/srv/irclogs.ubuntu.com/2019/04/08/#ubuntu-desktop.txt

tjaaltonduflu: care to try --enable-xwayland-eglstream?05:34
tjaaltonit's not enabled05:34
duflutjaalton, Yeah I don't think I need to. I just thought you had already done so05:54
dufluAlthough I am almost set up to test it easily today05:54
dufluSo maybe I will05:54
tjaaltoncool05:59
dufluI remember documenting Xwayland rebuilds as a TODO in passing in another bug06:00
dufluBut probably failed to log a bug for that specifically till now06:00
didrocksgood morning06:17
dufluMorning didrocks06:25
* didrocks reboots to answer duflu's new questions. Please think of asking everything at first06:25
didrocksthe ping pong of new -> incomplete is a little bit counter-productive06:25
didrocksespecially when new questions don't have links with previous06:25
* didrocks reboots with drm=106:26
dufludidrocks, I appreciate your frustration. I was working on Sunday to get on top of these new bugs06:31
didrocksduflu: don't work on Sunday. But I don't understand as well the passive-agressivness when it's your coworker to set to incomplete without pinging them.06:38
didrockshappened already on Friday, without pings, when both online, requesting for more info, which is fine but a ping would have been appreciated06:39
dufludidrocks, I wouldn't want to ping you on a Sunday. Even if I am working on a Sunday I keep chat closed06:39
didrocksand same with asking glxgears info. it sounds like the hope is to the reporter to never come back06:39
didrocksthe ping was for Friday06:39
dufludidrocks, that's incorrect. I asked you the question 21 hours ago on Sunday. For good reason, because I just thought of it and wondered if your bug might be related to something else I had discovered. I'm sorry to annoy you. Just trying to figure bugs out.06:41
didrocksbtw, I guess the importance can be reduced now in the bug as it's confirmed to be only drm = 1 and Xorg06:41
didrocksstill sad to have a regression for something that worked for a year though06:41
dufluYeah I agree06:41
didrocksduflu: no, on Friday, you set the bug as incomplete asking those: https://bugs.launchpad.net/ubuntu/+source/mutter/+bug/1823301/comments/306:41
ubot5`Ubuntu bug 1823301 in mutter (Ubuntu) "[nvidia] Shell hangs up at random times when using nvidia-drm.modeset=1" [High,New]06:41
didrockswhile we were both connected06:41
didrocksthen I answered them, had to set back to New, and on Sunday, yeah, the little game restarted06:42
didrocksthat I can understand when we are both not connected, but when people are connected, feels really passive agressive to me06:43
dufluI don't use chat for bugs. I like to keep a permanent record of the information so that (a) no one has to remember; and (b) so that anyone else can pick it up later06:43
dufludidrocks, I don't mean to appear that way. I am just eager to answer bugs quickly, and with a permanent record06:44
didrocksI don't think this is the most effective way to progress on bugs that coworkers have IMHO06:44
didrocksbut meh, let's disagree06:44
dufluAlso, on Friday I had no time left. There would be no sense me asking a question in chat and then pasting it into the bug06:45
didrocks(and again, I mean when both are online at the same time)06:45
didrocksyeah, so maybe that could have been reported to Monday?06:45
dufluThat's not true. I was out of time06:45
duflutjaalton, first attempt of that option failed06:53
dufluMaybe a typo or in the wrong place?06:53
dufluHmm, it still builds even with invalid flags06:58
duflutjaalton, it appears I built it correctly and still no acceleration.07:01
dufluDo you know how to verify the binary?07:01
dufluThough it might be another Nvidia driver bug. I will soon downgrade again and see if things work better07:02
dufluOh. tjaalton, "Xwayland -eglstream" is a thing. Maybe mutter needs to launch it differently?07:04
tjaaltonduflu: ok, i've no idea :)07:05
duflutjaalton, I launched another Xwayland using that option and still get llvmpipe in clients07:05
