminate quite a lot of variables.

For testing purposes I allways use an unmodified AmigaOS partition and boot from there. If the bug persists, the number of variables is infinitely smaller than on my everyday system volume.

It is a job of narrowing down variables, as some missbehaviours are not evident at first sight/guess.

Speaking of missbehaviours, it seems we just found one on the other Netsurf port.
 

Offline utri007

Re: NetSurf OS3 - testing!
« Reply #110 on: April 17, 2019, 09:44:43 PM »
Some outputs from MuForce / Segtracker / Wipeout

A1200 / Blizzard 1260

With lattest version (10th april) of Netsurf it quits without crash. But still getting cannot fetch documents. Some times it renders one page ie. amiga.org frontpage.
ACube Sam 440ep Flex 800mhz, 1gb ram and 240gb hd and OS4.1FE
A1200 Micronic tower, OS3.9, Apollo 060 66mhz, xPert Merlin, Delfina Lite and Micronic Scandy, 500Gb hd, 66mb ram, DVD-burner and WLAN.
A1200 desktop, OS3.9, Blizzard 060 66mhz, 66mb ram, Ide Fix Express with 160Gb HD and WLAN
A500 OS2.1, GVP+HD8 with 4mb ram, 1mb chip ram and 4gb HD
Commodore CDTV KS3.1, 1mb chip, 4mb fast ram and IDE HD
 

Offline utri007

Re: NetSurf OS3 - testing!
« Reply #111 on: September 20, 2020, 10:30:07 PM »
Lattest 68k version seems to be broken. NetSurf-gcc-5199.lha https://ci.netsurf-browser.org/builds/amigaos3/

Bitmap fonts doesn't display
Magic Menu doesn't work and eventually opening menus traditional way will crash computer.

ACube Sam 440ep Flex 800mhz, 1gb ram and 240gb hd and OS4.1FE
A1200 Micronic tower, OS3.9, Apollo 060 66mhz, xPert Merlin, Delfina Lite and Micronic Scandy, 500Gb hd, 66mb ram, DVD-burner and WLAN.
A1200 desktop, OS3.9, Blizzard 060 66mhz, 66mb ram, Ide Fix Express with 160Gb HD and WLAN
A500 OS2.1, GVP+HD8 with 4mb ram, 1mb chip ram and 4gb HD
Commodore CDTV KS3.1, 1mb chip, 4mb fast ram and IDE HD
 

Offline utri007

Re: NetSurf OS3 - testing!
« Reply #112 on: September 22, 2020, 10:02:14 PM »
last working version is 5152, build date 24.05.2020. Hope this helps.

It is really slow. Amigaworld.net wich hasn’t change a bit, used to render 14 seconds, not it renders 192 seconds. Some sort of SSL strip proxy solution would be nice.
ACube Sam 440ep Flex 800mhz, 1gb ram and 240gb hd and OS4.1FE
A1200 Micronic tower, OS3.9, Apollo 060 66mhz, xPert Merlin, Delfina Lite and Micronic Scandy, 500Gb hd, 66mb ram, DVD-burner and WLAN.
A1200 desktop, OS3.9, Blizzard 060 66mhz, 66mb ram, Ide Fix Express with 160Gb HD and WLAN
A500 OS2.1, GVP+HD8 with 4mb ram, 1mb chip ram and 4gb HD
Commodore CDTV KS3.1, 1mb chip, 4mb fast ram and IDE HD
 

Offline utri007

Re: NetSurf OS3 - testing!
« Reply #113 on: September 23, 2020, 06:44:23 AM »
Problem was codesets library, if there is a one, it needs to be lattest. I t works without it, but not with older versions.
ACube Sam 440ep Flex 800mhz, 1gb ram and 240gb hd and OS4.1FE
A1200 Micronic tower, OS3.9, Apollo 060 66mhz, xPert Merlin, Delfina Lite and Micronic Scandy, 500Gb hd, 66mb ram, DVD-burner and WLAN.
A1200 desktop, OS3.9, Blizzard 060 66mhz, 66mb ram, Ide Fix Express with 160Gb HD and WLAN
A500 OS2.1, GVP+HD8 with 4mb ram, 1mb chip ram and 4gb HD
Commodore CDTV KS3.1, 1mb chip, 4mb fast ram and IDE HD
 

Offline chris

Re: NetSurf OS3 - testing!
« Reply #114 on: October 09, 2020, 11:46:34 AM »
Problem was codesets library, if there is a one, it needs to be lattest. I t works without it, but not with older versions.

