| *** n2klb_ is now known as n2klb | 01:58 | |
| henk | poetaster: well, that’s nice. I just don’t have the spare money to comfortably afford buying another phone … that would be the third in five years and that’s just not really in the budget unless it absolutely needs to be. | 06:09 |
|---|---|---|
| poetaster | henk, yes, I know what you mean. I can't afford it now either and depend on the community. | 08:42 |
| poetaster | henk, I purchased the GS5s refurb years ago with extra batteries when I did have some cash. | 08:43 |
| henk | JhImsPWi1IRCp! | 08:52 |
| henk | geh, wrong window focused /-: | 08:53 |
| ender | haha | 08:59 |
| Mister_Magister | abr: hiiii in this wonderful monday didy you have time to look at the PR :) | 13:02 |
| abr | still not got time sorry. did you find anyone else to look at them? | 13:03 |
| Mister_Magister | abr: was I supposed to? Is there literally anyone else to look at them beside you? I made PR to repo i expect repo owners to figure that part out xd | 13:24 |
| abr | wow | 13:26 |
| Mister_Magister | abr: hm? | 13:47 |
| Mister_Magister | that's normal when making a pr? no? | 13:48 |
| Mister_Magister | Like if I made a PR to ffmpeg i wouldn't be searching for person to make me a CR and accept it, i assume they will see it and figure that part out on their own | 13:48 |
| Mister_Magister | I don't work at jolla | 13:48 |
| Mister_Magister | You don't seriously expect me to stalk every jolla employee, figure out who is responsible for this repo and email them asking for a CR? | 13:50 |
| poetaster | mal, I'm working on new backends for the image and video editors I took over. what would you choose: libav or gstreamer | 14:39 |
| Mister_Magister | poetaster: gstreamer is sfos way because you can tap into gstdroid | 14:46 |
| Mister_Magister | so hw decoding/encoding | 14:46 |
| Mister_Magister | ffmpeg does not give you that | 14:46 |
| poetaster | Mister_Magister, ah, that is a good arg.... | 14:47 |
| Mister_Magister | poetaster: and when abr reviews my PR and it gets merged you'll have VERY simple way of making a sink for it :)))) | 14:50 |
| Mister_Magister | thats why this pr is so importante | 14:50 |
| poetaster | Mister_Magister, PR in which repo? | 14:51 |
| Mister_Magister | uhhh which one was it again… | 14:55 |
| Mister_Magister | https://github.com/sailfishos/gst-plugins-good/pull/15 | 14:57 |
| Mister_Magister | its part of group of PRs collectively aimed at enabling qmlglsink | 14:57 |
| Mister_Magister | so gst sink in qml for free | 14:58 |
| Mister_Magister | no two self-coded sinks juggling | 14:58 |
| Mister_Magister | will help not only me but everyone else who wanna use gstreamer | 15:00 |
| poetaster | looks reasonable. I could take a crack at building them | 15:01 |
| Mister_Magister | poetaster: its all built if you wanna test it but you would need compatible port | 15:04 |
| Mister_Magister | it's all already done, just two prs need to get reviewed and merged | 15:04 |
| Mister_Magister | enabling it was small potatoes, making it work was the big issue | 15:05 |
| poetaster | compatible port? | 15:08 |
| Mister_Magister | yeah you need latest droidmedia and new enough gst-droid to contain the fixes | 15:10 |
| poetaster | xperia 10iii with 5.1? or even newer? | 15:13 |
| Mister_Magister | dunno but you can check the version | 15:14 |
| abr | Mister_Magister: You misunderstand. I (seriously) expect you to understand that we are all very busy and a deeply complex PR like this isn't something you can expect people to drop everything for hours to assess for you. | 15:18 |
| poetaster | 0.20260522.0-1.11.1.jolla | 15:19 |
| abr | I've already spent some time looking at it back when you posted it, and said I don't think it's the right fix in the right place. | 15:19 |
| Mister_Magister | abr: that's why i'm patiently waiting and just asking from time to time :) | 15:19 |
| Mister_Magister | abr: and i answered I think it is right fix when i understood the issue at hand hence i'm asking to reconsider :P | 15:19 |
| abr | and that means I will have to take even more time to find out the right place to put it. | 15:20 |
| Mister_Magister | and i'm patiently waiting just asking from time to time :) | 15:20 |
| Mister_Magister | if i don't keep asking youll forget | 15:20 |
| Mister_Magister | like my 2 year old fix that nobody merged | 15:21 |
| abr | I don't think it's the right fix, and I don't know what the right fix would be without spending a long time investigating | 15:21 |
