Qt6 Application Testing

[Pete]

Hi Eugene

Unzip QT6 package into a directory, eg d:\apps, keeping the directory structure - avoid mixing with existing *nix directory structure
Use a CMD file to set environment and start a qt6 app - see earlier post by Remy on page 39
Keep reading this thread for fine tuning/future releases - Martin updates page 1 with new releases


Hope the above helps - and maybe others can "fine tune" the above

Pete

[Remy]

Does QT6, Dooble use some process involving timers like the
SET NSPR_OS2_NO_HIRES_TIMER=1 used for mozilla ? 

[Paul Smedley]

Unzip QT6 package into a directory, eg d:\apps, keeping the directory structure - avoid mixing with existing *nix directory structure
Use a CMD file to set environment and start a qt6 app - see earlier post by Remy on page 39
Keep reading this thread for fine tuning/future releases - Martin updates page 1 with new releases

@Martin perhaps we need a wiki page on setting up Qt6 that could be referenced from page 1? Also a copy of Remy's script  to start dooble on page 1?

[Remy]

Hi Paul !

I could get 2 trp files using my priority settings (close when runing several times the nperf process from nperf selecting a server instead of using default one)
Do you find them usefull ?

Regards
 

[Paul Smedley]

Hi Paul !

I could get 2 trp files using my priority settings (close when runing several times the nperf process from nperf selecting a server instead of using default one)
Do you find them usefull ?

I'll take a look later in the week....

[Steven Levine]

Remy,

Are you using Paul's libcn0 build with the silent exit patch?  The trap appears to be in the code that implements the exceptq interface.

[Remy]

Remy,

Are you using Paul's libcn0 build with the silent exit patch?  The trap appears to be in the code that implements the exceptq interface.

Yes I use it else no elements

[roberto]

Hello
I've been testing, the new versions of qt6-626, and I'm not convinced, they work worse than the version 632,
 I can watch the videos for a while, from 30 minutes to 1 hour and ten minutes.
 I agree with everyone else in their comments on version 626.
 I have tried to install the 632.
 My mistake :
 [113:18:0920/200539.025000:ERROR:cert_verify_proc_builtin.cc(600)] No net_fetche
r for performing AIA chasing.

But looking for the error that I get on console I found this link:
https://github.com/nwjs/nw.js/issues/8040

And he says that the culprit of that mistake is libcups
https://github.com/nwjs/nw.js/issues/8032

In case it helps them.
Saludos

[Steven Levine]

Remy,

Are you sure that you did not get corresponding exceptq reports written to %LOGFILES%\app?  Normally, the silent exit code will generate an exceptq report with the description

   Exception 71785158 - Exceptq Debug Request

The file name will be similar to

  4294967295-001e_14-dooble-exceptq.txt

[Remy]

Remy,

Are you sure that you did not get corresponding exceptq reports written to %LOGFILES%\app?  Normally, the silent exit code will generate an exceptq report with the description

   Exception 71785158 - Exceptq Debug Request

The file name will be similar to

  4294967295-001e_14-dooble-exceptq.txt

No *dooble-exceptq.txt  at TRP date

Into the TRP, I could see many entries having time or timing text field (I do not know how to interpret this) but it looks like all have a ref to base_jumbo* )

[David McKenna]

 Been using Paul's latest QT6 with Dooble successfully as a Win11 VirtualBox 7.1.0 guest in ArcaOS 5.1 for awhile, with Remy's script and acpi powerman disabled it works the best it ever has. But it is not perfect. Never had a system hang yet, rarely a Dooble hang, very occasionally a silent exit, but the biggest problem I see is having a web page (usually heavy javascript) stop rendering, the mouse scroll wheel stops working (although I can grab the scroll bar with the mouse and scroll) and one of the 4 CPU's will show 100% and pass that 100% back and forth among the 4 CPU's. The system keeps working no problem, but Dooble at that point is no longer usable and I have to close it.  Then start it again, go back to the same page and it works fine.

  Does this seems like some kind of 'timing' or 'deadlock' issue? I'm curious if anyone else sees this phenomenon, and wonder what might cause it.

Regards,

[Remy]

May it possible to get a webengine build with -no-webengine-jumbo-build ?
Jumbo seems no more been supported by google if found information are good ones
(Due I see a lot of jumbo ref into the trp)

[Dave Yeo]

May it possible to get a webengine build with -no-webengine-jumbo-build ?
Jumbo seems no more been supported by google if found information are good ones
(Due I see a lot of jumbo ref into the trp)

Looking, if I understand correctly, jumbo-build is simply concatenating source files together before compiling them. Not using it would more then double build time, probably use less memory during compile but more during linking. It's linking where we have to be careful not to need more then 3.5GB's of ram to link.
The only benefit to users would be slightly more readable trp files, source file names instead of jumbo names.
Mozilla has a similar thing, unified builds, 4-8 source files concatenated together, sped up compiling and especially linking. Harder to read trp files.
It also seems that the QT Webengine did not remove the option in the interests of faster builds. Google itself has a pretty powerful build farm and is trying to introduce other build speedups.

[Remy]

What do you think about this ?
https://github.com/bitwiseworks/qtwebengine-chromium-os2/issues/27

[Paul Smedley]

May it possible to get a webengine build with -no-webengine-jumbo-build ?
Jumbo seems no more been supported by google if found information are good ones
(Due I see a lot of jumbo ref into the trp)

Not going to happen, for all the reasons  Dave mentions. For the end user,  jumbo-build makes no difference - end users aren't expected to intrerpret TRP files.