Yeah, for some reason they added features and didn't bump the version, only the revision.
"Miracles we do at once, the impossible takes a little longer" - AJS on Hyperion
Avatar picture is Tabitha by Eric W Schwartz
 

Offline utri007

Re: NetSurf OS3 - testing!
« Reply #115 on: October 09, 2020, 12:32:15 PM »
If possible, you could disable it. Using codesets.library with 68k Amiga makes Netsurf loose a speed a lot. Something like 300%.

ACube Sam 440ep Flex 800mhz, 1gb ram and 240gb hd and OS4.1FE
A1200 Micronic tower, OS3.9, Apollo 060 66mhz, xPert Merlin, Delfina Lite and Micronic Scandy, 500Gb hd, 66mb ram, DVD-burner and WLAN.
A1200 desktop, OS3.9, Blizzard 060 66mhz, 66mb ram, Ide Fix Express with 160Gb HD and WLAN
A500 OS2.1, GVP+HD8 with 4mb ram, 1mb chip ram and 4gb HD
Commodore CDTV KS3.1, 1mb chip, 4mb fast ram and IDE HD
 

Offline chris

Re: NetSurf OS3 - testing!
« Reply #116 on: October 09, 2020, 12:41:26 PM »
I can add a setting to disable it.  I'd be surprised if it was really that much slower, but possibly the way I'm using it isn't optimal.  I'll see if I can cache the internal structures for the charsets as it might be looking them up each time.
"Miracles we do at once, the impossible takes a little longer" - AJS on Hyperion
Avatar picture is Tabitha by Eric W Schwartz
 

Offline chris

Re: NetSurf OS3 - testing!
« Reply #117 on: October 09, 2020, 01:08:05 PM »
I haven't tested it at all, but try build 5216.
"Miracles we do at once, the impossible takes a little longer" - AJS on Hyperion
Avatar picture is Tabitha by Eric W Schwartz
 

Offline utri007

Re: NetSurf OS3 - testing!
« Reply #118 on: October 10, 2020, 01:58:34 AM »
It has now a problem without codesets.library and it introduced new graphical problem.

Would it be possible to revert back few years old version so that new amissl would work with it?  Those versions were almost usefull, these versions.. it doesn't matter if you fix or not these problems, there is no point to use it as it loads page like amigaworld.net about 5 minits. It used to load about 15 seconds, when Olaf fixed memory fragmentation issue three? years ago.
« Last Edit: October 10, 2020, 02:01:34 AM by utri007 »
ACube Sam 440ep Flex 800mhz, 1gb ram and 240gb hd and OS4.1FE
A1200 Micronic tower, OS3.9, Apollo 060 66mhz, xPert Merlin, Delfina Lite and Micronic Scandy, 500Gb hd, 66mb ram, DVD-burner and WLAN.
A1200 desktop, OS3.9, Blizzard 060 66mhz, 66mb ram, Ide Fix Express with 160Gb HD and WLAN
A500 OS2.1, GVP+HD8 with 4mb ram, 1mb chip ram and 4gb HD
Commodore CDTV KS3.1, 1mb chip, 4mb fast ram and IDE HD
 

Offline chris

Re: NetSurf OS3 - testing!
« Reply #119 on: October 10, 2020, 12:17:29 PM »
Quote
It has now a problem without codesets.library and it introduced new graphical problem.

I haven't touched the non-codesets code so that must be caused by something else.

Probably where the code is now offset differently the old bug causing it to access memory it shouldn't be is having a different effect.

This hasn't answered the question I wanted answered which was "is it any quicker now *with codesets.library*?"

Quote
Would it be possible to revert back few years old version so that new amissl would work with it?  Those versions were almost usefull, these versions.. it doesn't matter if you fix or not these problems, there is no point to use it as it loads page like amigaworld.net about 5 minits. It used to load about 15 seconds, when Olaf fixed memory fragmentation issue three? years ago.

I have no idea what caused it to go slow as I was told about this several months after it happened.  I have absolutely no chance of figuring out why now as I couldn't even at the time.

The memory model hasn't changed since then.

It's probably worth adding log_filter:level:CRITICAL to Choices because by default it is spewing out loads of Javascript-related errors which will be slowing it down.

I did switch to using AmiSSL as OpenSSL stopped working, I can't revert that change as not being able to make secure connections makes the browser useless.  There's no reason why AmiSSL should be slower than OpenSSL anyway, it's 99% the same thing.

What it needs is for somebody who can actually debug it properly to figure out what is causing the suspected buffer overflow problem and fix that.  That gets rid of the weirdness of it randomly crashing and doing odd things.  Then it's worth spending a bit of time optimising (again, by somebody who knows what they are doing).

« Last Edit: October 10, 2020, 12:24:09 PM by chris »
"Miracles we do at once, the impossible takes a little longer" - AJS on Hyperion
Avatar picture is Tabitha by Eric W Schwartz