/srv/irclogs.ubuntu.com/2020/02/18/#ubuntu-devel.txt

rbasakcpaelzer, kanashiro: yeah - figure out a way to break the cycle temporarily. I've been asked to avoid doing that live in the archive. An archive admin will be able to advise.00:33
gbit86__If anyone would look to take a stab at this problem I just posted.. then please do! https://stackoverflow.com/questions/60272229/detect-status-of-an-input-caret-under-any-or-most-linux-des01:08
gbit86__like* to take a stab01:08
krytarikgbit86__: I'm not sure why you still pound the development channel with that.01:14
gbit86__lol.. because I am relentless in wanting to find an answer. Pretty sure it is the last thing I need to solve to button this project up.01:15
gbit86__I was hitting the regular ubuntu channel but was directed to go here.01:16
gbit86__And someone in the ubuntu channel was actually very helpful, pointed me to look at ibus, which did and does work on ubuntu 19.10, but not all ubuntu/debian based systems though.01:17
gbit86__I tried installing ibus after the fact.. but that does not seem to be enough.01:17
rbasakmwhudson: systemd is not present in docker containers> well yes, but that's not soemthing that Debian/Ubuntu developers have ever formally considered a use case that they have to maintain. We may have to change that now, but until now Docker's use has effectively been an unsupported hack from our perspective.01:19
rbasakThat's what I meant by "not buggy"01:19
rbasakI accept that we should do what we can for DOcker users and find a solution here01:19
rbasakBUt to me it's Wishlist, not a bug.01:19
mwhudsonrbasak: if you depend on a feature from a package that you don't depend on and that package that's not essential, it's a bug per debian policy01:20
rbasakThough one that deserves some effort01:20
mwhudsonbut this is splitting hairs01:20
rbasakOnly for Debian01:20
rbasakAnyway, sure. I was just trying to explain what I meant.01:21
rbasakI don't care how we choose to state the problem :)01:21
mwhudsonright, i don't think we actually disagree :)01:21
rbasakmwhudson: nothing will process the tmpfiles fragments when the container starts> any chance of getting Docker upstream to provide a mechanism for that?01:23
rbasakIf we had a common language to tell Docker what it needs, and Docker understood that language, then that might be a clean way of making it all work end-to-end01:24
mwhudsonrbasak: i doubt it, but apparently the fact that /run and /tmp are not tmpfs in containers is sort of intended to make this sort of thing work01:25
rbasakSo we could provide an opt-in mechanism for Docker image builders?01:25
mwhudsonthe opt in mechanism is having postinst create the directories, no?01:26
rbasakMaybe. It doesn't feel generic enough to me.01:26
mwhudsonwell the dockerish answer is "RUN mkdir /run/wibble" in your Dockerfile01:27
rbasakI'm not sure there's a guarantee that in every case it's appropriate for the postinst to create all tmpfiles immediately, even if that's the default behaviour.01:27
rbasakYeah sure, but what I'm suggesting is "RUN mk-tmpfiles.d" rather than the image builder having to know specifics of each package01:27
rbasakOr something even more generic and not specific to systemd01:28
rbasakThen our general recommendation could be "if FROM ubuntu, then do that"01:28
rbasak(or Debian even)01:28
rbasakThat should work today without even needing debhelper mods I think?01:29
rbasak(if the mechanism is implemented, of course)01:30
mwhudsonrbasak: i have to run, but the idea of a solution that requires people writing dockerfiles to remember to do some extra thing to have things happen that happen automatically on installation on normal systems doesn't really appeal01:31
* mwhudson --> big blue (or rather dark grey today) room01:31
rbasakmwhudson: sure, hence my talk of getting Docker upstream to help with that part01:35
rbasakmwhudson: enjoy!01:35
xnoxmwhudson: I thought there was syntax to inject commands into whoever does FROM yourself.01:36
