Messages

16

Internet / otter-browser, Qt5 build

« on: September 15, 2024, 04:37:47 am »

Ok, up till now when I've built Otter, it crashes on startup. Andy reported success with Dooble and not running acpi-daemon. Testing, the otter browser starts up and runs fairly well without acpi-daemon running. It is crashy, I tried this post in Otter and probably uploading the file crashed the browser.
Anyways, for testing purposes, here is today's build of the Otter browser from the main repository, no OS/2 patches yet, it runs surprisingly well considering. It is using the same webengine as Dooble-Qt5 so some pages will not display correctly.
Edit: fix the markup

17

Programming / Re: Qt6 Application Testing

« on: September 15, 2024, 04:15:38 am »

After killing acpidaemon, both my Qt5 and your Qt6 builds of Otter run fine. Thanks to Andy for the suggestion, he also mentions that Dooble doesn't hang without apcpdaemon running, even without single-process. More testing to do here.
OK, YouTube comes up slowly, UI displays, video does not run. Closing otter results in a sigsegv.
Edit: the down side of not running acpidaemon is the computer runs hot, 64C instead of about 25C at idle.

Ok, retrying the Qt6 build of Otter sees the start page, DuckDuckGo, display but nothing else seems to draw the canvas. Also when hovering over a tab, the preview  thumbnail shows up in the upper left of the desktop. The Qt5 build, which I'm typing this in, the preview shows up over the tab. Niether seems to actually draw the preview.
I'll start a new thread on the Qt5 Otter browser.
Edit: Another difference is the Qt5 build, after a crash, starts up with a session restore window. The Qt6 build doesn't.

18

Programming / Re: Qt6 Application Testing

« on: September 15, 2024, 03:36:30 am »

After killing acpidaemon, both my Qt5 and your Qt6 builds of Otter run fine. Thanks to Andy for the suggestion, he also mentions that Dooble doesn't hang without apcpdaemon running, even without single-process. More testing to do here.
OK, YouTube comes up slowly, UI displays, video does not run. Closing otter results in a sigsegv.
Edit: the down side of not running acpidaemon is the computer runs hot, 64C instead of about 25C at idle.

19

Programming / Re: Qt6 Application Testing

« on: September 12, 2024, 02:05:48 am »

  The priority change is definitely what helps here - without it I get the usual hangs and silent exits regularly.  Much rarer with the priority change - any idea why that might be? Related to the processor optimization maybe? Or an interaction between that and the processor I'm using? Wish I knew more about the settings too - maybe it can be tweaked even more....

I think it's some issue with thread priorities - perhaps how they're mapped to the OS/2 API results in too low a priority and timing issues. Some of the silent exits Steven and I investigated a while back appeared to be due to variables having been freed whilst another thread was still expecting them to be valid.

I'll have to ask Steven for his thoughts. It's a fantastic discovery - we just need to work out *why* it's required and see if we can fix it in the codebase.

I tried adding PRIORITY=ABSOLUTE to config.sys to see if it helped not giving the foreground app a boost. Didn't help.

20

Programming / Re: Qt6 Application Testing

« on: September 11, 2024, 05:34:29 pm »

I tried the youtube url,
It start playing, then video stops and 2 to 4 seconds later, audio sound stops, I was able to click on an other video immediatly, the new one started to play and then system hung (power off/on to recover)

When playing youtube, the system hung generally occurs a few seconds (or starting an other video) after sound goes off while playing the selected video
It looks like something goes wrong between OS/2 audio and Dooble !  (how will ffmpeg chromium work with pthread disabled ?)     

When a kill just before the hung, saw something strange into the trps !
00AFFF5C  00010027   DOOBLE    0001:00000027  crt0.s#90 __text + 27 0001:00000000 (D:\Temp\cc7CpCIv.s)
00AFFF6C  00010030   DOOBLE    0001:00000030  crt0.s#94 __text + 30 0001:00000000 (D:\Temp\cc7CpCIv.s)
or
00B8FF58  00010021   DOOBLE    0001:00000021  crt0.s#94 __text + 21 0001:00000000 (D:\Temp\ccqVZDcA.s)
00B8FF4C  00010018   DOOBLE    0001:00000018  crt0.s#90 __text + 18 0001:00000000 (D:\Temp\ccqVZDcA.s)
00B8FF5C  00010021   DOOBLE    0001:00000021  crt0.s#94 __text + 21 0001:00000000 (D:\Temp\ccqVZDcA.s)
But not only Dooble !
00B8FFD4  9FE3004B   QT6CHRTS  0001:0000004B  dll0.s#99 __text + 4B 0001:00000000 (D:\Temp\ccvnLdQd.s)

 

Why and from where did it get this information for a drive D:\  ?   I have no D: drive

That's from Paul's system and perfectly normal.

