[05:51] good morning [08:27] good morning [10:34] hi, I'm creating the logical tag manually when merging amavisd-new, but version contains ':' ... could I use lp1946841/logical/1_2.11.1-5ubuntu1 instead of lp1946841/logical/1:2.11.1-5ubuntu1? [11:26] morning [12:08] good morning! [12:22] hi athos, andreas [12:23] hello miraspace [12:23] hello athos [12:52] sergiodj: hi, morning, around yet? [12:55] this test stared failing, it's like it's not setting the admin password correctly anymore. It used to work: https://pastebin.ubuntu.com/p/BSRMJtDswk/ [13:04] hm, something fishy is going on [14:05] mirespace, you probably already sorted that out by now :) anyway yes, but I think it's more common to use a '%' to replace ':' in those cases [14:07] the advantage of '%' is that it isn't a valid version string character, so it's made clear that the original version string has been modified [14:07] but that's it [14:13] FWIW, it's the standard for escaping Debian version strings into git tag names: https://dep-team.pages.debian.net/deps/dep14/ [14:15] ahasenack: hi [14:16] sergiodj: troubleshooting this weird problem, it's like debconf-set-selections stopped working === Nafallo_ is now known as Nafallo [14:18] ahasenack: does this seem like an openldap problem? because the test was passing normally before [14:18] and no changes have been made to the package since then [14:21] still troubleshooting [14:22] the one thing I spotted so far was that the domain (example.com) had a leading space ( example.com) [14:22] that might have tripped a regexp check [14:22] but also can't say why it became a problem now [14:22] /var/cache/debconf/passwords.dat has no password set, from the user (password{1,2}) or internally generated [14:23] /var/cache/debconf/config.dat has the domain name and org correct [14:24] sergiodj: the set-selection run was fine: https://pastebin.ubuntu.com/p/9zQv43jpnb/ [14:24] why didn't this work: info: Trying to set 'slapd/password1' [password] to 'secret' [14:25] slapcat confirms the example.com domain was used: olcSuffix: dc=example,dc=com [14:26] and *some* password was set, but it's a hash, I can't tell what it is [14:26] of course [14:26] olcRootDN: cn=admin,dc=example,dc=com [14:26] olcRootPW:: e1NTSEF9b1phRGhxWVpzNk1qWGljMm9PNzYzb0RKUlZ6OVArWWI= [14:26] * ahasenack scratches head [14:38] it's lost after the reconfigure: https://pastebin.ubuntu.com/p/m6sWqVvy37/ [14:38] but just the passwords [14:39] the rest (domain, org) are still set [14:51] oh, ok [14:51] # Now that we created the new directory we don't need the passwords in the [14:51] # debconf database anymore. So wipe them. [14:51] wipe_admin_pass [14:52] so that's a, what is it called again, red herring? [14:52] anyway [14:52] problem is not there [14:55] worked after a slapd restart... :( [14:56] that's... unexpected [14:57] * ahasenack has lunch [15:01] * oden heads home [15:02] ahasenack: I'm trying to debug multipath-tools here, but I'm keeping an eye on your progress [15:35] paride: yuo're right :), I got that a while a go ... but thanks a lot for answering! [15:50] sergiodj: I bet it's that debhelper bug with restarts [15:51] where was that email with the list of packages that needed rebuilding [15:51] let me find it [15:52] it's in the bug [15:52] #1959054 [15:52] yeah [15:52] and openldap is in the list [15:53] there you go [15:53] openldap (2.5.11+dfsg-1~exp1ubuntu3) jammy; urgency=medium [15:53] * No-change rebuild to update maintainer scripts, see LP: 1959054 [15:53] Launchpad bug 1959054 in libvirt (Ubuntu Jammy) "debhelper restarts services marked --no-restart-on-upgrade" [Undecided, Triaged] https://launchpad.net/bugs/1959054 [15:53] 1u3 should be fine then [15:53] but it's what I'm using [15:53] so back to square 1 [15:53] or maybe square 1.5 [15:54] ahasenack: sergiodj: as mentioned in standup - those rebuilds were without -v so if you had a merge in proposed the old changelogs will be gone [15:54] no auto-close of your LP bugs anymore :-) [15:55] the changelog mentions the rebuild explicitly [15:55] I have to see what