dufluSo yeah long term we will need that rebuild, but there's some other issue07:05
duflutjaalton, I wonder if we accidentally built mutter with EGLDevice support but no EGLStreams support?07:10
dufluThey are different options07:10
tjaaltonI'll check what fedora did07:34
dufluI guess I can also rebuild mutter easily and check that07:35
oSoMoNgood morning desktoppers07:39
didrockssalut oSoMoN07:40
dufluMorning oSoMoN07:41
oSoMoNsalut didrocks07:42
oSoMoNgood afternoon duflu07:42
RAOFtjaalton: Is GL *meant* to work in XWayland-EGLStream? My understanding was that it enabled *2D* (ie: XRender) acceleration, not GLX.07:54
RAOF(Also duflu, I guess :))07:54
dufluRAOF: I was wondering that too07:56
dufluHmm, and now I am testing 390 for another bug thats probably a new reason why EGLStreams isn't working07:59
tjaaltonRAOF: again no idea, this is new territory for me08:02
tjaaltonbut what about this https://src.fedoraproject.org/rpms/egl-wayland/blob/master/f/10_nvidia_wayland.json08:02
tjaaltonoh, we have that08:02
Laneymoin08:03
RAOFtjaalton: I'm *pretty* sure XWayland won't do GLX even with the EGLStreams patches.08:04
willcookemorning08:05
dufluMorning Laney and willcooke08:05
RAOF(and that's for EGL external platform, which isn't GLX 😀)08:05
tjaaltonRAOF: ah, right.. probably so08:05
dufluSo definitely a bug, and not fixable right now. Definitely avoid Wayland on Nvidia :/08:06
willcookehi duflu08:06
RAOFI don't even think there's a *plan* to make GLX on XWayland-EGLStream work08:07
dufluThat's unfortunate for the Wayland+Nvidia fans08:07
dufluRAOF: Though es2_info also reports LLVMpipe. Same thing?08:08
RAOFAs XWayland doesn't have the loadable DDX system that non-mesa platforms use in Xorg.08:08
RAOFduflu: es2_info queries the X11 EGL backend, which goes through GLX08:09
didrockshey Laney, willcooke08:09
oSoMoNgood morning willcooke, Laney08:09
dufluYeah I thought so08:09
willcookehi didrocks oSoMoN!08:09
willcookeFoggy here today, and not very warm #weatherchat08:09
RAOF(if you built es2_info on the Wayland platform, it should provide NVIDIA details)08:09
dufluSeems X11 will dominate for even longer than I thought08:09
tjaaltonhere's for another 35 years ;)08:10
willcooke:DD08:10
willcookeX11 becomes the new Fortran08:11
Laneyhi duflu tjaalton willcooke RAOF didrocks oSoMoN08:11
willcookemorning Laney, how goes?08:11
dufluwillcooke, if you've seen the Xorg source code it kind of is already08:11
willcooke:D08:11
dufluBut all this hinges on Nvidia still being important for 35 years08:12
tjaaltonduflu: according to this blog post nvidia driver support for xwayland won't arrive before fall https://blogs.gnome.org/uraeus/2019/04/03/preparing-for-fedora-workstation-30/08:12
duflutjaalton, it's fall right now ;)08:12
tjaaltonhaha08:12
dufluSome would say autumn08:12
seb128good morning desktopers08:24
oSoMoNsalut seb12808:24
seb128lut oSoMoN, en forme ?08:24
oSoMoNoui, et toi?08:25
seb128en forme ! passé le w.e en France seul, bien reposant :)08:25
willcookemorning seb12808:26
seb128hey willcooke! how are you? had a good w.e?08:26
willcookeseb128, worst golf competition ever.  I really sucked.  And got caught up with work yesterday.  So I've got that going for me...08:27
willcookeoh, and I feel like I'm getting a cold.08:27
willcooke:DD08:27
seb128:(08:27
seb128doesn't sound like a winning w.e :/08:27
willcooke#2ndtolast08:28
Laney:D08:28
dufluMorning seb12808:28
seb128hey duflu08:28
dufluRAOF: I seem to have a real problem still... glmark2-wayland uses LLVMpipe also08:29
RAOFOk, *that* is a real problem. That's my go-to test of NVIDIA under Mir.08:30
dufluYeah08:31
dufluHmm, fun fact: If your machine is slow enough then mutter can't match Shift with the key you are shifting. Seems it detects those two things a little too disparately08:33
RAOFPutting everything ona single eventloop has some downsides. Who knew!08:42
