00:40 <jenatali> alyssa: If you were curious I found the Vulkan CTS cases for the missing component border handling. I tripped over them because of a bug in WARP for integer texture sampling :)

00:41 <jenatali> Doesn't help for GL though :P

02:59 <alyssa> jenatali: Delight

09:50 <Siddh[m]> Hi, can anyone review https://lore.kernel.org/dri-devel/cover.1672957022.git.code@siddh.me/ ? Thanks. (For 1st patch, please consider the reply)

10:09 <sravn> Hi Siddh[m] - I like that we slowly move away from the deprecated logging functions.

10:10 <sravn> The first path - please cosider to split inline struct device *___drm_get_dev_ptr(const void *ptr, bool is_drm) in two funtions, so we avoid the is_drm flag.

10:11 <Siddh[m]> sravn: Oh okayy, noted. I will do that. Thanks!

10:11 <sravn> The confidence level for the following patches had been higher, if you hade used cocinelle to do the job you did manually here. Have you looked at that possibility?

10:13 <sravn> I looked brifly - and found no mistakes. But cocinelle just do a better job for this kind of changes

10:13 <Siddh[m]> sravn: No, and I really did forget cocinelle existed. Thanks for pointing out

10:13 <Siddh[m]> Though I have never really used coccinelle outside of static analysis

10:14 <sravn> And then it is easier to repaet the same edits in drivers etc. See other changelogs how they include the cocinelle script in the changelog

10:14 <Siddh[m]> Sure, I will take a look. Thank you!

11:55 <Nyaa> hi

11:56 <Nyaa> didn't know you were here too

12:11 <Nyaa> hooray, I think I figured out the vdpau segfaults for r300

12:41 <DavidHeidelberg[m]> Nyaa: nice, now can you do rewrite to VAAPI? :P

14:47 <Nyaa> and with https://gitlab.freedesktop.org/mesa/mesa/-/issues/8031 i think that might be enough code archaeology for me today

14:52 <Nyaa> DavidHeidelberg[m], if I attempt that I might just try to get shader-based h264 decoding working too

17:35 <Siddh[m]> <sravn> "The first path - please cosider..." <- I split it and also removed simplified the Generic. Please check https://lore.kernel.org/dri-devel/625541825d76ae5ef0d82f4a0ce5c8976ceff9f0.1673110890.git.code@siddh.me/. Thanks.

17:59 <javierm> sravn: I see that you didn't provide your r-b for https://lore.kernel.org/lkml/20221228014757.3170486-3-javierm@redhat.com/, did you forget?

18:28 <sravn> javierm: I forgot. It is r-b, I have deleted the mail, but can resurrect it via lore if you likethe r-b om the mailing list

18:32 <javierm> sravn: no need to sent to the ML, thanks!

18:35 <sravn> Siddh[m]: Do you need the extra functions, or could you embed it in __get_dev_ptr? I know I asked to split it up, but looking now it looks like this is too much.

18:41 <Siddh[m]> <sravn> "Siddh: Do you need the extra..." <- Null check has to be in a function, we cannot check for null in a macro. Generics only evaluate the thing if type matches, but the syntax regardless have to be correct, so I cannot put a drm->dev directly under first case.

18:41 <Siddh[m]> <"but looking now it looks like this is too much"> yeah, that's why I originally used the bool thingy to do the work one function. It was a compile time thing anyways, but it did feel like a hack when you pointed it out.

18:42 <javierm> sravn: I was meaning to look at b4 too. I wonder what could give me that patman doesn't

18:43 <Siddh[m]> in one function*

21:01 <sravn> javierm: git send-eamil did not work well with my current mail provider (one.com) and I was not happy using my google account for this. With b4 I can send patches without the smtp server hassle and I heard people say positive things about by. So far I have just used git + some basic scripting and never bothered with patman/quilt etc.

21:02 <sravn> s/by/b4

21:46 <javierm> sven: interesting, I should take a look

21:46 <javierm> sorry, meant sravn ^