| jbicha | duflu: good morning | 00:49 |
|---|---|---|
| jbicha | are you interested in merging mpv for the libva transition in bionic? | 00:50 |
| jbicha | also we had someone complaining about window decorations for mpv (but I don't use mpv) https://ubuntuforums.org/showthread.php?t=2376602 | 00:51 |
| duflu | jbicha, maybe... and yeah lack of server side window decorations is a Wayland "feature" for now | 00:57 |
| JanC | jbicha / duflu: people can get window decorations for mpv on Wayland by adding a 'vo=x11' line in '~/.config/mpv/mpv.conf', which makes it use XWayland :) | 03:09 |
| JanC | it's not perfect, but it works... | 03:09 |
| jbicha | that user claimed that he was sure he was using Wayland mpv with decorations on Arch | 03:10 |
| JanC | well, I switched from Wayland back to X11 because it has too many bugs still... | 03:14 |
| duflu | JanC, if I recall correctly that will disable hardware acceleration support. Because Xwayland only supports DRI3 and VAAPI only supports DRI2 | 03:16 |
| duflu | So not an option we can recommend | 03:16 |
| JanC | hm, I think mpv said it was using vaapi, but I'm not 100% sure | 03:17 |
| duflu | JanC, interesting. Please do check | 03:17 |
| JanC | maybe only for decoding | 03:17 |
| duflu | JanC, or try just comparing CPU usage in 'top' | 03:19 |
| duflu | *and* try just comparing CPU usage in 'top | 03:19 |
| duflu | ' | 03:19 |
| JanC | I might be totally wrong, but IIRC VAAPI can decode straight to video memory, or into a buffer, so possibly they use VAAPI for decoding, then copy to the screen buffer, or something? | 03:20 |
| JanC | which would of course cause higher CPU too | 03:20 |
| JanC | or they use OpenGL and VAAPI decodes into an OpenGL buffer (or whatever it's called) | 03:23 |
| JanC | duflu: if I remember next time I (have to) reboot I'll try | 03:23 |
| JanC | but I'm happy I don't have to reboot every 1-3 days as with Wayland :P | 03:24 |
| JanC | on X11 | 03:24 |
| duflu | JanC, yes all VAAPI use cases decode straight into video memory. At least the ones we use. There is no distinction there | 03:24 |
| duflu | One of the reasons why we never had hardware accelerated video decoding till 17.10 was that nobody knew or could tell if it was ON or OFF already. So I have tried to cover that here: https://wiki.ubuntu.com/IntelQuickSyncVideo | 03:26 |
| duflu | Essentially you should always have single digit CPU usage or close to it. If not, then something still needs improving | 03:27 |
| JanC | on Wayland the desktop gets unusably slow very quickly (sometimes after less than a day), and if I don't reboot myself in that case it hangs at some point with "Key repeat discarded, Wayland compositor doesn't seem to be processing events fast enough!" errors in dmesg | 03:28 |
| JanC | (I still have to file a bug about that) | 03:28 |
| JanC | "hangs" = doesn't update the screen; apps keep running, as e.g. music keeps playing | 03:29 |
| duflu | JanC, OK that's a completely different topic.... if you find that relates to memory usage of the gnome-shell process then subscribe here: https://bugs.launchpad.net/bugs/1672297 | 03:30 |
| ubot5 | Launchpad bug 1672297 in gnome-shell (Ubuntu) "gnome-shell uses lots of memory, and grows over time" [High,Triaged] | 03:30 |
| JanC | I'm not sure what causes it (I switched to X11 to see if it was Wayland-only, and that was 8 days ago ;) ) | 03:32 |
| JanC | I'll look at that bug and see if I can confirm it's related, but that was just to explain why I haven't been using Wayland while trying to get some work done :) | 03:34 |
| JanC | mpv has some "vaapi-copy" mode for hardware decoding ("copies video back into system RAM (Linux with Intel GPUs only)" according to the manpage) | 03:38 |
| JanC | I'm not sure exactly what it does or how it works :) | 03:38 |
| roasted | a 4k video with vaapi results in about 3% CPU usage on an i5 7200u processor | 03:39 |
| roasted | same video with vaapi-copy, 19% | 03:39 |
| roasted | no hardware accel (software only), ~60% | 03:39 |
| roasted | not to mention my battery life went from 18 hours (idle) to 14:30 using vaapi, and 7:30 using software accel (no hardware accel) | 03:40 |
| roasted | I was pretty certain I had window decorations working with mpv in Antergos + Wayland and was using gpu-context=wayland in my mpv.conf, but later I began to get errors when I launched mpv with gpu-context=wayland in mpv.conf | 03:41 |