| abr | and I might not even be the right person to ask | 15:21 |
| Mister_Magister | abr: well i'll be happy if you pointed me at another person to bother weekly :D | 15:22 |
| abr | It looks to me like it reaches up through the entire pipeline to rebuild the GL state, complaining that the scene graph isn't being invalidated, when the onSceneGraphInvalidated method is a dummy | 15:22 |
| Mister_Magister | abr: as i mentioned earlier implementing onSceneGraphInvalidated does nothing | 15:23 |
| Mister_Magister | it never gets called | 15:23 |
| abr | maybe try and make it more digestible so someone might be able to spare enough time to look at it properly | 15:23 |
| Mister_Magister | and fixing calling it is not right fix either | 15:23 |
| Mister_Magister | I tried | 15:23 |
| Mister_Magister | i tried implementing it i tried fixing calling it | 15:23 |
| Mister_Magister | they were both even worse ideas | 15:24 |
| abr | then I can't help you sorry | 15:24 |
| Mister_Magister | I cannot make it more digestable because it is complex issue | 15:24 |
| Mister_Magister | if not you then who? | 15:24 |
| Mister_Magister | like, it gotta get fixed somehwo | 15:24 |
| poetaster | abr, hm, should be a pipeline shortcut (maybe) | 15:25 |
| Mister_Magister | it's very important fix | 15:25 |
| abr | did you make a minimal reproduction? | 15:25 |
| abr | maybe step back and try to demonstrate the failure and intended recovery? find out why Qt goes wrong before jumping into gst with the answer when the question isn't very clear, to me anyway. | 15:27 |
| abr | is it definitely even a SFOS issue, or something upstream? | 15:29 |
| Mister_Magister | abr: there's no need for minimal reproduction its incredibly easy to reproduce | 15:36 |
| Mister_Magister | abr: "maybe step back and try to demonstrate the failure and intended recovery" already did | 15:36 |
| Mister_Magister | "find out why Qt goes wrong before jumping into gst with the answer when the question isn't very clear, to me anyway" already did | 15:36 |
| Mister_Magister | "is it definitely even a SFOS issue, or something upstream" definitely sfos | 15:37 |
| abr | none of that is written in the PR | 15:37 |
| Mister_Magister | yes it is | 15:37 |
| Mister_Magister | " Qt can destroy and recreate that context without tearing down the scene graph - on SailfishOS this happens whenever the display blanks" | 15:37 |
| abr | maybe simplify that a bit then, because all I see is a wall of LLM | 15:38 |
| Mister_Magister | and only sfos meshes such old qt with recent gstreamer | 15:38 |
| Mister_Magister | will do | 15:38 |
| abr | because otherwise I'm too stupid to review your PRs | 15:38 |
| Mister_Magister | it's incredibly simple: you turn screen off, turn it back on, the video is black | 15:39 |
| Mister_Magister | every single time | 15:39 |
| Mister_Magister | i made spaghetti meanwhile | 15:39 |
| abr | how do you know it's only sfos? | 15:40 |
| Mister_Magister | because no such error was reported upstream and only sfos meshes such old qt with recent gstreamer and in qt5.12/6 it was rewritten so the issue does not affect it | 15:40 |
| Mister_Magister | and | 15:41 |
| Mister_Magister | normal computers don't have screen blanking | 15:41 |
| abr | that doesn't mean it's a sfos specific problem | 15:42 |
| Mister_Magister | strongly suggests? | 15:42 |
| abr | if any device with old qt, recent gst and screen blanking would reproduce it | 15:42 |
| Mister_Magister | is there such a device? | 15:42 |
| Mister_Magister | beside sfos? | 15:42 |
| abr | doesn't matter. that decides whether to look towards droidmedia etc or old qt | 15:43 |
| Mister_Magister | and nobody upstream is gonna be fixing qt5.6 | 15:43 |
| abr | but you have to decide where to fix it now | 15:43 |
| Mister_Magister | I did | 15:43 |
| Mister_Magister | the fix belongs where the PR is | 15:43 |
| abr | that's what you need to prove | 15:44 |
| Mister_Magister | with what | 15:44 |
| abr | instead of just proclaiming it | 15:44 |
| Mister_Magister | how do I prove it then | 15:44 |
| abr | how should I know? | 15:44 |
| Mister_Magister | and how should I know? | 15:45 |
| abr | because you wrote the PR | 15:45 |
| Mister_Magister | Never before when making PR i was required to prove that it belongs there | 15:45 |
| Mister_Magister | and how "because i wrote PR" tells me what proof do you require of me? | 15:45 |
| abr | because you're the only one who investigated the problem and know all the details? | 15:46 |
