Testing SDL3 kmsdrm backend on our embedded systems I saw failures on a imx6 based hardware with mainline kernel and mesa3d/etnaviv driver. I didn't show anything on the display and later on produced errors from mesa:
MESA: error: etna_cmd_stream_flush:244: submit failed: -24 (Too many open files)
The same application works perfectly on imx8 with the manufacturer kernel/vivante driver.
I then let AI (claude) do its job and it came up with the two attached patches against release-3.4.14. Maybe this is also a solution for #12418
I read and understand the statement about AI usage in SDL but would like to leave the patches here anyway. At least to have something to discuss.
sdl3-kmsdrm-scanout-format.patch
sdl3-kmsdrm-atomic-request.patch
Testing SDL3 kmsdrm backend on our embedded systems I saw failures on a imx6 based hardware with mainline kernel and mesa3d/etnaviv driver. I didn't show anything on the display and later on produced errors from mesa:
MESA: error: etna_cmd_stream_flush:244: submit failed: -24 (Too many open files)The same application works perfectly on imx8 with the manufacturer kernel/vivante driver.
I then let AI (claude) do its job and it came up with the two attached patches against release-3.4.14. Maybe this is also a solution for #12418
I read and understand the statement about AI usage in SDL but would like to leave the patches here anyway. At least to have something to discuss.
sdl3-kmsdrm-scanout-format.patch
sdl3-kmsdrm-atomic-request.patch