xnoxmwhudson: Something like: ONBUILD echo after package installation your docker file must specify RUN systemd-tmfiles --create01:40
xnoxI do wonder if we should drop the /run/systemd/systemd guard, or like make systemd-tmpfiles allow to fail when not booted. And make debhelper to start generating depends on tmpfiles.d (== 210) depending on which directives where added when.01:42
xnoxAnd we probably should install systemd-tmpfiles in the docker container.01:43
rbasakNice. Sounds like that could make it all magically work?01:46
rbasakWithout any in-distribution changes and only in our official Docker images?01:46
rbasakonly -> the only changes required being in, I mean01:47
LocutusOfBorgblaze[m], not sure, better ask tsimonq207:33
LocutusOfBorgtsimonq2, http://debomatic-amd64.debian.net/distribution#unstable/qtcreator/4.11.0-2.1/buildlog07:36
mwhudsonxnox: turns out systemd-tmpfiles is linked to libsystem-shared.so which is linked to THE ENTIRE UNIVERSE but that's a solvable problem i guess07:48
mwhudson/usr/bin/ld: cannot find -lpthreads07:55
mwhudsonhow did that happen07:55
dokoseb128: the libxslt sync from experimental drops the xslt-config binary, causing a lot of ftbfs08:36
seb128doko, I guess those need to be fixed then08:36
dokoor you re-add that binary08:37
seb128or that08:38
seb128feel free to open a bug about that, might be easier to discuss the options08:39
dokoI don't want to discuss, I want to have it fixed ;p, just seeing the new ftbfs working on icu08:40
Skuggenahasenack: I think we've seen floating accuracy test failures before (but IIRC it was because of different maths library on PPC), yeah08:55
SkuggenLet me see if I can dig out some of those tests in MySQL itself, and I can see if they've changed recently08:56
Skuggenmdeslaur dropped that patch from mysql-8.0 when importing 8.0.1808:58
SkuggenActually, that looks like it might have been an urelated change making the same edits to our upstream tests09:06
Skuggen8.0.18 does have a floating point precision-related fix, though09:09
=== Wryhder is now known as Lucas_Gray
ahasenackrbasak: icinga2, the one with the boost fix, failed to build on s390x only12:12
rbasak:(12:13
ahasenackxnox: have you seen these before?12:13
ahasenack /usr/bin/ld: /usr/lib/s390x-linux-gnu/libboost_coroutine.so.1.71.0: undefined reference to `jump_fcontext'12:13
ahasenack /usr/bin/ld: /usr/lib/s390x-linux-gnu/libboost_coroutine.so.1.71.0: undefined reference to `make_fcontext'12:14
ahasenackrbasak: the mysql8 tests seem to take longer than the allotted timeout on arm*, do you remember how this was handled in the past? There is no existing hint for mysql-812:16
ahasenackhttp://autopkgtest.ubuntu.com/packages/m/mysql-8.0/focal/arm6412:17
ahasenackor is it hanging in that test, hmm12:17
rbasakI don't recall seeing it time out12:18
ahasenackthe test runner times out, I mean12:18
ahasenacktinoco had to retry once in the previous upload12:19
ahasenackand it passed after 4h23min12:20
ahasenackwell, better start early12:20
rafaeldtinocoyep i retried that one12:20
rafaeldtinocowe should add "quilt poop" command12:20
rafaeldtinocowhen we are merging and we do wrong things12:20
ahasenackarm64 takes twice as long as amd6412:22
rafaeldtinocoyep, that is what christian has also found12:24
rafaeldtinocowhen building qemu12:24
rafaeldtinocoat least for our build servers12:24
ahasenackwow, there are libreoffice dep8 runs taking 7h+ on arm6412:26
rbasakLocutusOfBorg: fancy merging lighttpd please? It's only on my list because of a no change rebuild :-/12:26
* rbasak doesn't understand why it must have such an extensive delta12:26
xnoxahasenack:  so, coroutine is new in 1.71 on s390x. and I have seen a couple of places like that. I somehow think, I have miss-cherrypicked boost patches for boost on s390x hence it doesn't work for neither ceph nor icinga12:31
ahasenackxnox: ceph is also failing to build on s390x with a similar error?12:31
xnoxahasenack:  yes, when it tried to use coroutine; it doesn't use coroutine on s390x anymore.12:35
ahasenackxnox: is this something you can fix for icinga2/boost?12:35
