On 19/01/2022 22:48, Stefan Dösinger wrote:
Am Mittwoch, 19. Jänner 2022, 22:51:47 EAT schrieb Gabriel Ivăncescu:
I tested on 32-bit version of Win7 and WinXP in a VM (not real hardware, I don't know if it matters; I doubt it affects focus results, but at least the device reset works I mean).
Hmm, I think WARP was added by Windows 8. If you look at https:// testbot.winehq.org/JobDetails.pl?Key=105331&f201=exe32.report#k201 (w7u_2qxl testbot results for your patch) you'll see that all the tests are effectively skipped.
Yeah it does pass here, and it fails if I change the test conditions (so they are not skipped). I wonder why the device fails to reset in your case or on the testbot, though.
Are you sure the tests did something useful when you ran them on the Win7 and XP VMs?
I also tested this on the original 64-bit Windows 10 version (1507) and it has the same behavior. It's always the same. No idea when it got changed though.
I'm happy to just drop the non-ddraw tests since the regression you are trying to fix is about ddraw only. The ddraw behavior is sensible in some way; d3d8, d3d9 and d3d9ex are all over the place though. That Win10's behavior changed at some point and didn't change back suggests (but doesn't prove) that there aren't any apps that care.
I don't mind, they're at the end of the patch series anyway so can simply be dropped / are independent.
Just to note that d3d8 is a very old API, it's not inconceivable that newer Windows versions can break some apps. Though very unlikely they care about this, so yeah. I don't think d3d8 is even available in 64-bit, looking at testbot results.