[12:33] <Chevek> Hi, I would like to propose a patch for #lp2057792  https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2057792 . It is a simple text file to add to the systemd package for /etc/sysctl.d/. I cannot find useful documentation on how to proceed. https://canonical-ubuntu-packaging-guide.readthedocs-hosted.com/en/latest/tutorial/ is mostly empty and
[12:33] <Chevek> older version I could find are obsolete. I'm reaching you for help on this.
[12:33] -ubottu:#ubuntu-devel- Launchpad bug 2057792 in linux (Ubuntu) "Some Games are crashing linked to a vm_max_map_count too low" [Undecided, Confirmed]
[12:35] <Chevek> I would like to build a systemd package with the patch on my ppa to provide the bug report some testing about the change.
[15:30] <tsimonq2> Chevek: I can follow up in a bit, otherwise I'd recommend ubuntu-devel-discuss@lists.ubuntu.com
[15:30] <tsimonq2> We're working on enhancing this documentation but we're also busy with Noble. :)
[15:30] <tsimonq2> !dmb-ping | Gooooood morning
[15:31] <tsimonq2> We have two pending packageset updates, and I've had to play sponsor lately.
[15:32] <tsimonq2> vorlon volunteered to write some scripts to make this view clearer but I haven't heard about progress on that front yet.
[15:32] <tsimonq2> https://lists.ubuntu.com/archives/devel-permissions/2024-March/002466.html
[15:33] <tsimonq2> https://lists.ubuntu.com/archives/devel-permissions/2024-March/002483.html (he neglected to mention this is for Jammy *and* Noble)
[15:33] <tsimonq2> We also have https://lists.ubuntu.com/archives/devel-permissions/2024-February/002458.html
[15:34] <tsimonq2> Please at least review these requests.
[17:04] <vorlon> tsimonq2: backburnered due to time_t; someone else could take a stab at it
[17:04] <vorlon> it's "germinate Ubuntu, germinate Kubuntu, set subtraction"
[17:41] <Chevek> tsimonq2 It should be easy, I just want to add a file with this line "vm.max_map_count=1048576" (and some comments to explain the situation), with name "10-vm-max-map-count.conf", which should be located in /etc/sysctl.d/. This should be achieved in the systemd package, e.g. here
[17:41] <Chevek> https://code.launchpad.net/~git-ubuntu-import/ubuntu/+source/systemd/+git/systemd/+ref/ubuntu/mantic-updates in order to test this on mantic with a PPA.
[18:50] <juliank> FWIW I've done anki appmenu-gtk-module aqemu atomes caribou
[18:53] <juliank> cegui-mk2
[18:53] <juliank> chatty
[18:55] <juliank> clearlooks-phenix-theme
[18:58] <juliank> cpupower-gui
[19:01] <juliank> cqrlog
[19:06] <dbungert> looking at wiipdf
[19:08] <juliank> deepin-movie-reborn
[19:08] <juliank> I'd have expected it to be a leaf, but it is not apparently?