xnoxyes, it's on my hit list for icu & boost transitions => is it blocking you somewhere else too?12:36
xnoxcause then i can look into it earlier.12:36
ahasenackI'm using https://bugs.launchpad.net/bugs/1863371 for icinga212:36
ubottuLaunchpad bug 1863371 in icinga2 (Ubuntu) "FTBFS with boost 1.71" [High,In progress]12:36
ahasenackxnox: yes, it's the only build failure in my upload12:36
xnoxack12:37
ahasenackthanks!12:37
ahasenackrbasak: those linux-* packages I blacklisted yesterday are blocking the importer runs from importing anything else12:41
ahasenackrbasak: should we build a new snap with that config and deploy?12:41
=== Wryhder is now known as Lucas_Gray
rbasakahasenack: if you kill and restart, it shouldn't reattempt unless another upload to that package is made12:41
ahasenackah, I'll do that then12:42
rbasakBut yes, happy for us to do another release if necessary12:42
ahasenackI never find the snap service name12:42
rbasakThere's also a command line option to override the blacklist locally, but I'd prefer to avoid that if possible to avoid it getting out of sync12:42
rbasaksystemctl status --user12:42
ahasenackah, it's a user service12:42
Mirvcan someone point me to (if exists) some explanation why daily-live 'current' points to 4 week old focal image? it seems a bit weird starting point for testing.12:46
ahasenackMirv: the current pending image is probably failing some test which is preventing it from being promoted to "daily"12:47
ahasenackit does look troublesome, though, if it has been failing checks for 4 weeks12:48
Mirvright, mainly the 4 weeks sounds a bit troublesome given release in 2 months12:48
xnox- default architecture     : none aha12:50
ahasenackxnox: what was that?12:51
xnoxprobably incomplete boost port to s390x!12:52
ahasenackyay12:52
LocutusOfBorgrbasak, .12:56
xnoxahasenack:  uploaded new boost1.71 into the archive, once that builds & publishes, it might be better on s390x, and then just a retry on icinga might fix it13:08
xnoxif that new boost builds13:08
ahasenackxnox: nice, thanks a lot, will watch it13:09
ahasenackxnox: 6ubuntu3?13:09
ahasenackyep13:09
rbasakLocutusOfBorg: thank you!13:20
ahasenackrbasak: restart didn't work as expected, it's trying again (linux-gke, linux-kvm, linux-gcp)13:22
ahasenackI'll ask bryce to do a new edge release and have that deployed13:23
mdeslauris anyone working on a samba merge? ahasenack?13:30
ahasenackmdeslaur: yes, I have one up for review for 4.11.513:30
mdeslaurahasenack: awsome, thanks13:31
ahasenackI pinged debian for 4.11.6, even proposed a pristine-tar branch to them so we at least get the same tarball13:31
ahasenackmdeslaur: https://code.launchpad.net/~ahasenack/ubuntu/+source/samba/+git/samba/+merge/37935013:31
ahasenackhttps://salsa.debian.org/samba-team/samba/merge_requests/45 for 4.11.613:31
rbasakahasenack: I'm gong to attempt a second restart13:35
ahasenackrbasak: why wouldn't it try again if the previous import didn't finish?13:36
rbasakIt gets marked as errored13:37
rbasakAnd deliberately isn't retried13:37
rbasakUntil the next time a publication is made13:37
rbasakRestarted13:37
rbasakThere were a bunch of uploads to different releases of those packages today13:38
rbasak(not -gke, I'm confused why you saw that, but the others)13:38
rbasakSo it's possible that the future uploads kicked off new import runs even though the previous ones were errored13:38
rbasakHence my second restart13:38
rbasakWe'll see13:38
xnoxahasenack:  so the new boost build looks a lot better; now just need to wait for it to publish, and then retry icinga13:43
ahasenackxnox: ack13:43
ahasenackppc64el hasn't even started yet13:43
rbasakahasenack: restarted the importer and it doesn't appear to be stuck on any of the kernel packages yet13:45
rbasakAnd none of them are marked 'needed' so I think we're good.13:45
rbasakMaybe we need a different mechanism13:46
rbasakJust blacklisting packages doesn't seem good enough if new oversized source packages keep being added like this13:46