21

Programming / Re: Qt6 Application Testing

« on: September 10, 2024, 06:08:21 pm »

Hi

Maybe a stupid question, but how to use that script from Remy, save it as a sh or cmd
Tellie

A cmd file, here I have it as dooble.cmd

22

Programming / Re: Qt6 Application Testing

« on: September 10, 2024, 05:22:40 pm »

Hi Pete, what is your VIRTUALADDRESSLIMIT set at? Marked most of the needed DLL's to load the code high? Tried a new profile?

23

Programming / Re: Qt6 Application Testing

« on: September 09, 2024, 11:46:12 pm »

Hi Paul!
ill ffmpeg be usable if adding --enable-pthreads option ?

FFmpeg will automatically enable os2threads, less over head then pthreads.

24

Programming / Re: Qt6 Application Testing

« on: September 09, 2024, 07:14:54 am »

Remy's cmd seems to work fairly well, on the 2nd try youtube works as expected for a while then the UI gets kind of stuck needing to close Dooble.

25

Programming / Re: Qt6 Application Testing

« on: September 09, 2024, 02:32:57 am »

Might be an idea to try rebuilding FFmpeg with --disable-avx --disable-avx2 --disable-avx512 --disable-avx512icl.
I find the newer GCC's are way to quick to enable AVX, though if targetting old CPU's, it shouldn't be a problem.

26

Programming / Re: Qt6 Application Testing

« on: September 06, 2024, 07:41:16 am »

Are you running out of memory on these builds? When I built the Qt5 webengine, I had to periodically reboot to defrag the memory.
Even with Mozilla, just building SeaMonkey to test a patch and got,

Code: [Select]

25:53.50 cc1plus.exe: out of memory allocating 65536 bytes after a total of 0 bytes
With a popup with a sys0147 which blocked until I allowed it to kill the program (cc1plus.exe I think)

Code: [Select]

SYS0147: Not enough resources are available to process this command.
The process for which this pop-up is displayed will end.

EXPLANATION: All available memory is in use.  If segment swapping is
active, the swap file may be full.

ACTION: Do one of the following and retry the command:
o  Reduce the number of running programs.
o  Reduce the value of the BUFFERS=, TRACEBUF=, DISKCACHE=, THREADS=,
   RMSIZE=, or DEVICE=VDISK.SYS statement in the CONFIG.SYS file and
   restart the system.
o  Remove unwanted files from the swap file disk and restart the system.
o  Install additional memory on your system.
o  Contact the supplier of the application.

An error with its roots in 16bit land.

Now the VIO window I was building in is frozen so have to reboot eventually to continue on.

27

Setup & Installation / Re: How to get rid of the second boot menue?

« on: September 04, 2024, 10:09:00 pm »

Add the drive letter as an option? My rEFInd menu entry for AOS 5.1 on drive M:

Code: [Select]

# This is an entry for ArcaOS 5.1 in UEFI mode
menuentry ArcaOS-5.1 {
    icon \EFI\refind\icons\os_aos.png
    loader \EFI\OS2\OS2LDR.EFI
    options "M:"
}



28

General Discussion / Re: Re: Do we still need a 64-bit version of OS/2? And a possible roadmap provided ...

« on: September 02, 2024, 07:11:43 pm »

You might want to test with only one cpu enabled. Add /MAXCPU=1 to the PSD=ACPI.PSD line in config.sys.
As for LibreOffice, one of the first things they did when they forked from OpenOffice was remove the OS/2 stuff and since then so much development has happened that re-adding it would be very hard. Even building OpenOffice is next to impossible without the right environment which I understand only one person has.

29

Programming / Re: Qt6 Application Testing

« on: September 01, 2024, 06:07:59 am »

Can't you do things like win32|os2 to follow that path on win32 or os2?

I was wondering the same, will test.

And yes, that works fine.

30

Programming / Re: Qt6 Application Testing

« on: September 01, 2024, 03:03:53 am »

Can't you do things like win32|os2 to follow that path on win32 or os2?

I was wondering the same, will test.