| Mister_Magister | Look mate I see it this way. I spend my time (a lot of it) and my money to improve Sailfishos. Is it crazy for Jolla to review and merge it? | 15:46 |
| Mister_Magister | abr: that still doesn't tell me what kind of proof you require of me | 15:46 |
| Mister_Magister | the proof is that it fixes the issue | 15:46 |
| abr | some problems have lots of solutions | 15:46 |
| Mister_Magister | and this is my solution | 15:47 |
| Mister_Magister | and i tired other solutions and they weren't solutions | 15:47 |
| abr | you're asking me to approve and merge something, and so from then on take responsibility for maintaining it, | 15:47 |
| Mister_Magister | no, I'm asking jolla to accept my effort to improve Sailfishos | 15:47 |
| abr | it's crazy to just think you can insist it's the right solution, but I really think it isn't, and expect me to shrug and merge it | 15:47 |
| Mister_Magister | well i dispelled all the arguments against it not being right fix i believe | 15:48 |
| abr | it might cause other problems elsewhere. we have to be sure that it's the right solution before we merge it | 15:48 |
| Mister_Magister | my testers experienced no issues with the fix | 15:48 |
| abr | you haven't dispelled any arguments. you've just said it's the only right fix and offered nothing to support that assertion. | 15:49 |
| Mister_Magister | you said there's onSceneGraphInvalidated, to wihch i said i tried implementing it and it did not fix it | 15:49 |
| Mister_Magister | i tried fixing it in other repos like qtwayland that was not correct place to fix it ether because it did not fix the isuse | 15:49 |
| Mister_Magister | what other doubts do you have? | 15:49 |
| abr | I give up, and I still have too much else to do. | 15:50 |
| abr | I've spent a lot of time trying to explain this to you but you don't want to listen. | 15:51 |
| abr | please bug someone else | 15:51 |
| Mister_Magister | i do wanna listen you just fail to explain | 15:51 |
| Mister_Magister | when i asked what proof do you want you didn't say | 15:51 |
| Mister_Magister | you said maybe it belongs somewhere else, i did try to fix it somewhere else and it wasn't working | 15:51 |
| Mister_Magister | i literally listened to everything you said? | 15:52 |
| Mister_Magister | Like I don't get what is your problem | 15:52 |
| Mister_Magister | i'm happy to provide anything you want you're just not telilng me what do you want | 15:53 |
| Mister_Magister | and as I said it does'nt have to be you, i made PR to repo owned by Jolla i expect Jolla to merge it, not you specifically | 15:53 |
| Mister_Magister | >you need to provide proof >what proof? >I don't know but proof >what kind of proof? >I don't know you know | 15:56 |
| Mister_Magister | like you're not making any sense | 15:56 |
| Mister_Magister | I harbour no ill will towars you i just want to help SFOS | 15:56 |
| poetaster | Mister_Magister, you need to simply provide a reproducible 'test case' in the PR, I believe and that would simplify things. | 15:58 |
| Mister_Magister | poetaster: literally use qmlglsink and turn the screen off and on like there really isn't anything for me to provide | 15:58 |
| Mister_Magister | I can provide my app or qmlglsink test app but its quite irrelevant, any use of qmlglsink will result in it | 16:00 |
| Mister_Magister | But i'll now prepare long winded in-depth description that'll nice and clearly explain the issue | 16:00 |
| Mister_Magister | as I said I would… because I do listen :P | 16:00 |
| poetaster | Mister_Magister, I'll re-read the PR. | 16:04 |
| Mister_Magister | basically once screen blanks qt yeets QOpenGLContext and creates new one and the gstreamer never notifies anyone | 16:05 |
| Mister_Magister | so all the textures come out black | 16:05 |
| poetaster | Mister_Magister, yeah, I got that but it's been 4 years since I took a crack at GEST as backend for the videoworks app so I have to get up to speed first. | 16:06 |
| Mister_Magister | yeah it's not simple issue, frankly i have no clue | 16:06 |
| poetaster | Mister_Magister, the first thing that popped at me is that the pipeline at the top should be notified when the screen blanks you'd usually go top down, but I haven't been there for ages. | 16:07 |
| poetaster | ah, different scope | 16:08 |
| Mister_Magister | poetaster: the pr does notify the pipeline, thats the thing :) | 16:09 |
| poetaster | Mister_Magister, ok, I'll have a look. Is it an issue with microtube? | 16:11 |
| Mister_Magister | no, as i said any use of qmlglsink | 16:12 |
