OS2 World Community Forum

OS/2, eCS & ArcaOS - Technical => Internet => Topic started by: Dave Yeo on January 16, 2016, 04:55:27 am

Title: Experimental SeaMonkey build
Post by: Dave Yeo on January 16, 2016, 04:55:27 am
I've uploaded an experimental build of SM 2.28 to https://bitbucket.org/dryeo/dry-comm-esr31/downloads (https://bitbucket.org/dryeo/dry-comm-esr31/downloads).
This build is optimized for Core2Duo and should work on most CPU's less then a decade old with best performance on Intel chips after P4, probably good performance on AMD as well, untested. If you get a sigill, your CPU is too old.
This is also linked against the new Fontconfig currently available in the Netlabs RPM experimental repository.
Other then above, same code, same requirements.
Title: Re: Experimental SeaMonkey build
Post by: muffetta on January 16, 2016, 07:21:39 am
Great job Dave, thanks for your efforts!
Unfortunately, still crash opening "Download Manager" window, and sometimes i got the "Flash plugin! - Failed to load Flash Win32 plugin." window.
Title: Re: Experimental SeaMonkey build
Post by: Greggory Shaw on January 16, 2016, 08:57:48 pm
HI Dave !

On new install of eCS 2.2 /w flash is working pretty good so far ! Only TRP was on Facebook, but that happens on all Mozilla apps.

The "Download Manager" works here too.


Thanks for the hard work ! Do you have a version of Enigmail that works with TB 31 ?

Greggory
Title: Re: Experimental SeaMonkey build
Post by: Dave Yeo on January 17, 2016, 04:40:24 am
@Barb. Sorry to hear the Download Manager crash is still there. I thought that some of the subtle bug fixes that dmik did fixed it. I notice you're using SNAP, does the crash happen in safe mode or with MOZ_ACCELERATED=0 set?
You do know about the workaround of not using the DownLoad Manager? Preferences-->Browser-->Downloads and click use a progress bar.

@Greggory, no I don't have a version of Enigmail that works with newer TB or SM. The problem is that Enigmail now uses CTYPES to load functions from native DLLs and I'm probably not skillful enough to write the JavaScript to open libc066.dll and use the libc functions to open and run PGP.

One thing I was curious about is how the fonts look for everyone?
Title: Re: Experimental SeaMonkey build
Post by: Greggory Shaw on January 17, 2016, 07:31:51 am
@Barb. Sorry to hear the Download Manager crash is still there. I thought that some of the subtle bug fixes that dmik did fixed it. I notice you're using SNAP, does the crash happen in safe mode or with MOZ_ACCELERATED=0 set?
You do know about the workaround of not using the DownLoad Manager? Preferences-->Browser-->Downloads and click use a progress bar.

It's fixed here and thx for the hard work !

Quote
@Greggory, no I don't have a version of Enigmail that works with newer TB or SM. The problem is that Enigmail now uses CTYPES to load functions from native DLLs and I'm probably not skillful enough to write the JavaScript to open libc066.dll and use the libc functions to open and run PGP.

One thing I was curious about is how the fonts look for everyone?

To bad Enigmail :( - but yes the fonts look good !