| roasted | I have no idea if I was troubleshooting too much at once and messed up my interpretation or if I got updates later that made it different, I'm not sure. | 03:42 |
| duflu | Yes roughly speaking, vaapi is an order of magnitude more efficient than vaapi-copy, and vaapi-copy another order of magnitude more efficient than software | 03:45 |
| roasted | The only downside I've seen is some youtube videos impose heavy gray artifacts when using vaapi. Not all, but some. | 03:45 |
| roasted | https://i.imgur.com/kBd3wf1.png | 03:46 |
| roasted | for local videos, and most online videos, it's brilliant. | 03:46 |
| JanC | that looks like there might have been video corruption | 03:47 |
| roasted | could be. If I comment out vaapi lines in my mpv.conf and rerun the same video in mpv, it's fine. | 03:47 |
| JanC | I mean, dropped frames, or such | 03:47 |
| roasted | albeit with my proc zooming but it functions | 03:48 |
| duflu | roasted, I will try to look into the grey corruption this week | 03:48 |
| roasted | regarding the GTK3 patch, it was my understanding that the patch introduces SSD via a KDE Protocol, thus allowing Wayland and apps like MPV to coexist. | 03:48 |
| roasted | my confusion that sparked the forum post was the fact that I didn't see anything really change. | 03:49 |
| roasted | I can bounce over to my Antergos partition and re-test then. I've been meaning to take another look sometime tonight. But if it's broken there then it stands to reason it could be a waiting for upstream thing yet. | 03:49 |
| roasted | duflu: pardon my ignorance, are you an mpv dev? mpv packager for ubuntu? | 03:50 |
| duflu | roasted, presently jack of all trades, and specifically responsible for Ubuntu hardware acceleration across the board | 03:50 |
| roasted | ah. well. kudos. :) loving it. | 03:50 |
| duflu | Not an upstream mpv dev | 03:51 |
| duflu | roasted, cool. Me too. I think people fail to realize it's not just about using less power, but also that modern 4K videos are simply unplayable without it | 03:51 |
| duflu | (modern = H.265/HEVC) | 03:52 |
| roasted | yeah. I hear you. I've been less-than-thrilled by some projects (ahem, mpv) insistance on pushing forth software acceleration | 03:52 |
| roasted | I've shot a lot of 4k video over the last week and it's certainly a necessity at this point | 03:52 |
| duflu | Interestingly somewhat by luck it's a technology feature Ubuntu introduced at about the same time as Apple did in High Sierra | 03:53 |
| JanC | I guess the grey corruption stuff could also happen when VAAPI doesn't implement some decoder feature, or implements it incorrectly | 03:53 |
| roasted | duflu: if there's anything within the capabilities of a "pretty fluent end-user" that I can do to test or help out regarding this please ping me. :) | 03:53 |
| duflu | roasted, OK thanks. Actually if you have raw HEVC files from an iPhone 8/X or from a GoPro Hero6 I would like some for testing | 03:54 |
| roasted | duflu: FYI, that gray artifact stuff, I noticed it happening a lot on youtube's more popular/top charted/main page vevo videos. One of Taylor Swift's recent music videos was where I snapped that imgur shot. | 03:54 |
| roasted | well, dang. not an iphone user, and a gopro is on my christmas list. :( | 03:54 |
| roasted | I just have a moto g5 atm | 03:55 |
| duflu | Such is the modern era; everything is on YouTube and people don't share their actual files. So I haven't found any to test with | 03:55 |
| roasted | I don't even know anybody with an iphone, sans a few users on very old models | 03:56 |
| roasted | nothing recent though | 03:56 |
| JanC | roasted: probably important to know what video quality you were playing too... | 03:57 |
| JanC | as YT usually has many | 03:57 |
| roasted | JanC: playing? I'm not sure. I didn't set anything in particular in my mpv.conf | 03:57 |
| roasted | I wonder what mpv in 17.10 repos && what mc3man's PPA of mpv default to | 03:57 |
| JanC | IIRC mpv uses the highest by default | 03:57 |
| JanC | for some value of "highest" | 03:58 |
| === Guest16249 is now known as RAOF | ||
| JanC | or maybe youtube-dl's default, which is that | 03:58 |
| roasted | could be. I'm not sure offhand. I recall seeing some flags to call specific qualities, though... | 03:59 |
