[00:10] arges: the state of bug 1401523 is rather odd? comment #8 mentions precsie but the lucid task was change and comment #9 mentions lucid and no task was changed. Do you know what might have happened? [00:10] bug 1401523 in Landscape Client "Update lucid, precise, trusty, utopic to landscape-client 14.12" [Medium,New] https://launchpad.net/bugs/1401523 [00:15] mlankhorst: hi, I've been looking over the last of the LTS stack uploads for 14.04.2, and noticed that xorg-server-lts-utopic is at version 1.16.2.901... when utopic itself is only at 1.16.0. Can you explain the discrepancy? [00:16] RAOF: ^^ or maybe you have some context for this [00:17] Hm. Did I fail to question mlankhorst about that? [00:18] Huh. Seems I didn't include that with my other queries. [00:19] RAOF: ah, do you have outstanding queries? [00:19] RAOF: I just reviewed the mesa one and it looked to me like it checked out, hope I haven't screwed anything up [00:19] A couple of nitpicks. [00:20] nitpicks can presumably be fixed post-accept then [00:20] Stuff like the empty Conflicts: stanzas and such. [00:20] Yeah, can be fixed post-accept. [00:22] I think I may be missing something, though. Maarten was resistant to fixing some clearly-incorrect things that I queried. They might be difficult to entirely script, but seemed like they could be trivial to fix (and quicker to fix than to argue about ☺) before uploading. [00:23] right, AIUI the lts-* package translations are completely automatic at this point [00:24] so if he fixed them by hand this time, would it just regress with the next upload etc [00:24] Well, it'd be something that would need to be done for each upload, yes. [00:25] Or fix the tools. [00:25] Because that's the only road to consistency. [00:25] "It's hard" isn't a good enough excuse to not fix something. [00:25] The ability to easily fix sed is limited :) [00:26] RAOF: Fix sed? :P [00:26] :) [00:26] RAOF: Oh, like empty Conflicts? That'll wash out in dpkg-gencontrol (it won't show up in the binary). [00:26] Fairly sure, anyway. [00:27] That said, line deletion in sed isn't rocket science. [00:27] And I guess won't show up in further debdiffs. [00:27] But an incorrect Homepage: field will. [00:28] An incorrect Homepage would be incorrect in both the source and the destination of the backport, no? [00:28] So, it should be fixed in both or neither. [00:28] Not worth a nitpick for just the backport. [00:28] No, the backport script *changes* the Homepage field so as to be incorrect. [00:28] RAOF: ... [00:28] Why would it even touch it? [00:29] Because it doesn't understand debian/control at all, it just runs sed s/xf86-video-mtrack/xf86-video-mtrack-lts-utopic/g over it? [00:29] RAOF: Oh, is it some overexhuberant "s/$package/lts-$package/" across all of debian/control? [00:29] Yeah. [00:29] That's silly. [00:30] Right. [00:31] For 1-line fields, that's easy to limit scope, but I can see if he wants to hit up descriptions or line-wrapped deps, it gets tougher without actually processing rfc822 properly. [00:31] Which would stop being a job for sed and start being a job for awk, at the leanest, or a perl or python module, at the heaviest. [00:32] '/Homepage/ ! { [...] }' [00:32] fixed [00:32] slangasek: Oh, sure, EXCLUDING fields is easy, I meant limiting inclusion to only certain ones can be tough. [00:32] true [00:33] but if the only bug so far is with Homepage, JFDI and move on [00:33] Yeah. [00:34] slangasek: Though, the Homepage thing is a cosmetic bug, the worrying bug is the unbounded replacement, which could theoretically have other consequences $later. [00:34] RAOF: and xf86-input-wacom-lts-utopic looks fine to me, have I missed something? [00:34] Not that I'd reject an upload because of a tooling bug that might break on a subsequent upload. :P [00:35] slangasek: No, I must have just missed that in my batch clicking. [00:35] infinity: right, well, you could also fix it by fixing the sed to word boundaries, since the package name isn't going to appear on its own in other