Greggory
Title: Re: Experimental SeaMonkey build
Post by: muffetta on January 17, 2016, 08:33:12 am
@Barb. Sorry to hear the Download Manager crash is still there. I thought that some of the subtle bug fixes that dmik did fixed it. I notice you're using SNAP, does the crash happen in safe mode or with MOZ_ACCELERATED=0 set?
You do know about the workaround of not using the DownLoad Manager? Preferences-->Browser-->Downloads and click use a progress bar.
Crash is still there, both 1/0 MOZ_ACCELERATED set, SNAP, PANO and old GRADD
One thing I was curious about is how the fonts look for everyone?
Looks right.
Title: Re: Experimental SeaMonkey build
Post by: muffetta on January 17, 2016, 08:40:56 am
It's fixed here and thx for the hard work !
Quote
Yes, i know Greggory, crash will occurr as soon as your downloaded file list will be too long, and of course, using the progrbar is a workaround for it, otherwhise, keep it quite short.
Title: Re: Experimental SeaMonkey build
Post by: muffetta on January 17, 2016, 08:45:25 am
It's fixed here and thx for the hard work !
Yes, i know Greggory, the crash will occurr as soon as your downloaded file list will be too long, and of course, using the progrbar is a workaround for it, otherwhise, keep it quite short.
Title: Re: Experimental SeaMonkey build
Post by: muffetta on January 17, 2016, 08:50:54 am
@Barb. Sorry to hear the Download Manager ...
Dave, here also the "Go back one page" button and the "Go forward one page" on the "Navigation Toolbar" are not available wile browsing, the "Back space" key instead works, while "Alt+LeftArrow" and "Alt+RightArrow" keys not.
Suggestions?
Title: Re: Experimental SeaMonkey build
Post by: Dave Yeo on January 17, 2016, 08:03:56 pm
That's strange. can you right click on the navigation bar (might have to try a couple of spots), choose Customize and drag them back or click restore default set?
Everything works fine here and I've been using this build for quite a while.
Title: Re: Experimental SeaMonkey build
Post by: muffetta on January 17, 2016, 08:36:18 pm
That's strange. can you right click on the navigation bar (might have to try a couple of spots), choose Customize and drag them back or click restore default set?
Everything works fine here and I've been using this build for quite a while.
Just after restore default set.
Yes, quite strange  :D
Maybe switching to a new, clean profile will solve.
Title: Re: Experimental SeaMonkey build
Post by: muffetta on January 18, 2016, 11:20:12 am
Maybe switching to a new, clean profile will solve.
..Indeed. Thanks to all!
Title: Re: Experimental SeaMonkey build
Post by: Dave Yeo on January 18, 2016, 08:03:32 pm
Did you try safe mode with your old profile? With luck it is an extension problem.
Title: Re: Experimental SeaMonkey build
Post by: muffetta on January 18, 2016, 09:24:39 pm
Did you try safe mode with your old profile? With luck it is an extension problem.
Honestly no, I do not have a lot of extensions and I simply export mails and bookmarks, i create a new profile, then resubmit the old data, fortunatly i never got issues.
Title: Re: Experimental SeaMonkey build
Post by: Andy Willis on January 21, 2016, 09:16:13 pm
Anyone using Chatzilla?  It crashes here shortly after start.  I have not tried a new profile yet.
Actually, just gave a new profile a shot and it crashes even faster... previously it would start to load the servers in the list, with new profile it crashes immediately on launch of IRC.
Title: Re: Experimental SeaMonkey build
Post by: Dave Yeo on January 22, 2016, 01:53:02 am
Seems to work fine here though I only have a couple of channels open. Are you getting a trp file? Is it crashing in an xmms register? And of course, what CPU, there's cpuinfo.zip on Hobbes to get all the CPU flags, handy for comparing which extensions are supported and at least one Phenom user had a sigill as Phenom supports SSE4a but not SSSE3 while my CPU is the opposite.
Title: Re: Experimental SeaMonkey build
Post by: Andy Willis on January 22, 2016, 08:17:51 pm
Seems to work fine here though I only have a couple of channels open. Are you getting a trp file? Is it crashing in an xmms register? And of course, what CPU, there's cpuinfo.zip on Hobbes to get all the CPU flags, handy for comparing which extensions are supported and at least one Phenom user had a sigill as Phenom supports SSE4a but not SSSE3 while my CPU is the opposite.
I don't know how to read the trp file but it does create one... it crashes in xul.dll on an invalid address is all I can tell for sure.  Here is the cpu information from cpuid and cpuinfo and a few lines from the trp files that seemed to stand out to me as maybe helpful.  I can send the whole thing if useful.