dufluRAOF: Yeah it's pretty terrible. I have performance bugs in progress for years now because of the single threadedness08:44
dufluAnd trying to work with it08:44
duflutjaalton, RAOF, my problem with glmark2-wayland was that uninstalling Nvidia drivers seems to have autoremoved libnvidia-egl-wayland1. Then you install a new Nvidia driver and everything works except Wayland clients on Nvidia08:45
dufluI feel like that should be autoinstalled or at least never uninstalled if it's that important08:46
tjaaltonnvidia doesn't depend on it08:47
tjaaltonbut there are file conflicts with the old version08:47
tjaaltonso it got removed because of that08:47
RAOFSomething should depend on it, right? It's the external EGL platform.08:48
tjaaltonsomething like.. the driver? :)08:50
dufluRAOF: Your ~/Desktop isn't a symlink to ~ or anything crazy right?09:00
RAOFNope!09:02
RAOFIt's a real (empty) folder!09:02
dufluRAOF: Hmm, 0755 ?09:03
dufluOr 0700 at least09:03
dufluMaybe bcachefs is confusing the file monitor?09:04
RAOF<duflu "Maybe bcachefs is confusing the "> That seems a stretch.09:06
dufluRAOF, yeah well the truth is going to be strange. I can't see any option you can accidentally set for those folders to appear09:07
RAOFMaybe xdg.dirs is set up strangely, though.09:07
RAOF<pre><font color="#0087FF">d</font><font color="#B58900"><b>r</b></font><font color="#DC322F"><b>w</b></font><font color="#859900"><b>x</b></font><font color="#B58900">r</font><font color="#808080">-</font><font color="#859900">x</font><font color="#B58900">r</font><font color="#808080">-</font><font color="#859900">x</font>     <font color="#808080">-</font> <font color="#B58900"><b>chris</b></font> <font09:07
RAOFcolor="#268BD2">10 Dec  2018</font>  <font color="#0087FF">Desktop</font></pre>09:07
RAOFdrwxr-xr-x     - chris 10 Dec  2018  Desktop09:07
RAOFYu.!09:07
RAOFYup!09:07
RAOF```09:08
* RAOF sent a long message: < https://matrix.org/_matrix/media/v1/download/matrix.org/wUSTMYvLZtBdvfdXPJEKhtmN >09:08
RAOFHello!09:08
dufluWe apologise for this temporary break in transmission09:09
dufluRAOF: How about ~/.config/user-dirs.dirs ?09:09
RAOFThat's what I just dumped.-09:09
RAOFEverything points to "$HOME/"09:09
RAOFOh, yeah, IRC.09:10
RAOFSorry.09:10
dufluOK, thanks.09:10
RAOFThat's why it shows $HOME as the desktop, but not why it hangs the compositor for a couple of seconds and leaks 50MB each time the desktop changes 😀09:17
dufluRAOF: I am guessing there's a linear search somewhere, or something else related to 10GB of memory09:21
RAOFIt's not doing something silly like recursively searching the path for desktop files or something?09:22
dufluRAOF: It will probably trigger garbage collection, which is slow and starts every 10 seconds (because you are invoking JS code more frequently than that)09:23
dufluExtra slow across 10GB09:23
dufluSo essentially a memory leak. But not one any other user is likely to hit in a hurry09:24
dufluIt's on my list of performance things to fix still09:24
dufluI hope you are OK with changing those dirs so they don't appear09:25
RAOFOh, I've already changed that file.09:26
RAOFAnd it hangs the compositor from the get go, even when the memory usage is only 300MB09:26
dufluWeird. Some other customisation or upgrade bug no doubt09:27
=== cpaelzer__ is now known as cpaelzer
RAOFEditing files on the desktop seems like a not-infrequent use case.09:29
RAOFWhich would also trigger this, as autosave kicks in.09:29
RAOF(but not as quickly)09:29
RAOFAnyway, I'll switch to the plain GNOME session and tell you how long it takes to leak 10G of memory 😀09:30
dufluYeah there are leaks but those will likely be different leaks09:30
dufluI guess it's up to me to come up with a user-friendly procedure for enabling heap profiling09:31
dufluand retrieving results09:31
RAOFThat'd be nice!09:39
RAOFYou could also just point me at some documentation.09:39
RAOFIt doesn't have to be *terribly* user-friendly 😀09:39
