Qt6 Development

[Paul Smedley]

OK I got a DLL and some TRPs to go with it (attached)

The maps.exe one seems to be OpenGL related (not surprisingly) the second looks more straightforward

[Paul Smedley]

About 500 files to go and I get the following:

Code: [Select]

{1}[u:\dev\qt6-webengine-os2-6.5.x\build\src\core\release\i386] mksnapshot --tur
bo_instruction_scheduling --target_os=os2 --target_arch=x86 --embedded_src gen/v
8/embedded.S --embedded_variant Default --random-seed 314159265 --startup_src ge
n/v8/snapshot.cc --no-native-code-counters
Assertion failed: _UM_LUMP_STATUS (olump) == _UMS_FREE, file ./libc-0.1.10/src/emx/src/lib/malloc/ialloc.c, line 116

Killed by SIGABRT
pid=0x33b0 ppid=0x0041 tid=0x0001 slot=0x007d pri=0x0200 mc=0x0001 ps=0x0010
U:\DEV\QT6-WEBENGINE-OS2-6.5.X\BUILD\SRC\CORE\RELEASE\I386\MKSNAPSHOT.EXE
Creating 33B0_01.TRP
Moved 33b0_01.TRP to C:\var\log\app\64313478-33b0_01-MKSNAPSHOT-exceptq.txt

Edit: Cheated for now and replaced mksnapshot with the version from Qt 6.4.x

OK so this wasn't the best solution. Worked through a couple of other issues and now get:

Code: [Select]

# Fatal error in , line 0
# Version mismatch between V8 binary and snapshot.
#   V8 binary version: 10.8.168.25
#    Snapshot version: 10.2.154.17
# The snapshot consists of 288000 bytes and contains 1 context(s).
#
#
#
#FailureMessage Object: 0x12f104
Killed by SIGTRAP


[Andy Willis]

xpdf is working with 6.5

Hi Andy.
Please post the binaries to test it out. I don't know where those are.

Regards

Ha!  I was able to sneak something past you!  :-) :-)
https://www.os2world.com/forum/index.php/topic,3252.msg38877/topicseen.html#msg38877

[Paul Smedley]

Sometimes it pays to not look at something for a few days. Finally, I noticed some differences that needed implementing in page_allocator_internals_os2.h and got past the previous trap (in Qt 6.4.3).

Commits in https://github.com/psmedley/qt6-webengine-os2/commit/f49a4993adc8f58091a7396dfe91dda1252ce874 - the key fix was the first section, adding the additional cases
New TRP attached which I'll investigate shortly.

[Paul Smedley]

Some analysis from Steven on the latest trp that I'm still digesting...

Code: [Select]

>> Is this an alignment problem?
> Not sure yet. The trap appears to be happening at
> https://github.com/psmedley/qt6-webengine-os2/blob/6.4.x/src/3rdparty/chromium/third_party/blink/renderer/platform/graphics/parkable_image_manager.cc#L33

> but not sure why yet...

FWIW, the caller is:

  src\3rdparty\chromium\third_party\blink\renderer\platform\exported\platform.cc:251
    base::trace_event::MemoryDumpManager::GetInstance()->RegisterDumpProvider(
        &ParkableImageManager::Instance(), "ParkableImages",
        base::ThreadTaskRunnerHandle::Get());

I think the generated code in the constructor is trying implement a "fast
memset" to clear the instance object.  The PXOR clears the XMM0 and the
MOVAPS instructions zero 16-byte chunks of memory.  This is going to be
faster than the rep movsd we typically see.

The problem here is that 0xfb31248 needs to be 16-byte aligned and it's
not.

[Paul Smedley]

Not sure where else to post it, but I'm in the midst of changing ISP so the website is currently offline. Will hopefully be back online in the next few hours.

[Paul Smedley]

Should be back up now, DNS has been updated

[Martin Iturbide]

Hi Paul

I confirm I see your website: https://os2ports.smedley.id.au/
... and you Mantis page. https://mantis.smedley.id.au

Regards

[Paul Smedley]

Some analysis from Steven on the latest trp that I'm still digesting...

Code: [Select]

>> Is this an alignment problem?
> Not sure yet. The trap appears to be happening at
> https://github.com/psmedley/qt6-webengine-os2/blob/6.4.x/src/3rdparty/chromium/third_party/blink/renderer/platform/graphics/parkable_image_manager.cc#L33

> but not sure why yet...

FWIW, the caller is:

  src\3rdparty\chromium\third_party\blink\renderer\platform\exported\platform.cc:251
    base::trace_event::MemoryDumpManager::GetInstance()->RegisterDumpProvider(
        &ParkableImageManager::Instance(), "ParkableImages",
        base::ThreadTaskRunnerHandle::Get());

I think the generated code in the constructor is trying implement a "fast
memset" to clear the instance object.  The PXOR clears the XMM0 and the
MOVAPS instructions zero 16-byte chunks of memory.  This is going to be
faster than the rep movsd we typically see.

The problem here is that 0xfb31248 needs to be 16-byte aligned and it's
not.

OK I got a potential patch for this overnight, unfortunately I'd made a separate change which triggered a partial rebuild which is almost complete and I will test.

[Paul Smedley]

Hi Paul

I confirm I see your website: https://os2ports.smedley.id.au/
... and you Mantis page. https://mantis.smedley.id.au

Regards

Thanks - unfortunately IPv4 only at this stage, but working on getting an IPv6 tunnel setup.

[Paul Smedley]

Not the most exciting screenshot in the world, but this is the first page successfully rendered in QtWebEngine 6.4.3

This is only possibile due to a dirty hack in the code, and any page more complex than this simple HTML only page results in another crash....

Just wanted to show some progress

[Paul Smedley]

Actually, with javascript disabled - it's quite stable - not terribly useful, but stable...

[Paul Smedley]

OK, the TRPs with 6.4.3 and 6.5.0 are driving me nuts. I'm going to try and stabilise the qt 6.3.2 webengine and fix and traps there. That might shed some light on stuff I've missed...

[Mentore]

OK, the TRPs with 6.4.3 and 6.5.0 are driving me nuts. I'm going to try and stabilise the qt 6.3.2 webengine and fix and traps there. That might shed some light on stuff I've missed...

Most probably. I second your idea. Keep letting us know how things are going.
Mentore

[Martin Iturbide]

OK, the TRPs with 6.4.3 and 6.5.0 are driving me nuts. I'm going to try and stabilise the qt 6.3.2 webengine and fix and traps there. That might shed some light on stuff I've missed...

Hi. Sounds like a great plan to me. Just keep pushing out the 6.3.2 releases and I will try it out.
Regards