[e:\download]cpuid
CPU Vendor String: "GenuineIntel"

CPU supports MMX instructions
CPU supports SSE instructions
CPU supports SSE2 instructions
CPU supports SSE3 instructions
CPU supports SSSE3 instructions
CPU supports SSSE4.1 instructions
[e:\download\cpu\cpuinfo]cpuinfo.exe
vendor_id       : GenuineIntel
cpu family      : 6
model           : 23
stepping        : 6
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov
pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe pni dtes64 monitor ds_
cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 lm lahf_lm
cpu MHz         : 2500.105
model name      : Intel(R) Core(TM)2 Duo CPU     T9400  @ 2.53GHz

 Filename: E:\BROWSERS\WARPZILLA\XUL.DLL 12/25/2015 14:34:00 35,707,368
 Address:  005B:14F31D63 (0001:009B1D63)
 Cause:    Attempted to read from 0000001C
           (not a valid address)

______________________________________________________________________

 Failing Instruction
______________________________________________________________________

 14F31D54  SUB  ESP, 0x4c              (83ec 4c)
 14F31D57  MOV  EBX, [ESP+0x68]        (8b5c24 68)
 14F31D5B  MOV  EDI, [ESP+0x60]        (8b7c24 60)
 14F31D5F  MOV  ESI, [ESP+0x6c]        (8b7424 6c)
 14F31D63 >TEST BYTE [EBX+0x1c], 0x40  (f643 1c 40)
 14F31D67  JNZ  0x14f31d79             (75 10)
 14F31D69  MOV  DWORD [ESP+0x4], 0x0   (c74424 04 00000000)
 14F31D71  MOV  [ESP], EBX             (891c24)

 Call Stack
______________________________________________________________________

   EBP     Address    Module     Obj:Offset    Nearest Public Symbol
 --------  ---------  --------  -------------  -----------------------
 Trap  ->  14F31D63   XUL       0001:009B1D63  between gfxFontGroup::TryAllFamilyMembers + 13 and gfxFontGroup::WhichSystemFontSupportsChar - 19D  (both in gfxFont.cpp)

 Lost Stack chain - invalid EBP FFFF7F00
Title: Re: Experimental SeaMonkey build
Post by: Dave Yeo on January 23, 2016, 12:47:48 am

 Filename: E:\BROWSERS\WARPZILLA\XUL.DLL 12/25/2015 14:34:00 35,707,368
 Address:  005B:14F31D63 (0001:009B1D63)
 Cause:    Attempted to read from 0000001C
           (not a valid address)

______________________________________________________________________

 Failing Instruction
______________________________________________________________________

 14F31D54  SUB  ESP, 0x4c              (83ec 4c)
 14F31D57  MOV  EBX, [ESP+0x68]        (8b5c24 68)
 14F31D5B  MOV  EDI, [ESP+0x60]        (8b7c24 60)
 14F31D5F  MOV  ESI, [ESP+0x6c]        (8b7424 6c)
 14F31D63 >TEST BYTE [EBX+0x1c], 0x40  (f643 1c 40)
 14F31D67  JNZ  0x14f31d79             (75 10)
 14F31D69  MOV  DWORD [ESP+0x4], 0x0   (c74424 04 00000000)
 14F31D71  MOV  [ESP], EBX             (891c24)

 Call Stack
______________________________________________________________________

   EBP     Address    Module     Obj:Offset    Nearest Public Symbol
 --------  ---------  --------  -------------  -----------------------
 Trap  ->  14F31D63   XUL       0001:009B1D63  between gfxFontGroup::TryAllFamilyMembers + 13 and gfxFontGroup::WhichSystemFontSupportsChar - 19D  (both in gfxFont.cpp)

 Lost Stack chain - invalid EBP FFFF7F00

Seems to be getting a bad font. The new Fontconfig will try more fallbacks when it can't find a match so that may be the problem.
Someone else had a similar crash which was for the same reason or a SMP problem. The Cairo font code has lots of warnings about being careful in a multi-threaded app and the different optimization probably changes some timings. One thing is to make sure using the latest Freetype as upstream did a bunch of thread safety fixes and you could always do a quick test with only one core enabled.
Hopefully these problems will go away when we switch to the Pango font route.
BTW, I ran Chatzilla for a few hours last night with no problems.