dufluRAOF: Heh in that case: https://gperftools.github.io/gperftools/heapprofile.html09:43
dufluAlso, isn't it night time?09:44
dufluYou could set it up system wide but disabled till you send a signal09:45
dufluI've never tested the stability of that though09:45
dufluOops. That's the CPU profiler only that's off till you turn it on09:46
seb128ricotz, hey, did you look at libdmapsharing failing to build with the new vala?11:22
ricotzseb128, oh, I thought I fixed this upstream11:30
seb128ricotz, I looked at the commits in git but maybe the title/description wasn't obvious to me11:30
andyrockrbalint: hey hey can you create an ubuntu/bionic branch for software-properties ?11:34
ricotzseb128, hmm, I think this was something weird about the source management there since the generated source are in git11:37
seb128ricotz, k, I'm not sure to understand what it means for the patch though, do you have one we can use?11:42
ricotzseb128, https://paste.debian.net/plain/107665311:43
ricotzI assume the package carries further patches to resolve a linker issue?11:44
seb128ricotz, thx11:44
seb128ricotz, no, no patch atm, build might still fail then, let me try11:45
ricotzah, I built the master branch11:46
ricotzseb128, https://paste.debian.net/plain/107665411:50
seb128ricotz, thx!11:51
ricotzLIBDMAPSHARING_3_0 branch has more issues :\11:52
ricotznot surprising with the last commit being from 201711:52
rbalintandyrock, looking11:59
rbalintandyrock, done12:07
andyrockrbalint: out of curiosity which command did you use to generate it?12:34
andyrockand thanks!12:34
=== chrisccoulson_ is now known as chrisccoulson
jibeltsimonq2, Hey, if you want to play with a multi-layer image, there is today's build at http://people.canonical.com/~platform/ubuntu-canary/20190408/12:53
jibelit's completely untested12:54
seb128ricotz, do you plan to mp the fix on gitlab?13:02
ricotzseb128, did it work out? -- no, not in its current state, there are more issues with the vapi build in 3_0 and the unlinking in master13:03
ricotz*underlinking13:04
seb128ricotz, yes, it fixed the build applied on the current package13:04
seb128well, I'm going to open at least an issue then so I've a reference for the patch header13:04
seb128thx13:04
ricotzok, thanks13:05
tsimonq2jibel: Awesome, thank you!13:12
seb128ricotz, https://gitlab.gnome.org/GNOME/libdmapsharing/issues/713:12
tsimonq2Does this mean ubuntu-cdimage has to put the images elsewhere? How'd you get those there?13:12
gitbotGNOME issue 7 in libdmapsharing "2.9.39 fails to build with vala >= 0.43" [Opened]13:12
didrockstsimonq2: we proposed some patches for it to be reviewed by sil2100. Unsure if this is enough though13:16
didrockstsimonq2: /!\ the script has some small failures, we are testing/fixing13:16
tsimonq2ack13:17
tsimonq2didrocks: Just to ubuntu-cdimage?13:18
didrockstsimonq2: yep https://code.launchpad.net/~jibel/ubuntu-cdimage/support_ubuntu_subproject/+merge/36559813:19
didrocksbut maybe some other bits are needed13:20
sil2100eeek13:20
* sil2100 forgot about the MP last week13:20
sil2100Looking at it in a bit13:20
didrocksno worry sil2100, this is why we asked you to copy the current image :)13:20
didrockssil2100: thanks! Maybe that's not enough, but we are quite sure this is needed13:21
seb128andyrock, hey, did you see my ping about https://gitlab.gnome.org/GNOME/gnome-control-center/merge_requests/423 on friday?13:36
gitbotGNOME issue (Merge request) 423 in gnome-control-center "online-accounts: Don't segfault if get_all_providers_cb is called during init" [1. Crash, 6. Component: Online Accounts, Opened]13:36
andyrocknope :)13:37
andyrockbut to fix it properly upstream the best way is to change goa API13:37
andyrockto make it sync or cancellable13:37
andyrockI'll take another look later this week13:38
seb128andyrock, k13:40
=== ddstreet_away is now known as ddstreet
andyrockrbalint: https://code.launchpad.net/~azzar1/software-properties/+git/software-properties/+merge/365668 can you take a look? thx!13:55