| JanC | when run from the commandline, it can tell you the exact URL | 04:00 |
| JanC | which might be useful for duflu to get the same | 04:00 |
| JanC | (he can even download it then) | 04:02 |
| oSoMoN | good morning desktoppers | 06:44 |
| duflu | Morning oSoMoN | 06:47 |
| oSoMoN | hey duflu, good afternoon! | 06:50 |
| didrocks | good morning | 07:14 |
| oSoMoN | salut didrocks | 07:23 |
| didrocks | ça va oSoMoN ? | 07:24 |
| oSoMoN | bien, et toi? | 07:24 |
| didrocks | j'ai été au lit tout le week-end :/ | 07:25 |
| didrocks | fièvre et rhino | 07:25 |
| didrocks | je commence tout juste à émerger | 07:25 |
| oSoMoN | ah merde, pas cool | 07:34 |
| didrocks | on survivra :) | 07:35 |
| oSoMoN | bon rétablissement! | 07:39 |
| didrocks | merci ;) | 07:45 |
| seb128 | good morning desktopers | 07:48 |
| seb128 | lut didrocks oSoMoN | 07:48 |
| jamesh | hi seb128 | 07:52 |
| seb128 | hey jamesh, how are you? | 07:53 |
| jamesh | good. Went for another ride yesterday training for a charity bike ride later this month | 07:54 |
| jamesh | 60 km around the river | 07:54 |
| seb128 | nice | 07:56 |
| jamesh | the charity ride is going to be 70 km, but should be a bit easier, going on roads and not having to stop as much | 07:57 |
| Guest86060 | ahoy there | 09:02 |
| Guest86060 | GOD | 09:02 |
| Guest86060 | THIS AGAIN | 09:02 |
| Guest86060 | i should fix that logon thing | 09:03 |
| === Guest86060 is now known as Laney | ||
| * Laney tries to remember how to do that | 09:03 | |
| jamesh | setting your nickserv password as the irc network password seems to be the most reliable option | 09:05 |
| Laney | nick:password | 09:06 |
| Laney | I remember, a couple of weeks ago I switched to the SSL certificate thing and it doesn't work properly for me | 09:06 |
| * Laney tries | 09:06 | |
| Laney | seemed to work | 09:07 |
| === pesari_ is now known as pesari | ||
| willcooke | morning all | 09:24 |
| jibel | morning willcooke | 09:28 |
| willcooke | hi jibel, wb | 09:30 |
| willcooke | good hols? How did the move go? | 09:30 |
| jibel | holidays were fantastic. Sun and high temperature end of October is priceless | 09:30 |
| jibel | and the move is done, one more before next :) | 09:31 |
| willcooke | :)_ | 09:33 |
| seb128 | hey willcooke | 09:37 |
| seb128 | lut jibel | 09:37 |
| Merlijn_S | I found a potentially critical issue: Ubuntu 17.10 fails to boot after a firmware update: https://github.com/rhboot/fwupdate/issues/86 | 09:41 |
| oSoMoN | salut jibel ! | 09:41 |
| Merlijn_S | I had this issue myself, and now another user has reported the same issue. | 09:42 |
| Merlijn_S | Bug in launchpad: https://bugs.launchpad.net/fwupd/+bug/1730343 | 09:42 |
| ubot5 | Launchpad bug 1730343 in gnome-software (Ubuntu) "firmware update breaks Ubuntu" [Undecided,New] | 09:42 |
| Merlijn_S | Sadly, there is no way to debug this issue because Ubuntu _still_ doesn't preserve logs from previous boot.. | 09:43 |
| Merlijn_S | It might be a good idea to disable firmware update in the software center until this issue is resolved. | 09:43 |
| duflu | Hmm, I didn't know it was even an option to forget history. Seems we do (journalctl --list-boots) | 09:45 |
| Merlijn_S | The issue with forgetting history is reported here: https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/1618188 | 09:47 |
| ubot5 | Launchpad bug 1618188 in ubuntu-meta (Ubuntu) "systemd journal should be persistent by default: /var/log/journal should be created; remove rsyslog from default installs" [Wishlist,Triaged] | 09:47 |
| Merlijn_S | The fix is incredibly simple, but they don't want to enable it because that might cause some logs to be written to disk twice. | 09:48 |
| Merlijn_S | max 2x disk space for logs seems like a small price to pay for the benefits of being able to debug fwupd, but that's another issue.. | 09:49 |
| Merlijn_S | Can I do something else to get the desktop team's attention for the fwupd issue, or do they check this channel regularly? | 09:50 |
| duflu | Merlijn_S, they should mostly be here now (mostly in Europe) | 09:52 |
| Merlijn_S | Great, thx | 09:52 |
| duflu | Merlijn_S, although as is always the case, it's not that simple. It seems to be in the hands of the Ubuntu foundations people, not desktop :) | 09:53 |
| Merlijn_S | How do you mean? | 09:53 |
