08:11pq: ascent12, make sure you release the *previous* buffer on pageflip event, not the one you just programmed in. Otherwise, weston should be seeing issues too, and I don't think it is.
08:12pq: ascent12, at least intel seems to tend to send the pageflip event a little before the timestamp, which is why weston has to handle "negative event delta time".
15:11jcdutton: Hi. Does anyone know what this means: amdgpu 0000:43:00.0: No GEM object associated to handle 0x0000000D, can't create framebuffer
15:12jcdutton: It is failed to start X.
15:21MrCooper: jcdutton: which version of Mesa?
15:39jekstrand: anholt: If someone puts their CI farm behind a squid caching proxy, will it actually cache the build artifact downloads or does HTTPS and authentication blow that up?
15:40jcdutton: MrCooper, Which command tells me the version? I am using ubuntu bionic
15:43kisak: if it's distro installed: apt policy <any mesa package like libgl1-mesa-dri>
16:02jcdutton: kisak, thank you. It had some old version, 18.0.0. I did an update, so it now has 19.2.8, and X is working.
16:17MrCooper: right, that was https://bugs.freedesktop.org/show_bug.cgi?id=110903 then
16:45Venemo: panfrost CI is flaky again: https://gitlab.freedesktop.org/Venemo/mesa/pipelines/114171
20:19robclark: so, anyone have any hint about what step I'm missing to get "Mesa: Texture Obj 0 incomplete because: Image[baseLevel=0] == NULL" when trying to sample from an FBO that I rendered to? https://hastebin.com/xivasexeyo.cpp
20:20robclark: I can see from cmdstream that it's actually doing the part that renders to the fbo, so fbotex should be good (and have a backing buffer)
20:26imirkin: robclark: min filter
20:26imirkin: or something dumb like that
20:26imirkin: er no, you did that
20:27robclark: yeah, I'm pretty sure it is something dumb ;-)
20:27imirkin: is there a render somewhere in between?
20:28imirkin: otherwise there's no bakcing for the texture
20:28imirkin: anyways... you could spend time figuring out what's causing it
20:28imirkin: or you could fix it
20:28imirkin: the fix is to use glTexStorage instead of glTexImage
20:29imirkin: there's _obviously_ a way to make it work without that, but do you care
20:30robclark: hmm, still getting the same thing..
20:30robclark: I'm *assuming* I can setup an fbo once, and then use it each frame..
20:30imirkin: can you share the complete thing?
20:30robclark: fwiw, more complete thing.. https://hastebin.com/osedawebol.cpp
20:31imirkin: will take a look in a bit
20:35imirkin: robclark: i assume draw_shadertoy is where it fails?
20:36imirkin: right. so what you're doing it illegal
20:36robclark: well, if I use glTexStorage2D() it does, because incomplete framebuffer
20:36imirkin: but irrespective of that
20:36imirkin: the issue is that when you do the first draw, there is no backing texture allocated
20:37imirkin: with glTexStorage2D, that shouldn't be an issue
20:37imirkin: but you can't sample and render to the same texture
20:37imirkin: without glTextureBarrier() in the middle
20:37imirkin: and even then, you can only sample from the texel corresponding to gl_FragCoord
20:37robclark: hmm, switching fbo doesn't imply that, I guess?
20:38robclark: so first pass doesn't actually use any samplers..
20:38robclark: (at least not intentionally)
20:38robclark: it just renders to fbo.. then 2nd pass tries to sample from it
20:38imirkin: first pss = draw_shadertoy
20:38imirkin: second = cube?
20:39imirkin: so draw_shadertoy is only called once?
20:39robclark: once per frame
20:39imirkin: right, wtvr
20:40imirkin: and what's the problem?
20:40imirkin: the draw in the cube is getting an incomplete texture?
20:41imirkin: but it's trying to render to the same fbo? or am i missing something?
20:41robclark: shouldn't glBindFramebuffer(GL_FRAMEBUFFER, 0); be switching rendering back to the framebuffer?
20:41robclark: instead of the fbo
20:42imirkin: oh. end of draw_shadertoy. yes.
20:48imirkin: robclark: uTex is a samplerExternalOES
20:48imirkin: i think you want sampler2D?
20:48robclark: ahh, yes
20:48robclark: copypasta fail
20:50robclark: yeah, that seemed to be the issue..