rbalintandyrock, i used pull-lp-source -d and gbp import-dscs , with some shell history, but don't have a single script yet14:29
brlinCan anyone review this? : https://code.launchpad.net/~buo-ren-lin/gnome-system-monitor/fix-snap/+merge/36476815:19
=== ecloud_ is now known as ecloud
Trevinhomorning folks15:45
willcookehi Trevinho15:47
popeyhttps://usercontent.irccloud-cdn.com/file/Na1tpRPD/Current%2019.04%20on%20nVidia15:48
popey^^ Just updated my 19.04 rig and now I have *massive* fonts in the indicator and shell activities thing. Is this known?15:48
willcookepopey, I think didrocks and/or duflu saw that before, but they're both EOD now.15:58
popeykk15:58
willcookeIf you can ask again tomorrow morning, please do.  I will try and remember15:58
popeywill do15:59
popeywillcooke: just switched to 200% scale and back again and it made the big stuff go away16:16
Eickmeyerclobrano: I created a patch based on the upstream commit in materia-gtk-theme made by the author. I can confirm that it works. Thanks for all the help!16:18
clobranoEickmeyer: great!16:19
willcookepopey, ah, kk. That sounds familiar16:22
Laneythat's what didier has been reporting16:37
cyphermoxLaney: are you preparing a new ubiquity upload?16:46
Laneynope, but I could do16:48
LaneyI think I asked on one of the MPs :>16:48
cyphermoxoh, I didn't see16:48
cyphermoxit's fine, I was about to go merge rik's changes16:48
Laneygo for it16:48
cyphermoxI can do the upload if you want me to16:48
cyphermoxok16:48
Laneyubiquity uploads are annoying for me16:48
Laneygbp buildpackage / debuild don't make a proper package (they lose the d-i stuff)16:49
Laneyso I have to use one of those and then unpack the package it makes, run debian/rules update again and rebuild the source package16:49
Laneydunno why16:49
cyphermoxoh16:50
cyphermoxI just make -C d-i/  and gbp buildpackage --git-ignore-new or whatnot16:50
cyphermoxafter making sure things are properly committed, of course, but we don't want to include and commit d-i bits16:51
cyphermox(in the upload yes, in the tree no)16:51
Laneyright16:53
=== pstolowski is now known as pstolowski|afk
Trevinhopopey: yeah, it's already fixed in a MP waiting for sponsor18:44
Trevinhopopey: you can also try pkgs at https://launchpad.net/~ci-train-ppa-service/+archive/ubuntu/369518:56
=== tdaitx_ is now known as tdaitx
ricotzseb128, could you pick up vala 0.44.3 bugfix release from debian git?19:24
seb128ricotz, is anyone going to upload it (in which case we can sync)?19:25
ricotzseb128, not sure I waited a week and 0.44.2 wasn't uploaded19:25
ricotzj_bicha seems afk for now19:26
seb128ricotz, k, I have a look but probably tomorrow it's getting late tonight and I don't plan to stay at the computer long19:26
ricotzseb128, ok, please don't forget it19:26
seb128yeah, he said he needed to step back from Debian/Ubuntu work and that he would blog about it19:26
seb128k19:26
ricotzthanks19:27
seb128ricotz, you don't have any idea about bug #1823426 by any chance? ;)19:29
ubot5`bug 1823426 in librsvg (Ubuntu) "librsvg ftbfs in disco (i386 only)" [High,New] https://launchpad.net/bugs/182342619:29
seb128     = note: `N1024` must be defined only once in the type namespace of this module19:29
seb128For more information about this error, try `rustc --explain E0428`.19:29
seb128error: Could not compile `typenum`.19:29
seb128I wonder if that's a rust toolchain issue or a librsvg one19:30
seb128(I didn't find a report/commit upstream in librsvg that seems to correspond)19:30
ricotzseb128, librsvg git built on disco earlier today19:31
seb128the issue is i386 only19:32
seb128I guess you tried on amd64?19:32
ricotzah correct19:32
tjaaltonRAOF: intel-opencl-clang and -graphics-compiler are now on the disco NEW queue, if you have time to have a look. -opencl-clang has an older upload there that should be dropped21:24

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!