Works fine. Tested with GLMark2 and MPV (Vulkan HW accelerated video with mesa-vulkan-drivers-freeworld package).

This fails the new vkcube --validate test, but it's not segfaulting. It's just exiting non-zero (I guess). A warning vkGetDeviceProcAddr(): pName is trying to grab vkCreateDisplayPlaneSurfaceKHR which is an instance level function is shown on the console - https://openqa.fedoraproject.org/tests/4019348#step/desktop_graphics_validation/13 . The earlier warning about a driver being skipped also shows up when the test passes, so we can ignore that one.

This is actually the first time the test has run on F42, so it might be the same even without this update, not sure.

I'll poke into this a bit more (at least find out whether it fails without the mesa update).

FWIW, I'm getting (on an AMD Radeon RX6700):

$ vkcube --validate
Cannot find layer: VK_LAYER_KHRONOS_validation
vkEnumerateInstanceLayerProperties failed to find required validation layer.

The exit code is 1.

It looks like you are missing vulkan-validation-layers, installing this package should fix it:

$ sudo dnf install vulkan-validation-layers

Thanks, installing that package solves the issue:

$ vkcube --validate
Selected WSI platform: xcb
Selected GPU 0: AMD Radeon RX 6700 (RADV NAVI22), type: DiscreteGpu
WARNING : VALIDATION - Message Id Number: 582089644 | Message Id Name: WARNING-vkGetDeviceProcAddr-device
    vkGetDeviceProcAddr(): pName is trying to grab vkCreateDisplayPlaneSurfaceKHR which is an instance level function

Bug 2417967 submitted.

@rathann what is the exit code of that command when you run it? I suspect it's not 0.

Looks like the openQA test is also failing on an unrelated update so I guess it's a pre-existing issue not caused by this update. I'll have to try and figure out if we can make the check workable on f42 or we'll just have to skip it there.

@adamwill

what is the exit code of that command when you run it? I suspect it's not 0.

It's still 1.

Right, that matches what I'm seeing.

I've filed https://bugzilla.redhat.com/show_bug.cgi?id=2418077 and worked around it in openQA for now. It doesn't need to hold up this update - it exists outside of it.

Works

I'm split about this update.

On one side about 3rd part project coordinator, I'm willing to apply any update that are provided by the mesa-freeworld maintainer.

On the other end, I'm experiencing a regression with mutter triggered by this mesa update. Where the workaround was to revert the to mesa version in f42... https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4786#note_2619170

So at the end I'm not going to block on this, but I still consider a case of a major version upgrade in fedora stable as "spreading bugs everywhere".

I've backup the previous mesa for regression at https://dl.kwizart.net/pub/fedora-42-mesa-25-1/ (signed with fedora-42 gpg key). so I can do the paper work for the regression I have with Tegra.

@jexposit we can update the package as needed (I had an early pushed today, but reverted it).

Hey @kwizart

On the other end, I'm experiencing a regression with mutter triggered by this mesa update. Where the workaround was to revert the to mesa version in f42... https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4786#note_2619170

Sorry, I didn't understand if the regression is caused by mesa or by mutter.

Anyway, there is no need to rush this update (or to push it) if there are regressions. If you found a case were this mesa update cause an issue, I prefer to drop this update and leave the current mesa as it is in Fedora 42.

If you confirm that mesa caused problems in your set up, I'll revoke this update to avoid issues.

no regressions here, but 0 due other user's comment

This update has been unpushed.

I unpushed this build due to @kwizart's comment. I don't want to risk a regression in Fedora 42, let stay in the safe side.

So... Was there any progress? It would be great to update Mesa in F42 to 52.2.x (and if possible also Mesa in F43 to 25.3.x). Both already have very recent kernel 6.18, but still using an outdated Mesa version without any of the recent improvements and support for newer hardware.

*25.2.x

Sorry for the silly typo.

You can have latest mesa if you like: https://copr.fedorainfracloud.org/coprs/che/mesa/

Thanks, but that's not a proper solution for end-users.

@asciiwolf this update caused a regression as commented by kwizart, so I unpushed it. I'd prefer to keep a stable version in Fedora 42, so there are no plans to push this at the moment.

True and it is the right approach. I just hope this wasn't the last 25.2.x build for F42. :)