GUI startup issue in v0.9.0 #142
Comments
|
Hey Jo, I see this too, Additionally I noticed that we no longer get the binary Sorry I skipped packaging 0.8 because I was too busy :( |
|
oh man, wayland makes me sad. not clickable might be related to #137 but in the log above (this is with wayland startup, not with x11 glfw, right?) it is unable to create the swapchain: which is what i regularly get from nvidia on wayland. @korrat this is a laptop right? with dedicated AMD GPU that is confused about access to the display.. @paperdigits : glfw now ships both for runtime switching. as korrat mentioned, you can steer it via environment variable, so there is no need for an explicit binary for wayland any more. @korrat : which driver is this? amdgpu, amdgpu-pro, or radv? also, is this a debug build ( will try to install wayland/something once again.. |
Right. The x11 startup does not produce that error.
No, this is a Desktop (with an Intel CPU & AMD GPU).
That should be
No. Building with |
|
okay, great, thanks for confirming a few details there. if you could install the validation layers and re-run the debug build maybe it will tell us some more? btw just pushed 0ee0cb3 which may or may not fix the mouse interaction for wayland. hard to tell for me since i'm currently on an nvidia machine and running wayland is.. interesting. |
|
Sorry, took some fighting with the validation layers, but here's the log: |
|
excellent, thank you. i am going to say this is a driver/vulkan bug, much like KhronosGroup/Vulkan-Docs#2440 but let me force fix this on our side too. are you using a rotated screen at all? |
|
@korrat if you have the chance to compile from git, can you confirm that the above mentioned commit fixes the issue? |
Yeah, HDMI-A-6 is rotated to portrait orientation. Is there any way to force vkdt to use the displayport screens and ignore the HDMI one? |
Still no dice, sadly. But the error is somewhere else, I think. |
|
now i'm a bit at a loss. the code literally does exactly what the validation layer says, i.e. queries or uninstall amdvlk or some such. |
|
New tests: forcing radv through
|
|
awesome. so i'll consider this fixed then. i don't trust any more errors coming out of amdvlk :) such semaphore problems are pretty severe and i did debug this with validation layers that seemed to actually know what they are saying. and this works on windows, macos/intel, debian/(nvidia,intel,amd), so i doubt there is an actual deadlock here. thanks for providing these detailed validation layer outputs though, they are really helpful. to answer your other question about monitors: this is just for colour management. for wayland it's probably broken anyways. the only thing that could be a workable approach is to determine one main monitor you care about and put the icc profile into all the profiles vkdt loads. the detection code "where is my window" does not work on wayland. i will currently ignore VCGT/correction curves, so it's essential that during profile creation the white point of the monitor matches quite well (through hardware adjustments). |
I am unable to open the GUI under Hyprland/Wayland on NixOS.
The log ends with
double free or corruption (!prev)without ever showing a window.Additionally, I see two initialization errors in the log.
Forcing GLFW to use X11 using
XDG_SESSION_TYPE=x11gets to an initial window, but no interaction is possible. It also seems that after the first interaction attempt, the log freezes as well.Execution log:
The text was updated successfully, but these errors were encountered: