imx6: update kodi patches, second pass

[vpeter4]

No description provided.

[mk01]

@vpeter4
patch to rendererimx breaks all deint methods but BOB.
(this is proper refactor - mk01/xbmc@e1cac5f)

[vpeter4]

Actually I already have applied all patches from @warped-rudi like https://github.com/OpenBricks/openbricks/blob/0e0d33112608f421f5bd8e60849a881f6a063d1e/packages/multimedia/kodi/patches/0042-IMXRender-Refactor-to-improve-readability-and-perfor.patch. If I see correctly.

Why doesn't such kode come to kodi upstream? Seems to me just a big mess all over.

Update: Now I see that it is not the same code. So would would be the perfect one?

[mk01]

@vpeter4

I'm commenting indeed the fact that you applied and broke deinterlacing.

[vpeter4]

Obviously I don't understand anything of this but from what I see it is broken in kodi krypton upstream anyway. So I didn't broke anything - the code was/is already broken.

[fritsch]

The big question is: Why does a distribution like LibreELEC that wants to ship official kodi pick multiple external patches from elsewhere != upstream, without even knowing if they fix something or if the do something good?

@mk01 is the official imx maintainer of kodi and it's sad that this single dev now needs to go towards downstream and fix the pickery issues.

[mk01]

@vpeter4

• first two screens are from advanced(half)/advanced resp. - as they should look correctly (upstream or the simplified code in commit in 1st comment)

• then two screenshots following are advanced(half) and adv(half) zoomed. the last one is advanced - patched.

(video is burosch1.mpg (720x576x25) played at 1920x1080x50 from @fritsch 's pool of test files)

upstream - advanced(half) ((yadif 1x))

upstream - advanced ((yadif 2x))

patched - advanced(half)

patched - advanced(half)

patched - advanced

[fritsch]

@mk01 you don't really use yadif right? Just reusing the name :-) (Yeah: https://github.com/xbmc/xbmc/blob/master/xbmc/cores/VideoPlayer/DVDCodecs/Video/DVDVideoCodecIMX.cpp#L638)

[fritsch]

@mk01 while having you here. I asked @vpeter4 for the reason to pick the stuff from openbricks. It seems there are some bugs left, he mentioned "seeking" would cause issues. Is that okay to file these bugs on trac.kodi.tv or where do you want them?

I want to make sure that no distro has to pick valid fixes from elsewhere, but making sure upstream works as it should - features are something else, obviously.

@vpeter4 could you name the issues you have with "standard v17"? - then we can track them and fix them.

[vpeter4]

Fro a start - SIGSEGV on fast forward video. This fixes it: https://github.com/OpenBricks/openbricks/blob/0e0d33112608f421f5bd8e60849a881f6a063d1e/packages/multimedia/kodi/patches/0040-IMXCODEC-Prevent-lkFIFO-from-returning-uninitialized.patch

[warped-rudi]

@fritsch: Upstream had issues and did not even compile cleanly from august to december...

Besides the problem @vpeter4 mentioned, I see here hangs on exit (usually the video plane gets switched on again so that the last frame of the previously played video is shown while Kodi waits infinitely for CIMXContext::Process() to exit).

Most of the other patches are attempts to make the code better understandable for someone who has not written it, remove unneeded stuff and tweak performance. They are not really new as I sent a link to the repo that contains them to @mk01 by the end of september. But I did not get feedback on that. I quickly adapted them to the december commit and probably broke something. S**t happens. The current source can be found here https://github.com/warped-rudi/xbmc/commits/Krypton-iMX6 . Maybe someone can comment.

[fritsch]

@warped-rudi:

Upstream had issues and did not even compile cleanly from august to december...

Could you perhaps think of the fact, that upstream - with only one imx dev left - would have been happy to get that compile fix from you?

[warped-rudi]

@fritsch:
Well, I thought it would be sufficient to report the issue to the maintainer. BTW, where can I find your pool of test files?

[fritsch]

There are some files that I used back at the time I was writing IMX code: http://solidrun.maltegrosse.de/~fritsch/ - mostly for deinterlacing and throughput testing 2017-01-08 17:28 GMT+01:00 Rudi Ihle <[email protected]>:

…

@fritsch <https://github.com/fritsch>: Well, I thought it would be sufficient to report the issue to the maintainer. BTW, where can I find your pool of test files? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1119 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABCfHZU7YzlTWVefUu8Z4WTVJz00fiuGks5rQQ6YgaJpZM4LZC7G> .

-- Key-ID: 0x1A995A9B keyserver: pgp.mit.edu ============================================================== Fingerprint: 4606 DA19 EC2E 9A0B 0157 C81B DA07 CF63 1A99 5A9B