[19:09] <juliank> ah it is
[19:09] <juliank> vorlon: please remove-package -b -a armhf deepin-movie libdmr-devlibdmr0.1
[19:09] <juliank> vorlon: please remove-package -b -a armhf deepin-movie libdmr-dev libdmr0.1 (fixed)
[19:10] <dbungert> looking at xbase64
[19:15] <juliank> Yeah no, not fixing deepin-movie, this is silly
[19:15] <juliank> * dhcp-probe
[19:20] <juliank> emacs-pdf-tools
[19:23] <juliank> emerald
[19:30] <juliank> exaile
[19:31] <juliank> fastdds
[19:33] <dbungert> unity-gtk-module
[19:36] <juliank> Turns out libfastcdr1t64 has wrong symbols file still saying libfastcdr1, hence fastdds got wrong dep
[19:36] <juliank> Updating Debian NMU and syncing
[19:37] <juliank> So fastdds is going to need another no-change rebuild once fixed fastcdr has been built
[19:43] <dbungert> telepathy-mission-control-5
[19:43] <juliank> vorlon: in the meantime while we get new fastcdr built; remove-package -b -a armhf fastdds-tools libfastrtps-dev libfastrtps2.11 fastddsgen sounds reasonable (fastddsgen is a leaf afaict, script doesn't show it, and is reverse-testsuite-trigger)
[19:43] <juliank> not sure this blocks anythging
[19:43] <juliank> freeipa
[19:50] <dbungert> rapid-photo-downloader
[19:53] <juliank> dbungert: You went a bit wrong there with telepathy-mission-control5, libmission-control-plugins0t64 is the right library, the depends was correct, the culprit is the libmission-control-plugins0 dependency
[19:53] <juliank> ▸   dh_installdocs -plibmission-control-plugins-dev \
[19:53] <juliank> ▸   ▸   --link-doc=libmission-control-plugins0
[19:54] <juliank> This creates it and needs to be adjusted, not debian/control
[19:54] <juliank> dbungert: The global-ben.rebuild-for.txt shows you the library name that we want, not the broken one :)
[19:58] <dbungert> juliank: thanks for looking.  I'll check again.
[19:58] <juliank> Should I add versions to the report?
[19:58] <juliank> Not sure
[19:59] <juliank> Ah no, then rebuild-for picks up the old version even if somebody uploaded a newer one
[19:59] <juliank> that would be silly
[20:02] <jbicha> Chevek: I believe properly fixing #2057792 at the distro level should be done through the kernel configuration and not through adding a file to /etc/sysctl.d/
[20:02] <jbicha> bug 2057792
[20:02] -ubottu:#ubuntu-devel- Bug 2057792 in linux (Ubuntu) "Some Games are crashing linked to a vm_max_map_count too low" [Undecided, Confirmed] https://launchpad.net/bugs/2057792
[20:02] <jbicha> kernel's build configuration
[20:05] <dbungert> juliank: ok, double checking telepathy-mission-control5, I made a legit change to reduce hardcoding but I'll revisit the doc part
[20:06] <juliank> dbungert: I don't agree fwiw, libmission-control-plugins0t64 is in the same source package, the dependency is there to enforce a = relationship on the same binary version, probably due to use of private APIs
[20:07] <juliank> be careful :)
[20:08] <juliank> It just gets sed at ABI bumps anyhow :)
[20:09] <juliank> I'll sign out, it's 9pm on a Saturday, but of course IRC pings go to my phone so be careful :)
[20:09] <juliank> I should eat some dinner
[20:12] <doko> nice, the list is half the size now
[20:15] <doko> looking at keybinder
[20:16] <dbungert> I'm double checking all mine based on Julian's feedback above
[20:18] <vorlon> juliank: happy to - appreciate it if you point to the reason as well though for completeness
[20:19] <vorlon> juliank: (in part because I open proposed-migration bugs for packages where appropriate for tracking, and tag them time-t)
[20:19] <juliank> vorlon: reason: depends on lots of hardcoded libs and too lazy to fix for leaf package
[20:20] <vorlon> juliank: for deepin-movie-reborn?
[20:20] <vorlon> fwiw it showed as a candidate on https://ubuntu-archive-team.ubuntu.com/proposed-migration/update_output_notest.txt
[20:20] <doko> looking at ibus
[20:20] <vorlon> but maybe broken only on armhf
[20:22] <juliank> vorlon: yeah for deepin-movie-reborn
[20:22] <vorlon> ack
[20:22] <juliank> I think the fastdds has explanation
[20:23]  * juliank is busy cooking spaghetti with Beyond meatballs
[20:23] <vorlon> juliank: just because it bothers me, can you say why fastcdr's .symbols file was wrong? was this a case of multiple libraries in the same package? (early versions of the script unfortunately only fixed the dependency for the first one)
[20:24] <juliank> vorlon: sorry I'm afraid I don't know, I think there's only lib in there
[20:25] <vorlon> juliank: and yeah, I'll file a bug on the deepin-movie-reborn one - just because deps on the old package name are satisfiable doesn't mean we want to ship with them around, as you well know that can make the resolver angry on upgrade
[20:30] <vorlon> juliank: removed fastdds binaries, but fastddsgen seems to only have a test dep?