postinst does [15:55] it's the most complicated postinst I have ever seen in a package [15:56] I had to hack on it a lot during last cycle [15:56] I'll try to look into it later [15:57] still working on multipath-tools/postgresql now, so my plate is currently full :( [15:58] it does call this at the very end [15:58] + update-rc.d slapd defaults 19 80 [15:58] + invoke-rc.d --skip-systemd-native slapd start [15:58] but it's running, so that should be working === genii-core is now known as genii [16:23] rbasak, https://git.launchpad.net/~bryce/+git/excuses-kicker/ is the tool I've been using. [16:30] sergiodj: got it [16:30] what's happening, I mean, not the cause yet [16:30] openldap from the release pocket works [16:30] the one from proposed doesn't [16:30] so good test, caught a real problem [16:30] waveform: around? I might need you and your debhelper restart thingy fix [16:30] ahasenack, I'm here [16:31] the fix for the debhelper restart problem *should* be in proposed IIRC [16:31] waveform: sergiodj how it works in the release pocket for openldap: https://pastebin.ubuntu.com/p/FF4JytPQmq/ [16:31] key thing is ldapwhoami working with that password [16:31] and we can see that slapd was restarted after dpkg-reconfigure [16:31] now with proposed: [16:32] https://pastebin.ubuntu.com/p/xgMRrzGX9g/ [16:32] no restart after dpkg-reconfigure, ldapwhoami is not working until I restart slapd myself [16:33] openldap | 2.5.11+dfsg-1~exp1ubuntu1 | jammy | source [16:33] openldap | 2.5.11+dfsg-1~exp1ubuntu3 | jammy-proposed | source [16:33] u3 and u2 are no-change rebuilds [16:35] waveform: what is the binary package that has your fix? debhelper? [16:35] that's correct -- the debhelper 13.6ubuntu1 merge also incorporated the changes to restart behaviour [16:36] Preparing to unpack .../35-debhelper_13.6ubuntu1_all.deb ... [16:36] I'll have a look at the what's been generated in the openldap maintscripts in proposed [16:36] and it was used in the openldap u3 build [16:39] hmmm ... I can see the start line in the generated postinst, which is correct ... let's see the prerm [16:41] waveform: this is my quick diff: https://pastebin.ubuntu.com/p/GmYFjfHQR9/ [16:42] s/restart/start/ ? [16:42] no, that much is correct -- that was the root of the restart issue [16:42] no, I see the issue [16:42] dpkg-reconfigure runs that [16:42] it's only *stopping* the service on removal, rather than unconditionally [16:43] (it should be unconditional in the case of something using --no-restart-after-upgrade which openldap is) [16:43] damn [16:44] first, let me thank you for working on this very thorny issue :) [16:44] it's the kind of thing that would keep me awake at night after I uploaded [16:48] waveform: is that enough to file a bug? [16:49] hang on, I'll dig out the original bug -- it may as well go there as this'll need to be fixed upstream in debian too and that's got the link [16:49] LP: #1959054 [16:49] Launchpad bug 1959054 in libvirt (Ubuntu Jammy) "debhelper restarts services marked --no-restart-on-upgrade" [Undecided, Triaged] https://launchpad.net/bugs/1959054 [16:49] ok, I'll paste the diff there (or attach) [16:50] yeah, the relevant bit is the prerm one [16:50] doubly annoying because I knew openldap was one of the affected packages (not affected by the restart bug in that issue *directly*, but by the fix that caused the restart bug in that issue, and which I had to undo and redo differently to fix that issue) [16:51] thought I'd checked the maintscripts were being generated correctly, but evidently not [16:51] the openldap maintscripts are very complicated [17:02] oh, hang on -- no, the prerm was *meant* to only stop on upgrade; it's the pre*inst* that's meant to handle the stop in the --no-restart-after-upgrade scenario now ... just a mo [17:02] stop on remove even ... bah [17:04] that was the whole crux of the openldap issue (sorry, too many things on fire today -- only just swapping back all the debhelper stuff into my head) [17:04] careful with those juggling knives, they are sharp :) [17:08] this dpkg-reconfigure