| duflu | And foundations sounds right. Foundations = internals, Desktop = graphical etc | 09:54 |
| Merlijn_S | oh, ok | 09:54 |
| duflu | Merlijn_S, try emailing Dimitri and Bryan directly, I recommend | 09:56 |
| duflu | Likely they just lost the email thread for the bug | 09:56 |
| Merlijn_S | Which Bryan are you talking about? | 10:01 |
| duflu | Merlijn_S, the only one mentioned in the bug :) | 10:02 |
| Merlijn_S | oh, ok, I searched the thread for "brian" instead of "bryan" xD | 10:03 |
| jbicha | good morning | 10:18 |
| willcooke | morning jbicha | 10:30 |
| jbicha | Merlijn_S: did you link to the wrong github issue for 1730343 ? | 10:32 |
| Merlijn_S | Yes, my bad! | 10:34 |
| Merlijn_S | jbicha: fixed, thanks! | 10:36 |
| Laney | Merlijn_S: what about pulling this bad update from the LVFS? | 11:27 |
| oSoMoN | seb128, do you know what keeps re-triggering the amd64 build for https://launchpad.net/ubuntu/+source/libreoffice/1:5.4.2-0ubuntu2 ? | 11:28 |
| oSoMoN | there's a consistent unit test failure for which I'm backporting a debian patch, rebuilding over and over won't help | 11:28 |
| jbicha | oSoMoN: I did a rebuild once on Saturday because sometimes LO has a one-time test failure, but other people have clicked the retry button since | 11:31 |
| oSoMoN | ack, so anyone please refrain from retrying that, I'll have another upload ready during the day | 11:37 |
| jbicha | I clicked Cancel on that build now | 11:42 |
| Merlijn_S | Laney: It's not caused by the update itself, it seems, it's more of a race condition in fwupd itself. The annoying thing is that I can't reproduce it by downgrading and upgrading again, but another user has it too, so it's obviously still happening.. | 12:01 |
| seb128 | oSoMoN, likely people who are unsure of the issue is transient or not and want to see the build green so do a retry to see | 12:21 |
| seb128 | oSoMoN, I've been one of those :p | 12:21 |
| seb128 | oSoMoN, I assumed that it failed to build realiably on amd64 you would have hit that issue before upload so that it meant that it's probably random | 12:22 |
| oSoMoN | seb128, yeah, looks like the issue is due to another update in the archive, although the upstream patch in debian doesn't elaborate on which | 12:22 |
| oSoMoN | it was building in my PPA before the upload | 12:23 |
| seb128 | k | 12:23 |
| seb128 | start of cycle fun :p | 12:23 |
| seb128 | things keep changing under you | 12:23 |
| oSoMoN | what I don't understand though is that the new test build in my PPA which I uploaded 6+ hrs ago failed (for another reason apparently), and when I went back to it to see the log I saw that it had been restarted | 12:24 |
| oSoMoN | so now I don't have a build log, and I have to wait for another 6 hours | 12:25 |
| seb128 | is that your own ppa? | 12:25 |
| seb128 | or a team one? | 12:25 |
| oSoMoN | yes, my own | 12:25 |
| seb128 | weird :-/ | 12:25 |
| seb128 | that's worth asking on #launchpad how that's possible | 12:25 |
| oSoMoN | and no, my other schizophrenic self didn't retry it | 12:26 |
| seb128 | the one reason I could see if that they had an infra issue and decided to restart the build from everything that hit it | 12:26 |
| seb128 | but I would ask on #launchpad, just feels weird to me | 12:26 |
| oSoMoN | yes, I'll do that | 12:26 |
| seb128 | thanks | 12:34 |
| === dobey_ is now known as dobey | ||
| === alan_g is now known as alan_g|lunch | ||
| === alan_g|lunch is now known as alan_g | ||
| === pavlushka_ is now known as pavlushka | ||
| oSoMoN | seb128, so according to Colin, what happened in my | 16:21 |
| oSoMoN | PPA can happen if the build crashes the builder | 16:21 |
| oSoMoN | (stupid misplaced return key) | 16:21 |
| oSoMoN | or if LP looses connectivity to the builder for too long | 16:22 |
| seb128 | oSoMoN, oh ok, good to know | 16:57 |
| seb128 | I guess they consider it as a builder issue and auto retry in such cases? | 16:57 |
| oSoMoN | seb128, yeah, and when the builder crashes the logs can't be retrieved anyway, so it makes sense to trigger a new build | 16:58 |
| oSoMoN | have a good evening everyone! | 17:37 |
| === alan_g is now known as alan_g|EOD | ||
| === JanC is now known as Guest24940 | ||
| === Guest48995 is now known as fredp | ||
| === fredp is now known as Guest25304 | ||
| willcooke | night all | 20:47 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!