[20:31] <juliank> Yeah
[20:31] <juliank> Can keep it and force-badtest it?
[20:31] <vorlon> badtest which?
[20:31] <juliank> fastddsgen/armhf if it is passing now
[20:32] <juliank> Otherwise removed test dep would be a regression
[20:32] <juliank> I'm on my phone tossing pasta I haven't checked
[20:33] <vorlon> I'll migration-proposed/0 it
[20:37] <vorlon> LocutusOfBorg: libfm built, trying the copy again
[20:39] <dbungert> juliank: OK, I understand what you're saying now on telepathy-mission-control-5.  Reverting previous upload and fixing libmission-control-plugins-dev
[21:58] <Chevek> jbicha: One could change this line https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/include/linux/mm.h?h=v6.2.12#n203  but I wonder if there is a value that fits all use cases. Having a file in /etc/sysctl.d/ seems a better way for people to adjust the value to their needs, don't you think ?
[22:05] <Chevek> This value seems hardcoded in the kernel source. I'm not aware of a possible configuration for this value, except patching the source file directly.
[22:09] <Chevek> AFAICT either you add a file to /etc/sysctl.d/, either you patch the kernel like this https://aur.archlinux.org/cgit/aur.git/tree/max_map_count.patch?h=linux-steamos
[22:12] <Chevek> Fedora choose the /etc/sysctl.d/ way : https://src.fedoraproject.org/rpms/systemd/blob/f39/f/10-map-count.conf
[22:32] <juliank> File probably ought to be shipped in procps
[22:32] <juliank> procps is what ships almost all sysctl.d files
[22:33] <juliank> our systemd only ships /etc/sysctl.d/99-sysctl.conf
[22:33] <juliank> Chevek: So you want to add a task for procps package, do one of (1) apt source procps, patch it to add the file, generate a debdiff, and attach it to the bug (2) git ubuntu clone procs, patch it, submit merge request
[22:34] <juliank> Chevek: Or just attach the sysctl.d file and hope someone picks it up; if you attach a patch, well a debdiff (and subscribe ubuntu-sponsors) or open a merge proposal, it lands in the sponsoring queue
[22:35] <juliank> But then I'm not sure this could get sponsored easily
[22:35] <juliank> This needs coordination from like kernel team, desktoop team, and possibly product management
[22:39] <juliank> Chevek: So I subscribed desktop team; added tasks for procps and gamemode - gamemode is supposed to enable specific game modes, so it could just set the sysctl when running games I suppose
[22:41] <doko> qutebrowser, but not so cute dependencies ...
[22:44] <juliank> doko: Seems you inadvertently dropped various deltas in libcanberra
[22:45] <juliank> At the very least pipewire-pulse | pulseaudio became pulseaudio | pipewire-pulse again which it shouldn't be
[22:46] <juliank> This causes germinate to seed wrong
[22:47] <juliank> OK I think the rest is reasonable, but the ubuntu changelog also got lost in the merge
[22:47] <juliank> -02_disable_login_sound.patch
[22:47] <juliank> -04_nodisplay_autostart.patch
[22:47] <juliank> oh these patches are missing now too
[22:48] <doko> hmm, ok, looking
[22:48] <juliank> The last ubuntu upload was 0.30-10ubuntu7 , I think you may have accidentally? pulled the Debian source
[22:49] <juliank> But yeah I think these three parts need to be reinstated or desktop will shout at you :)
[22:50] <juliank> So we'll have a bunch of libcanberra rebuilds to do later then, they should show up on https://magenta.jak-linux.org/ubuntu-archive/distcheck/noble.armhf/global-ben.rebuild-for.txt eventually
[22:51] <juliank> no-change rebuilds would be taken care by my run tomorrow morning fwiw
[22:53]  * juliank just did `rebuild-for "libfastcdr1t64" fastdds` now that the new fastcdr has been built
[22:53] <juliank> (well manually but still counts :D)
[22:56] <doko> yes, Debian. sorry.