| Mister_Magister | if you use qmlglsink in any way the issue will be there | 16:12 |
| Mister_Magister | i was using qlmglsink in my picoplayer app but i also used test app and both showed it | 16:12 |
| Mister_Magister | poetaster: abr: here's nice summary https://gist.github.com/Michal-Szczepaniak/ea481fa4280cc52252a30082f46cbf75 | 16:15 |
| Mister_Magister | you can try doing it in qt all you want but if gstreamer never notifies anyone… | 16:16 |
| Mister_Magister | I'll try asking gstreamer folks if they think this fix belongs upstream | 16:17 |
| poetaster | Mister_Magister, the gist is does get the, ah, gist accross :) | 16:18 |
| Mister_Magister | :D | 16:18 |
| Mister_Magister | mind you and i'm not hiding it, the PR and the gist were written by claude because this whole issue is way above my head and I couldn't count on Jolla to fix it and i really want it to get merged because it will help both users and developers | 16:19 |
| poetaster | Mister_Magister, it might really be an upstream patch, but doesn't hurt to test here | 16:19 |
| Mister_Magister | poetaster: yeah will ask the gstreamer guys, although it probably doesn't affect qt5.12 or 6 | 16:20 |
| Mister_Magister | also once you mention qt5.6 nobody wants to listen | 16:20 |
| poetaster | Mister_Magister, ah, hmmmm. ah. | 16:20 |
| Mister_Magister | they say "ah, good luck with 5.6" | 16:20 |
| Mister_Magister | and they completely disregard the issue | 16:20 |
| Mister_Magister | i should deploy nemo on pinephone, theres qt6 there and similar pipeline | 16:21 |
| Mister_Magister | Like i have no intentions of upsetting anyone, all i want is to get this bug fixes ;-; | 16:22 |
| Mister_Magister | fixed* | 16:22 |
| poetaster | sure, got you. | 16:26 |
| poetaster | Mister_Magister, is the source for picoplayer around anywhere? | 16:27 |
| Mister_Magister | my github of course, I'll even update qmlglsink branch for ya | 16:27 |
| Mister_Magister | https://github.com/Michal-Szczepaniak/picoplayer/tree/qmlglsink | 16:28 |
| Mister_Magister | I've also used this https://github.com/dv1/qmlglsink-example | 16:28 |
| Mister_Magister | I've added the two as a comment | 16:29 |
| Mister_Magister | and have asked gstreamer people where should I be asking xd | 16:29 |
| Mister_Magister | poetaster: also 20260522 is too old | 16:32 |
| Mister_Magister | to answer your droidmedia version | 16:32 |
| Mister_Magister | you need this commit https://github.com/sailfishos/droidmedia/commit/81db5a7b2a50083f70f1249d99b9c1a735ff4114 and this commit in gst-droid https://github.com/sailfishos/gst-droid/commit/f06bd0296038a80196f1e79f24e9edf8aff1f48d | 16:33 |
| Mister_Magister | or you can simply disable use of gst-droid then it will work fine | 16:33 |
| Mister_Magister | beacuse turns out that when you use qmlglsink which forces gst-droid to convert droidmedia format to something normal, things go badly xd | 16:34 |
| Mister_Magister | poetaster: abr: i'll forward the PRs to upstream and we shall see | 16:39 |
| poetaster | Mister_Magister, ok, too old means 5.2. Does the bug show up in 5.0 or older? | 17:15 |
| Mister_Magister | poetaster: yes sfos version does not matter | 17:24 |
| poetaster | ok. so GstGLVideoItem is broken. | 17:29 |
| Mister_Magister | poetaster: wdym? | 17:31 |
| poetaster | Mister_Magister, it's the sink item in question. doesn't matter. I see how it's put together. | 17:39 |
| Mister_Magister | poetaster: the sink works fine | 17:39 |
| Mister_Magister | it's just that without the droidmedia/gstdroid fixes you cannot use droidmedia/gst-droid | 17:40 |
| Mister_Magister | but you can use qmlglsink | 17:40 |
| poetaster | sure, I meant in the context of gst-droid, GstGLVideoItem is borked. | 17:42 |
| Mister_Magister | no, i wouldn't say so no | 17:42 |
| poetaster | Well, obviously not all mediaplayer like objects exhibit this effect, so it's specfic to some subset of objects. | 17:45 |
| Mister_Magister | yeah mostly 3 qmlglsink, glupload and glcolorconvert | 17:48 |
| Mister_Magister | I'll have to make the PRs to gstreamer some other day because they require me to understand the code and the code makes my head boil over :D | 18:14 |
| poetaster | it's just the goto statements. ignore those and you'll get it. | 20:05 |
| piggz[m] | 20 GOTO 10 | 20:19 |
| *** Werty_ is now known as Werty | 21:04 | |
| *** n2klb_ is now known as n2klb | 21:05 | |
| *** mia_ is now known as mia | 21:05 | |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!