ahasenackyep13:51
rbasakSo what's the story with the /usr/bin/python name?14:01
rbasakFrom doko's email we shouldn't be using it14:01
rbasakBut how do I explain this to upstreams? Is this in preparation for a future pointer to Python 3, or are we doing that immediately in Focal?14:01
ahasenackIIRC there would be a default python late in the cycle, pointing at py314:03
ahasenackfocal still14:04
ahasenackbut doko would know for sure14:04
dokorbasak: we will have two packages, one linking to 2, one to 3, but none must be used in the archive14:08
dokoI have to follow-up again on this14:09
rbasakdoko: what can users expect to see?14:22
rbasakBecause if we have something by default, they'll inevitably start using that in upstream shebangs, so we'll have be be permanently patching14:22
dokopython-is-python2-but-deprecated and python-is-python314:23
rbasakI can point upstreams to https://www.python.org/dev/peps/pep-0394/#for-python-script-publishers I guess14:25
rbasakahasenack: do you have a current temlate for your regular typedef patch please?14:34
ahasenackrbasak: just the dep3 headers14:35
ahasenackrbasak: I grep for my_bool, then check if the same file includes mysql.h, and add the typedef beneath that14:35
ahasenackif mysql.h is not included there, I grep to see which other file includes it, and change that one14:35
rbasakI was just after your standard dep3 headers and changelog text :)14:35
ahasenackjust a sec14:36
ahasenackrbasak: https://pastebin.ubuntu.com/p/7XqPzbGSGj/ with a helper hack/script14:37
rbasakThank you!14:37
rbasakI suppose I should replace the author?14:38
ahasenacksure14:41
rbasakahasenack: uploads ready for apr and apr-util: https://paste.ubuntu.com/p/SxNYxRTPbD/ and https://paste.ubuntu.com/p/NYphZ9yshg/14:51
ahasenackrbasak: oh, I had those ready already, but cool :)14:51
rbasakOh14:51
rbasakI did file an upstream bug14:51
ahasenackin my ppa, I did a test build yesterday14:51
rbasakAnd for apr14:51
rbasak(in dep3 also)14:51
rbasakHow do you want to proceed?14:52
ahasenackyou go ahead14:52
ahasenackjust drop the ~dev114:52
ahasenackrbasak: I uploaded a few packages with the mybool patch, but I really would prefer for mysql8 to mgirate first14:52
rbasakTo be clear, to upload to the archive now, with these latest pastebins, dropping the ~dev1?14:52
ahasenackbut that will take a while, given the libreoffice tests take hours and hours14:52
ahasenackrbasak: yes, let me just check the second one quickly14:52
ahasenackrbasak: yeah, +1 to both14:53
rbasakThanks!14:53
ahasenackwith the !~dev1 drop14:53
rbasakYep14:53
rbasakapr uploaded14:55
rbasakapr-util will have to wait until apr is built and published, or else it will FTBFS14:55
ahasenackyes14:55
ahasenackbeen there, done that :)14:56
ahasenackboost is built14:56
* ahasenack checks if it's published14:56
ahasenacklooks like it14:56
* ahasenack rebuilds icinga214:57
ahasenackand that was done already14:57
* ahasenack watches it14:57
rbasakI'll look at zoneminder next shall I?14:59
xnoxahasenack:  icinga has now built15:08
xnox(there is buildlog, meaning that it is uploading now)15:09
coreycbinfinity: do you know if 20.10 final release would be oct 15 or oct 22?15:48
ahasenackxnox: awesome, thanks again16:00
knoctecan a maintainer help here? https://bugs.launchpad.net/ubuntu/+source/mono/+bug/1520033 we need to drop an arch (s390x) so that mono can get a hotfix that will make nuget work16:49
ubottuLaunchpad bug 1520033 in mono (Ubuntu) "ftbfs on s390x" [Undecided,Confirmed]16:49
knocteotherwise nuget (the .NET package manager) will crash when running with the mono version shipped by the upcoming LTS16:49
knocteI understand that s390x is not a critical arch for ubuntu anyway16:50
xnoxknocte:  s390x is an official ubuntu architecture with support on-par with ppc64el amd64 arm6416:52
xnoxknocte:  mono itself is in universe, and not supported as much as other packages in ubuntu are.16:52
xnoxknocte:  however current mono in ubuntu does not ftbfs.16:52
