Bug 98869

Created attachment 128216 [details]
screen2

mind doing a git bisect and checking which commit broke it?

I tried to compile mesa-git from aur but it failed. I really won't be able to compile it multiple times on this machine to search which commit broke it. Is this a big problem? Will be this not be resolved if I am not able to do all thing?

13.0.2 the same

If you can't bisect mesa, then providing an apitrace of the workload will allow someone else to bisect it.

Created attachment 128444 [details]
Log from aplication

Created attachment 128445 [details]
apitrace

Similar bug is present in sequel.

Created attachment 128475 [details]
skl retrace

skl shows correct rendering

Created attachment 128497 [details]
apitrace

apitrace: loaded into /usr/bin/apitrace32
apitrace: unloaded from /usr/bin/apitrace32
apitrace: loaded into /home/kacper/.local/share/Steam/steamapps/common/ElectronicSuperJoy/ElectronicSuperJoy
Set current directory to /home/kacper/.local/share/Steam/steamapps/common/ElectronicSuperJoy
Found path: /home/kacper/.local/share/Steam/steamapps/common/ElectronicSuperJoy/ElectronicSuperJoy
Mono path[0] = '/home/kacper/.local/share/Steam/steamapps/common/ElectronicSuperJoy/ElectronicSuperJoy_Data/Managed'
Mono path[1] = '/home/kacper/.local/share/Steam/steamapps/common/ElectronicSuperJoy/ElectronicSuperJoy_Data/Mono'
Mono config path = '/home/kacper/.local/share/Steam/steamapps/common/ElectronicSuperJoy/ElectronicSuperJoy_Data/Mono/etc'
apitrace: tracing to /home/kacper/.local/share/Steam/steamapps/common/ElectronicSuperJoy/ElectronicSuperJoy.trace
apitrace: redirecting dlopen("libGL.so.1", 0x102)
apitrace: attempting to read configuration file: /home/kacper/.config/apitrace/gltrace.conf
apitrace: warning: glVertexPointer: call will be faked due to pointer to user memory (https://github.com/apitrace/apitrace/blob/master/docs/BUGS.markdown#tracing)
apitrace: warning: glTexCoordPointer: call will be faked due to pointer to user memory (https://github.com/apitrace/apitrace/blob/master/docs/BUGS.markdown#tracing)
apitrace: warning: glColorPointer: call will be faked due to pointer to user memory (https://github.com/apitrace/apitrace/blob/master/docs/BUGS.markdown#tracing)
apitrace: unloaded from /home/kacper/.local/share/Steam/steamapps/common/ElectronicSuperJoy/ElectronicSuperJoy

"skl shows correct rendering" What does that mean?

I also tested app with LIBGL_ALWAYS_SOFTWARE=1 and then rendering is correct.

Comment on attachment 128497 [details]
apitrace

Longer apitrace

Still present.

Informacje o komputerze:
    Producent: Unknown
    Model: Unknown
    Rodzaj: Komputer stacjonarny
    Nie wykryto panelu dotykowego
    
Informacje o procesorze:
    Dostawca CPU: GenuineIntel
    Marka procesora: Intel(R) Core(TM)2 CPU         T5500  @ 1.66GHz
    Rodzina CPU: 0x6
    Model CPU: 0xf
    Stepping CPU: 0x6
    Typ CPU: 0x0
    Częstotliwość: 1667 MHz
    2 rdzenie logiczne
    2 rdzenie fizyczne
    HyperThreading:  Nieobsługiwane
    FCMOV:  Obsługiwane
    SSE2:  Obsługiwane
    SSE3:  Obsługiwane
    SSSE3:  Obsługiwane
    SSE4a:  Nieobsługiwane
    SSE41:  Nieobsługiwane
    SSE42:  Nieobsługiwane
    AES: Nieobsługiwane
    AVX: Nieobsługiwane
    CMPXCHG16B: Obsługiwane
    LAHF/SAHF: Obsługiwane
    PrefetchW: Nieobsługiwane
    
Informacje o sieci:
    Szybkość sieci:  
    
Wersja systemu operacyjnego:
    "Manjaro Linux" (64-bitowy)
    Nazwa kernela: Linux
    Wersja kernela: 4.9.2-1-MANJARO
    Dostawca serwera X: The X.Org Foundation
    Wydanie serwera X: 11901000
    Menedżer okien X: Xfwm4
    Wersja biblioteki Steam: <Runtime disabled>
    
Karta graficzna:
    Sterownik: X.Org R300 Project Gallium 0.4 on ATI RC410

    Wersja sterownika: 2.1 Mesa 13.0.3
    Wersja OpenGL: 2.1
    Głębia kolorów pulpitu: 24 bitów na piksel
    Częstotliwość odświeżania monitora: 59 Hz
    Ident. producenta: 0x1002
    ID urządzenia: 0x5a62
    Nie wykryto wersji
    Liczba monitorów: 1
    Liczba logicznych układów graficznych:  1
    Rozdzielczość głównego ekranu:  1280 x 800
    Rozdzielczość pulpitu: 1280 x 800
    Rozmiar głównego ekranu: 13,03" x 8,15" (15,35" diag)
33,1cm x 20,7cm (39,0cm diag)
    Nie wykryto głównego VRAM
    
