09:59 <warpme_> hi all !

09:59 <warpme_> just quick Q: is there any progress with t720?

11:39 <raster> bbrezillon: btw - did you resolve the partial update/buffer/clearing thing because ... latest mesa git master is horribly flickery and broken for me in enlightenment

11:41 <raster> bbrezillon: it still seems to be that it ends up clearing non-updated areas with 0's etc.

11:41 <raster> and it leads to all sorts of nasties down the pipeline

11:41 <bbrezillon> raster: nope

11:42 <raster> ok. so still pending :)

11:42 <raster> sorry - been distracted by other things of late too

11:43 <raster> i just reset my mesa tree and did an update to latest master and enjoyed my flicker-heaven :)

11:43 <bbrezillon> well, I started looking at how I could get things working without passing the context, but couldn't find a clean solution

11:44 <raster> perhaps for now it's best to revert the original changes in master until this works right?

11:47 <bbrezillon> raster: that's an option

11:48 <raster> i know it's not wonderful... but better than a broken experience in a mesa release... :)

11:48 <bbrezillon> daniels, alyssa: what's your opinion ^

11:49 <raster> i know the patch is good for performance reasons... :)

11:49 <raster> but bad for correctness (at least until fixed)

11:49 <raster> just raising the question for now anyway

12:52 <alyssa> bbrezillon: Reverting probably isn't terrible (and then cc mesa-stable I guess)

12:53 <bbrezillon> alyssa: I guess I should revert everything then

12:53 <alyssa> bbrezillon: Er actually

12:53 <alyssa> can't you just patch out the few lines in panfrost that enable damage tracking?

12:53 <alyssa> So then we don't expose the ext but no need to invasively mess with reverts?

12:53 <bbrezillon> it's actually the whole thing that's broken

12:54 <bbrezillon> other drivers implementing the ->set_damage_region() will be impacted too

12:55 <alyssa> Oh. Don't break other people's drivers ;)

12:59 <bbrezillon> alyssa: it's not the case, yet

12:59 <bbrezillon> :p

13:00 <bbrezillon> drivers have to implement ->set_damage_region() to expose the feature

13:48 <bbrezillon> and the only driver implementing this method is panfrost

14:45 <warpme_> bbrezillon: quick Q: I have again white screen on t820 (you remember - we were bisecting this weeks ago). is this regression on wip?

14:48 <warpme_> on->or

16:32 <bbrezillon> warpme_: I reverted the patch you said was responsible for the regression

16:32 <bbrezillon> so it's probably something else

16:36 <warpme_> right :-( current master is huge progress regarding lima but a bit regression regarding t820....

16:42 <robmur01> just tried mine, and indeed the cube in glmark -btexture now comes up blank (but still shaded)

16:43 <robmur01> (also, I finally solved my glmark problem :D)

16:44 <robmur01> amusingly, the totally-unsupported T620 is still texturing fine...

16:51 <HdkR> silly

16:57 <raster> bbrezillon: patch -> mesa -> thumbs up. do you want me to create some test stuff for you so you can easily see the issue?

17:05 <robmur01> ooh, T820 actually seems to have DATA_INVALID_FAULTs correlated with missing textures

17:12 <robmur01> ...or not - actually it seems to spew a couple of those at the start of whatever run it's doing. Must just be an artefact of the FPGA being hella slow :(

17:30 <bbrezillon> raster: if you can make sure the patch fixes the issue that'd be great! (I don't have a board at hand)

19:09 <bbrezillon> alyssa: can you have a quick look at https://gitlab.freedesktop.org/bbrezillon/mesa/commit/16e2d060f1c6b901dfe9fb8477c30df05c387216 and https://gitlab.freedesktop.org/bbrezillon/mesa/commit/710b1384ab522742ff21d7e5c56dc2f98a1ffad0 and let me know if you're happy with this version

20:28 <alyssa> bbrezillon: I'm in prepare-for-XDC-crunch mode so 'fraid I can't look until Thursday evening :(

20:30 <bbrezillon> alyssa: np