xnoxknocte:  so what is your question ? about which package / version and in which release?16:53
xnoxupdated that bug report to be marked as fix released16:54
knoctefix released? what?16:55
xnoxknocte:  Click https://launchpad.net/ubuntu/+source/mono/5.18.0.240+dfsg-3 look at the builds portion of the page: note how it is built correctly on all 6 architectures including s390x16:56
xnoxis there soemthing wrong with https://launchpad.net/ubuntu/+source/mono/5.18.0.240+dfsg-3/+build/17185141 ?16:56
xnoxi see there is -5 version in unstable, not yet in ubuntu.16:57
xnoxbug report and its status was out of date, w.r.t. reality. Do you have any other concerns around mono or nuget in Ubuntu?16:58
knoctewe need 7ac7d35 in ubuntu, otherwise upcoming LTS will crash with nuget16:58
knocte(that git commit I mentioned above is taken from https://code.launchpad.net/~usd-import-team/ubuntu/+source/mono/+git/mono/+ref/ubuntu/eoan-proposed )16:58
xnoxknocte:  i see16:59
knocte(it comes from upstream's "* [d8d5d6b] Fixes __MonoCS__ handling of value types")16:59
xnoxknocte:  indeed something odd is happening16:59
Laney-5 got removed from eoan-proposed because of s390x build failures17:00
Laneyyou can see that on https://launchpad.net/ubuntu/+source/mono/+publishinghistory17:00
xnoxvorlon:  infinity: looking at https://launchpad.net/ubuntu/+source/mono/+publishinghistory it seems like at archive opening we have dropped -5 upload from eoan-proposed, but did not publish it into focal-proposed17:00
xnoxi would have expected it to roll over17:00
* xnox syncs it again17:00
knoctewell, both mono-packaging maintainers have agreed that the -5 fix is more important than the s390x arch17:00
xnoxknocte:  sure, i understand the issue now. It seems that we have dropped -4 and -5 uploads of mono on the floor.17:01
xnoxwhich is a process / archive administration issue.17:02
xnoxknocte:  i'm now trying to get up to date mono upload into ubuntu. thank you for raising this.17:02
knoctethanks for your help xnox17:02
xnoxLaney:  actually i guess i need to do a binary copy from eoan-proposed, or a no change rebuild.17:03
vorlonxnox: -5 was removed from eoan-proposed manually by infinity as 'FTBFS'17:03
Laneyif it does fail to build on s390x then you'll have problems @ proposed-migration17:03
xnoxright17:03
* xnox uploads a no change rebuild to see what happens17:03
xnoxwe have now raised to z13 which might do better17:03
dokocoreycb, jamespage: please could you add a lasso merge on openstack's list? needed for python317:53
coreycbdoko: yes will do17:53
dokota17:54
rbasakahasenack: here's the fix for zoneminder: https://pastebin.ubuntu.com/p/GcRKP5Yyt5/18:08
rbasakI've sent a patch to Debian BTS, just waiting on the bug number18:09
rbasakHow do you want to integrate it? Is the paste sufficient?18:09
rbasakProbably worth testing that in your PPA18:09
rbasakhttps://bugs.debian.org/cgi-bin/bugreport.cgi?bug=95160018:09
ubottuDebian bug 951600 in src:zoneminder ""debian/rules build" fails to call defined build-indep target, causing FTBFS" [Serious,Open]18:09
ahasenackrbasak: you rock18:13
ahasenackrbasak: the paste is sufficient18:14
ahasenackrbasak: I'll upload to my ppa18:14
rbasakThanks!18:18
rbasakI do like the interesting fixes :)18:18
dokorbasak: sent the python2 removal update18:35
rbasakThank you!18:38
ahasenackrbasak: apr has built, will you upload apr-util now?18:47
rbasakahasenack: sure18:50
rbasakSorry I'd forgotten about that18:51
ahasenacknp, thanks!18:51
rbasakDone18:51
seb128doko, does the patch on https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=951101 looks fine? your cracklib2 upload failed to build...19:36
ubottuDebian bug 951101 in src:cracklib2 "cracklib2 FTBFS: debian/rules:6: /usr/share/python/python.mk: No such file or directory" [Serious,Open]19:36
seb128| debian/rules:6: /usr/share/python/python.mk: No such file or directory19:36
=== jdstrand_ is now known as jdstrand
=== ben_r_ is now known as ben_r
mwhudsonmorning20:37
=== gbit86__ is now known as gbit86

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