Karta dźwiękowa:
    Dźwięk: Realtek ALC861-VD
    
Pamięć:
    RAM: 1877 MB
    
Dodatki:
    Język interfejsu: Polski
    JĘZYK: pl_PL.utf8
    Mikrofon:  Not set
    Przedłużka i kabel Steam Controllera: Not set
    Dostępne miejsce na twardym dysku: 54060 MB
    Największy wolny blok na twardym dysku: 39965 MB
    Gogle VR: Nie wykryto
    
Ostatnie raporty niepowodzeń:

(In reply to cosiekvfj from comment #11)
> "skl shows correct rendering" What does that mean?
> 
> I also tested app with LIBGL_ALWAYS_SOFTWARE=1 and then rendering is correct.

This means that the trace renders correctly on Intel Skylake (i965) driver.

I was able to narrow it a bit.

Latest working: lib32-mesa-11.2.2-1-x86_64.pkg.tar.xz
First not working: lib32-mesa-12.0.1-1-x86_64.pkg.tar.xz

Hi. I was searching for r300 in 11.2 and 12.0 branch. And it's look like this five patches are in 12 and not in 11.2. It is likely one of them that breaks the game?


https://cgit.freedesktop.org/mesa/mesa/commit/?h=12.0&id=b7da8fa11d5c6ec71113350eed1959191a7d5990

https://cgit.freedesktop.org/mesa/mesa/commit/?h=12.0&id=84b961dd53a0509a6865d8417301838b34a40096

https://cgit.freedesktop.org/mesa/mesa/commit/?h=12.0&id=172bfdaa9e80342ade3f023f72d455d76713b866

https://cgit.freedesktop.org/mesa/mesa/commit/?h=12.0&id=e382bc649b95aa2ab6e86d60b0520236b2bf2947

https://cgit.freedesktop.org/mesa/mesa/commit/?h=12.0&id=f75a1084349f2332aa080d77acc175ddbe0ab886

172bfdaa9e80342ade3f023f72d455d76713b866 Is cause of the problem!
https://cgit.freedesktop.org/mesa/mesa/commit/?h=12.0&id=172bfdaa9e80342ade3f023f72d455d76713b866

I have compiled couple of times. Here are my results.

b7da8fa11d5c6ec71113350eed1959191a7d5990 working
84b961dd53a0509a6865d8417301838b34a40096 working
e54b2e902aba22f697c0ba8622cd0a905f1edfff working
172bfdaa9e80342ade3f023f72d455d76713b866 not working
f75a1084349f2332aa080d77acc175ddbe0ab886 not working

Created attachment 129459 [details] [review]
hack

I guess not many developers still have r300 hardware, so you'll have to test some patches. Here is one you can try.

Yes of course.
I have tested this patch and it resolved the problem :)
Thank you very much!!!

Can you explain this bug and patch to my? :) What now?

(In reply to cosiekvfj from comment #21)
> Can you explain this bug and patch to my? :) What now?
The patch has been rejected (more or less), see the mailing list for details: https://lists.freedesktop.org/archives/mesa-dev/2017-February/144118.html

You can take the default case statement in the unbyteswap function on LE. That should safely disable the whole thing on LE.

Ah thats what the patch does. I'll reply on mesa-dev.

I can do more testing if needed.

I pushed Grazvydas's patch.

Created attachment 129912 [details]
Piglit dri2 w/o patch

I have done piglit test you guys was talking about on the mailinglist. I can do it again with patch if you want. Also I have rendering issue with dri3 and steam login screen. I think I will make another bug report. Also I want to test hyperz in piglit so it can be flipped on by default.

Created attachment 129930 [details]
Piglit dri2 patched

This is piglit test with patched mesa. If someone want to take a look and write proper patch I will do my best with testing. Also I will provide more test in this bug-reports:
https://bugs.freedesktop.org/show_bug.cgi?id=37724
https://bugs.freedesktop.org/show_bug.cgi?id=37724

https://bugs.freedesktop.org/show_bug.cgi?id=99970

Created attachment 129963 [details]
piglit mesa at e54b2e902aba22f697c0ba8622cd0a905f1edfff

I also tested mesa before the commit which broke my usercase. After earlier tests I had small image artefact on terminal window. Not with this one. :)
If you want more tests or you want me to redo previous tests or if you are unsure about results, I'm happy to help.

I think it's not the end. I think this workaround proposed in this patch is not an actual workaround. I've opened another ticket for this issue, but I think it would be better of to revert back this patch. https://bugs.freedesktop.org/show_bug.cgi?id=101382

OK. I might have said i to early. Patch was merged in
https://www.mesa3d.org/relnotes/17.0.1.html
https://www.mesa3d.org/relnotes/13.0.6.html
https://www.mesa3d.org/relnotes/17.1.0.html