<zmike>
I think there's still a missing piece in kopper somewhere to make this trigger shutting down the driver or something
<zmike>
because on nv the result is if you click the close button on glxgears the window closes gracefully but then the process continues rendering
<ajax>
yeah, there is
<zmike>
at the least, with that MR it should be possible to use the return value of zink_kopper_update() for whatever needs to happen
<zmike>
I think
<ajax>
a piece missing, i mean. i was reading through this code while sick last week and i think i didn't hallucinate this, but
* zmike
braces
<ajax>
afaict the notification chain for DestroyWindow stops in the dri frontend code in the driver. it calls into st but nothing calls into the driver.
<zmike>
maybe?
<zmike>
it seems like st is still chugging away just fine
<ajax>
so if, like waffle, you do display teardown with DestroyWindow before DestroyContext, DestroyContext would normally think it has to do all the dereferencing cleanup
<ajax>
but we're likely to have gotten notified from the wsi thread already
<zmike>
hm
<ajax>
when we could just explicitly shoot down the swapchain from DestroyWindow
<zmike>
annoying
<zmike>
sounds plausible
<ajax>
i'm writing up some other stuff for a fedora change atm, i'll look into it after lunch once i've rebased
<zmike>
cool
LexSfX has quit [Ping timeout: 480 seconds]
LexSfX has joined #zink
lygstate has joined #zink
lygstate has quit [Write error: connection closed]