[22:57] <juliank> It's good sort of, we needed to merge it, but forgot
[22:57] <juliank> We didn't do t64 transition on it apparently
[22:58] <doko> ´jbichaplease could you have a look at the libcanberra merge? no t64 upload yet ... :-/
[22:58] <juliank> vorlon: We need a KPI for current count of failed tests per architecture
[22:58] <doko> jbicha: ^^^
[22:58] <doko> too tired now
[23:00] <juliank> I'm probably happy to keep misbuilt libcanberra on armhf for now maybe and sort this out after everything else migrated
[23:01] <juliank> I don't expect we have lots of tests exercising it
[23:01] <juliank> Not sure which symbols changed there tbh
[23:02] <juliank> back to KPIs
[23:02] <juliank> Debian has really nice dose-distcheck report in https://qa.debian.org/dose/debcheck/unstable_main/index.html
[23:02] <juliank> including historical view
[23:03] <juliank> they also have number of uninstallable dependencies graphs
[23:03] <vorlon> why does purple-lurch consistently break the test bed
[23:03] <juliank> seems fitting for the name?
[23:04] <vorlon> it's only a no-change rebuild too
[23:04] <juliank> armhf and s390x autopkgtest queues are empty again
[23:04] <juliank> amd64 is at 17k tests now
[23:05] <juliank> i386 at 8k
[23:05] <juliank> arm64 at 3k
[23:05] <doko> shlibs.local, fixing
[23:05] <vorlon> doko: which one?
[23:06] <doko> purple-lurch. uploaded
[23:13] <juliank> I changed my cron job to run global-ben and removal-candidates first before it spends the ~15 mins on dose-distcheck
[23:17] <juliank> So you should get results at like 1-2 minutes past the hour now, as fresh as possible :)
[23:17] <doko> vorlon: purple-lurch ftbfs on armhf, implicit ... I can look at that tomorrow, but not now
[23:22] <Chevek> @juliank: Thank you for pointing me to the right package and for your participation to #2057792. I'll try to patch procps.
[23:25] <juliank> https://magenta.jak-linux.org/ubuntu-archive/distcheck/noble.armhf/global-ben.log has a bit nicer log messages now
[23:25] <juliank> Lots of missing proposed builds; or old proposed builds
[23:25] <juliank> Rebuilding log messages like the following are useful for context:
[23:25] <juliank> # Rebuilding gst-plugins-good1.0 due to gstreamer1.0-plugins-good 1.24.0-1ubuntu4 depending on {'libsoup2.4-1'} - should be {'libsoup-2.4-1'}
[23:26] <juliank> These are issued for every `rebuild-for` command in the `global-ben.rebuild-for.txt` report
[23:26] <juliank> that one seems buggy though
[23:27] <juliank> vorlon: libsoup2.4-1 maps to libsoup2.4-1 (itself) in your https://people.canonical.com/~vorlon/armhf-time_t/ftp-override-map.txt
[23:28] <juliank> as does libsoup-gnome-2.4-1
[23:28] <juliank> Does this mean no change needed?
[23:29] <juliank> oh there is a new dash in there
[23:29] <juliank> sorry!
[23:30] <juliank> I *think* global-ben only logs "missing proposed build" for binaries that exists in armhf release pocket
[23:31] <juliank> God polishing up these tools will be messy but I think they are all useful
[23:32] <juliank> Probably shipping them in like apt-archive-reports in apt-team on salsa because lolz
[23:38] <juliank> oof 00:38
[23:38] <juliank> smell you later, as some would say :)
[23:41] <vorlon> juliank: yes those dashes :)
[23:42] <vorlon> doko: purple-lurch> thanks, removing from release pocket then and filing a bug