ChanServ changed the topic of #zink to: official development channel for the mesa3d zink driver || https://docs.mesa3d.org/drivers/zink.html
Akari has quit [Read error: Connection reset by peer]
Akari has joined #zink
fahien has joined #zink
fahien has quit [Quit: fahien]
<zmike> kusma: nice work finding it!
<kusma> np :)
<zmike> I think there are a lot of issues with frontends ignoring caps
<kusma> Well, there hasn't generally been a rule that states that aren't supported not be set, just ignored
<kusma> It's *nice* not to set them, because the CSO cache will love you... But it's pretty common that these kinda contain "junk"... or at least that used to be the case.
<zmike> I'm talking more generally
<zmike> not specific to this one thing
<kusma> Yeah, sure.
<kusma> It's very common that people add a cap and only respect it in the GL state-tracker ;)
<kusma> I've been guilty of quite a few cases of that myself, I'm sure
<zmike> not a real mesa developer if you haven't done that
<kusma> Oh boy. NV30 seems to be the only GPU in Mesa ATM that doesn't support PIPE_CAP_NPOT_TEXTURES...
<zmike> I wish there was an easier way to see what caps were supported with which value
<zmike> would be cool if ci could figure that out and have a site to track it
<kusma> Yes.
<kusma> Yeah, I kinda want our CI runs to start tracking stuff like extensions etc supported so we can retire features.txt
<kusma> These things would fit right into that.
<kusma> It just always gets nasty as soon as I start thinking about the details
<zmike> yep
<kusma> I don't like to think about nasty stuff.
<kusma> So instead I think of nice stuff. Like moving to a farm and giving up this "programming" stuff.
<zmike> I know someone who did one of those things
<zmike> now he's smoke proofing his house because of all the wildfires
<zmike> ajax: any luck?
<ajax> some, kinda. which issue do we mean here?
<zmike> any? :D
<ajax> heh
<ajax> i think i have a strategy for avoiding xcb_wait_for_... at least for the find-back-buffer part of dri3
<zmike> nice
<ajax> !17685 seems to cure glx-visuals-stencil
<ajax> !17634 is still a nice set of ideas but michel's being a stickler for that whole correctness thing
<ajax> the nerve!
<zmike> gasp
<ajax> i wonder just how many piglits that fixes
<zmike> hard to say until I can see the flakes coming through with it merged I think
<ajax> why can i not get lavazink working anymore
<zmike> LIBGL_ALWAYS_SOFTWARE?
<ajax> my wrapper script looks like https://paste.centos.org/view/raw/110af34b
<zmike> elaborate
<ajax> 'zink lavapipe vulkaninfo' works but z l glxinfo does not
<ajax> (have to run away for a second, just complaining out loud here)
<zmike> obviously dumb question but you're using the right libGLX right?
<ajax> datura:~/git/zink% zink lavapipe ldd `which glxinfo` | grep GL libGL.so.1 => /var/home/ajax/git/zink/build/src/glx/libGL.so.1 (0x00007fcc89b2f000)
<ajax> bbiab
<zmike> seems like it should work
<zmike> maybe try VK_LOADER_DEBUG=all with glxinfo to make sure it's opening the driver
<ajax> DEBUG: Instance Extension: VK_EXT_debug_utils (/var/home/ajax/git/zink/build/src/gallium/targets/lavapipe/libvulkan_lvp.so) version 0.0.2
<ajax> libGL error: glx: failed to create drisw screen
<ajax> libGL error: failed to load driver: zink
<ajax> oh good, that means this is my fault.
<zmike> could be zink returning early from screen creation?
<ajax> pilot error, forgot export before LIBGL_ALWAYS_SOFTWARE=1
<zmike> classic
<ajax> do we only have ci jobs for anv and lvp at the moment?
<zmike> yep
<zmike> though there's a turnip trace job that's sort of trying to get in there once glx/gbm issues are fixed
Soroush has quit []
Soroush has joined #zink
eukara has quit []
eukara has joined #zink