Intel Mesa Gives Problems With KDE's KWin, Again

[energyman]

intel driver devs fuck up all the time. Just look up Linus' rants whenever they drop a load of crap into the kernel.

They could also learn a lot from the kernel: never change behaviour expected by the userspace.

Such a change in a point release is more than stupid. Once again intel's devs fucked up.

[marek]

Oh right, so now it's the kwin developers' fault that the free graphics driver developers fucked up once again?

They didn't fuck up at all. The Intel devs only changed the renderer string. It's a valid change, a normal one, it may happen from time to time and apps shouldn't rely on it. On the other hand, detecting KMS/GEM/DRI2 via some string in OpenGL is clearly a hack. There are better ways of detecting those than some random string that happened to contain "GEM" by luck. (I am surprised too that someone parses that string.)

Anyway, the fact the driver devs haven't tested it is, of course, their fault. But the real problem lies in the Kwin code, Martin has even admitted that with the code snippet in his blog post. It's pretty clear that hacks like that would cause breakage sooner or later.

[pingufunkybeat]

Marek, how would you test correctly if a driver provides appropriate functionality?

It is known since the last KWin fiasco that you can't trust the advertised functionality, and checking for direct rendering is also a hack which doesn't even work anymore. Blacklists caused their own set of problems, and if there is no check, KWin will crash all the time, like 4.5 did.

What is the correct way to check this?

[TheBlackCat]

There are better ways of detecting those than some random string that happened to contain "GEM" by luck.

Such as...?

[Nille]

Kwin do it again

BTW. AMD/ATI is not mentioned because no one from KDE test there Stuff on AMD/ATI and if only on an old Code Base.

[mdias]

wtf.. checking on a renderer string to make assumptions? realy!?

I believe the best way to handle this is to make the DE check the speed and availability of the needed operations on the active driver whenever it changes.

[pingufunkybeat]

I believe the best way to handle this is to make the DE check the speed and availability of the needed operations on the active driver whenever it changes.

It does that already. And disables compositing if it notices that it is too slow.

[mdias]

It does that already. And disables compositing if it notices that it is too slow.

Oh, sorry, I didn't know that since I don't use KDE.

But then why do they check on the renderer string? To make extra sure? I don't get it, sorry, it's a genuine question.

[NeoBrain]

Oh, sorry, I didn't know that since I don't use KDE.

But then why do they check on the renderer string? To make extra sure? I don't get it, sorry, it's a genuine question.

KWin needs to know which effects can savely be enabled and which cause glitches or even crashes. You can check the OpenGL extensions to do that OR use feature tests. KWin used the latter ones previously, but apparently even the functionality tests caused crashes for some drivers. Thus, KWin now checks for the OGL render string and only enables per-driver and per-driver-version hardcoded feature sets. Which is kinda lame, but there's no other way to do it until the drivers get fixed properly (and are reliable enough to enable functionality checks again)..

[mdias]

KWin needs to know which effects can savely be enabled and which cause glitches or even crashes. You can check the OpenGL extensions to do that OR use feature tests. KWin used the latter ones previously, but apparently even the functionality tests caused crashes for some drivers. Thus, KWin now checks for the OGL render string and only enables per-driver and per-driver-version hardcoded feature sets. Which is kinda lame, but there's no other way to do it until the drivers get fixed properly (and are reliable enough to enable functionality checks again)..

Understood, even though I agree that it's lame to do that... Intel users should just complain to their driver devs if their drivers don't work...

Well, I can't complain that KDE people are worried about my experience, but I still think that's just wrong...

Thank you for the info.