sequence ... doesn't seem to be running all the maintscripts? Theoretically what *should* be happening with the scripts generated by the new debhelper (for openldap specifically) is: prerm does nothing (previously it stopped the service but it shouldn't anymore); preinst stops the service (if $1 == upgrade, which it presumably should); postinst starts the service again (if $1 == configure which it should do) [17:09] bryyce: no docs :-/ [17:11] waveform: if you want I can add set -x to all of them and run dpkg-reconfigure [17:12] ahasenack, that could be helpful -- I have a sneaking suspicion it's running the postinst (which now just "starts" the service instead of "restart" -- it can't do the latter as that was the crux of the restart-on-no-restart bug), but it's skipping the preinst (which would stop the service) or it's running it with some parameter set which is causing it to skip stopping the service [17:19] ok [17:29] rbasak, yeah sorry, there is `excuses-kicker -h` but probably would help to have some better docs [17:40] waveform: attached [17:41] ahasenack, thanks -- I'll take a look this evening [17:41] cheers [18:13] hi, I need review on the MP for amavisd-new merge https://code.launchpad.net/~mirespace/ubuntu/+source/amavisd-new/+git/amavisd-new/+merge/415808 [18:35] thanks in advance, of course [18:54] mirespace: I'm taking a look [19:18] ahasenack: thanks for the time spent on openldap [19:18] np, an sssd test is failing because of it, and I need sssd to pass due to nfs-utils (it's a rdeps) :) [19:34] ahasenack, okay -- figured it out. To my way of thinking this is a bug in dpkg-reconfigure which has been exposed as a result of the debhelper fix. I'll write up the details in the original bug (and upstream in Debian), and propose a (trivial) patch for dpkg-reconfigure (and then wait and see what breaks as a result of *that* ... urgh, what a rabbit hole!) [19:35] ugh, nasty [19:35] thanks for digging in [19:36] the crux of it, though, is that dpkg-reconfigure runs d/prerm (incorrectly IMO because that's meant to be from the *prior* version of the package, but it's where the "service stop" went historically so that's probably why), skips the d/preinst (again, incorrectly, but historically almost nothing used it and it's only now it's used as the new location for "service stop"), runs d/config (correctly), skips d/postrm (correctly), then runs d/ [19:36] postinst (which runs "service start") [19:37] (so the "correct" fix would be "change out running d/prerm for d/preinst" but that'll almost certainly break a ton of stuff so I'll propose adding d/preinst to the list of things run and leaving d/prerm alone for now) [20:13] I'd like to start jotting down docs on "how to transition php". Where would be the best place to put it? [20:14] for now I'll work on adding it to the UMH, unless anyone has a better idea [20:17] part of me says wiki.ubuntu.com [20:17] but then another part says nope [20:17] heh [20:17] maybe in the future the server guide could grow a development component [20:17] but UMH works [20:18] (and it's markdown) [20:18] biiig plus [20:18] yeah I used the wiki up til now - https://wiki.ubuntu.com/ServerTeam/Transition/Php8.1 [20:18] umh? [20:18] but yeah markdown and being able to edit with $EDITOR would be nice [20:19] if'n y'all figure out a place to publish markdown for developer guides, I'd love to do the same for the security team's processes documentation. [20:19] ohai sbeattie, funny seeing you here.. :) [20:19] umh = https://github.com/canonical/ubuntu-maintainers-handbook [20:22] is there an ubuntu namespace in github? [20:22] https://github.com/ubuntu/ yep [20:23] sbeattie, github's been working well for us. We can RTFM to brand new hires even before their creds are set up :-) [20:26] bryyce: ah yeah, nice. [21:03] ahasenack: thank you for the review :) [21:03] Have you all a nice weekend [21:19] cya mirespace [22:03] ooh, that's some nice documentation (re: umh) -- is it built anywhere? (i.e. rtd-ish?) === JanC is now known as Guest5415 === JanC_ is now known as JanC