fields [00:35] RAOF: ok, accepted [00:35] RAOF: did xserver-xorg-video-intel-lts-utopic also get reviewed? [00:36] slangasek: No. [00:36] slangasek: Right, and that fixes the most obvious potential bug, which is "what if your package name is a substring of another package name that happens to also show up in debian/control?" [00:37] q fetch xserver-xorg-video-intel-lts-utopic-antidisestablishmentarianism [00:37] Oh, you have some extra tools not in ubuntu-archive-tools? [00:38] mlankhorst: Fixing RAOF's Homepage complaint, and the above potential bug could be done by implementing slangasek's suggestion to make your packagename substitutions bound on word separators. [00:38] * RAOF was meaning to ask about that; it was somewhat cumbersome to review those things [00:38] RAOF: q is queue. [00:38] yeah [00:38] RAOF: For hysterical raisins, many people have it aliased. [00:38] cjwatson brand raisins [00:39] (I'm sure he'll appreciate that highlight) [00:39] Ah, I'm missing pytz so didn't get the help for that. [00:40] mlankhorst, RAOF: similar issue with xserver-xorg-video-intel; the version in the SRU queue doesn't match the version in utopic (2.99.916 vs. 2.99.914) [00:41] Also, I haven't checked the fallout yet, but were these all being overridden to the right component (main/universe, as appropriate, based on their location in utopic) before accepting? [00:41] infinity: I'm pretty sure I got the overrides right [00:41] RAOF: Kay. I'll be double-checking anyway, but saves a bunch of pain if it turns out to be right. :) [00:42] mesa-lts-utopic is wrong. [00:42] Was that vorlon's fault just now? [00:42] Yes :P [00:43] I wonder if he highlights on vorlon on networks where here's slangasek. [00:46] infinity: what did I miss? oh, the component? [00:46] yeah sorry [00:50] slangasek: I'll fix. [00:50] slangasek: Did you miss any others too? [00:51] infinity: alright, I'm up on the new -45 w/ all my boxen. [00:51] Nothing's crashed yet. [00:51] infinity: xf86-input-wacom-lts-utopic [00:51] infinity: I'm consistent [00:54] slangasek: I'll just override the rest in the queue now. [00:54] ack [00:54] And done. [00:58] slangasek: I thiiiiink 'q' was a Keybuk thing, but not sure [00:58] infinity: he does [00:58] elmo: And so do you? [00:58] would've been back when new-cycle auto-syncs were a multi-day nightmare [00:59] infinity: elmo just highlights on everything, to save time [01:00] Smart. [01:03] cjwatson: Hey, launchpad guy. If a source is accepted in universe, goes dep-wait, and is overriden to main before the dep-wait clears, will the build happen in main or universe? [01:05] cjwatson: I'm pretty sure it does the lookup at dispatch time, and that's what buildd-manager feeds to the slave, but I can never remember if that's fact or wishful thinking. [01:05] wgrant: Since it's more your working hours. ^ [01:07] dispatch time, yes [01:08] As Colin says. [01:08] cjwatson: Also, I'd be inclined to blame pitti, since I think his entire life revolves around single-char aliases. But that might not work, cause 'q' might predate him being an AA. [01:09] The best gift we could give him would be a keyboard with twice as many keys. [01:09] could be [01:09] y [01:09] I think m for madison-lite was my fault though === Malsasa_ is now known as Malsasa === neunon_ is now known as neunon [02:48] bdmurray: i reviewed it for precise/lucid; When I did the precise 'sru-review' I ctrl-c'd the window accidentally and forgot to flip the bugstate after adding the message. === marcusto_ is now known as marcustomlinson_ === marcustomlinson_ is now known as marcustomlinson [05:29] arges: okay, just wanted to make sure some tool wasn't broken === highvolt1ge is now known as highvoltage [06:13] pitti: Howdy. [06:13] LocutusOfBorg1: samba merge> if you want to, please :) Happy to sponsor [06:15] hey Unit193 [06:15] Good morning [06:15] infinity: h! [06:18] xnox: \o/ lazr.restfulclient, thanks! I took the liberty of syncing it from exp [06:29] morning === kickinz1|afk is now known as kickinz1 [07:40] good morning [07:56] hi dear folks [08:29] hey dholbach [08:30] hi ochosi [08:30] may i bug you for a sec? [08:30] sure [08:30] thanks! [08:30] ok, so just quickly, because i'm going away over the weekend... [08:30] since you approved the MR https://code.launchpad.net/~thad-fisch/ubuntu/vivid/xdg-utils/lp1363540/+merge/246820 [08:30] I don't think I'm the best person to make a decision about this one [08:31] i wanted to mention that we plan to SRU this to trusty, so it'd be great to get it into the archive soonish [08:31] ah [08:31] seb128, ^ do you think somebody on the desktop team can review this? [08:31] right, i mean it's not too problematic, all in all it's just a shell script ;) [08:31] well.... [08:31] but sure, if we need somebody else to review it, that's fine and good to know [08:32] dholbach, sure, I was sort of hopping that didrocks would get to it during his sponsoring shift on tuesday, but it seems that didn't happen there [08:33] thanks seb128 (and dholbach)! [08:33] since i'm going to be away for a week after today, i just felt like following up on a few things [08:33] dholbach: yeah, I was quite busy yesterday, but as told early this morning to seb128, I'm going to shift with today (this afternoon, to be precise) [08:34] yoohoo [08:34] didrocks, ok, letting it to you then, thanks ;-) [08:35] yep :) [08:35] didrocks: if there are any questions about it, feel free to bug me! [08:35] at least today i should be aroundish [08:39] ochosi: will do! [08:39] thanks [08:46] infinity: afaict word bounaries don't include hyphens, which are used in package names.. [08:49] meh weird.. dpkg-control looks like it should be smart enough [08:50] hm, description, vcs-git and vcs-browser are unmodified, i should do the same for homepage then [08:50] that was easy to fix.. [08:51] the reason it was renamed was because xf86-input-mtrack was in the mapping file, and I didn't special case homepage. Not because it was the source name. [08:53] something like -lts-utopic-lts-utopic could never have happened on the debian/control file, because dpkg-control parses it [08:56] - elif 'Description' == prop or 'Vcs-Git' == prop or 'Vcs-Browser' == prop: [08:56] + elif 'Description' == prop or 'Vcs-Git' == prop or 'Vcs-Browser' == prop or 'Homepage' == prop: [08:56] print orig_line [09:01] prop.lower() in ("description", "vcs-git", "vcs-browser", "homepage") [09:02] you know it makes sense [09:02] none of the affected packages have that :P [09:02] You can still make the code less repetitive :) [09:03] it's code that runs once every 6 months, cost/effort analysis points to keep it like it is === Guest54456 is now known as Laney [09:21] Hmph, removing /etc/machine-id doesn't seem to work: http://paste.ubuntu.com/9816701/ [09:21] (I'm switching to systemd by doing https://wiki.ubuntu.com/systemd#Installing_systemd without the PPA bit; is that correct?) [09:51] Odd_Bloke: please rather use https://wiki.ubuntu.com/SystemdForUpstartUsers#Switching_init_systems [09:51] Odd_Bloke: don't bother with trying this on anything < 14.10; on 14.10 you can experiment with systemd but there are lots of known failures especially on package upgrades [09:51] Odd_Bloke: it's only really supported on vivid [09:52] pitti: OK, cool. [09:54] huh? [09:54] how did mesa build without llvm 3.5 [09:54] oh right, only on archs without llvm [09:56] Odd_Bloke: I added an outdated warning and a ref to the above page [09:58] pitti: Thanks! [10:08] pitti I feel stupid, but doesn't MoM grab the proposed stuff automatically? [10:08] https://merges.ubuntu.com/s/samba/ [10:08] still listing ubuntu3 [10:08] LocutusOfBorg1: it might be a bit behind; but indeed I don't believe it's looking at -proposed at all [10:08] and I don't see it http://people.canonical.com/~ubuntu-archive/proposed-migration/update_output.txt [10:08] http://people.canonical.com/~ubuntu-archive/proposed-migration/update_excuses.html#samba [10:08] what is blocking the migration? [10:08] ah, held by freeze [10:09] look at excuses first, then output [10:09] ah, yes, the html page is more comphrensive [10:09] so what should I do? just wait for -release, wait for MoM and ping you? [10:12] LocutusOfBorg1: or grab the diff from proposed and apply it to your merge [10:12] http://launchpadlibrarian.net/195481264/samba_2%3A4.1.11%2Bdfsg-1ubuntu3_2%3A4.1.11%2Bdfsg-1ubuntu4.diff.gz [10:13] dear LP developers, I love this feature [10:13] mmm yes, but how? patch -p1 and resolve? [10:13] LocutusOfBorg1: pretty much, yes; there's a chance that the security fix is already applied in Debian, in which case it's only adding the new changelog [10:14] hm, no, it's not [10:14] or maybe it's fixed in 4.1.13 already? [10:14] nope [10:14] seems not fixed [10:14] but I'm trying to apply right now [10:14] at least in source.debian seems not applied [10:15] https://sources.debian.net/src/samba/2:4.1.13%2Bdfsg-4/librpc/idl/security.idl/ [10:15] nope 4.1.13 is from 3 months ago [10:15] the git commit id is from two days ago [10:20] done! [10:20] mail sent, can I forget about samba? [10:20] looking to fix the 6 virtualbox CVEs now [10:48] xnox: jamespage: So it looks like removing /etc/machine-id from the image makes systemd sad because /etc isn't writable when it wants to re-create it. [10:48] Odd_Bloke: you aren't using vivid, right? [10:48] Odd_Bloke: because that got fixed there [10:49] pitti: "that" being /etc/machine-id being missing, or /etc not being writable? [10:49] Odd_Bloke: having a more clever mechanism to create /etc/machine-id, and write it to disk after it becomes writable [10:49] Odd_Bloke: i. e. "first boot" initi [10:51] pitti: I am on vivid, but I'm looking at creating cloud images (which we build from scratch). How was the fix introduced? (We might just have missed it) [10:53] Odd_Bloke: didrocks wrote a systemd-machine-id-setup helper (http://cgit.freedesktop.org/systemd/systemd/tree/src/core/machine-id-setup.c); at boot, systemd creates a tmpfs one while the root fs isn't writable yet, and then later on once it becomes writable that helper writes it to the root partition [10:54] didrocks: ^ BTW, you missed to put a correct copyright there :) [10:55] Odd_Bloke, didrocks: erk, sorry, wrong file: http://cgit.freedesktop.org/systemd/systemd/tree/src/machine-id-commit/machine-id-commit.c [10:55] Odd_Bloke: supposedly the cloud images shouldn't have a machine-id, it should be created on first boot? [10:56] Hey, I want to make some developmental contribution to ubuntu foundation. How should i get started? [10:58] pitti: Yeah, we want to create it on first boot (otherwise all cloud instances will have the same machine-id). [10:58] Odd_Bloke: right, that's pretty much what we do on the desktop install images too [10:58] pitti: So if /etc/machine-id is _empty_, we seem to be fine. [10:58] Odd_Bloke: pitti: if the systemd is read-only at boot, you need still to have the file created, by empty [10:58] so that you can machine-id-setup can mount it [10:58] It's just if it's not present at all that we hit issues. [10:58] didrocks: I thought with that fix we wouldn't need an empty file any more? [10:58] yeah [10:59] pitti: we need to have it empty if the system is read-only [10:59] however, the id is then committed once it's rw [10:59] and you keep the same [10:59] (which wasn't the case before) [10:59] didrocks: oh, I see -- to have a mountpoint [10:59] exactly [11:00] didrocks: interesting. does it make it persistent? [11:01] xnox: yeah, it does [11:01] * xnox is confused why systemd can't remount the fs rw without a machine-id.... [11:01] it's not like firstboot is *that* interesting [11:03] oh, that's recent only been done in december... [11:04] abhinav_: Find something that interests you to hack on? [11:04] xnox: You said something about writing an upstart job to create machine-id if missing; are we going to be supporting both upstart and systemd in vivid? [11:05] And from where to get that list of things to hack on? [11:05] (Or, more generally, is there anywhere I can read a summary of the init system decisions/roadmap?) [11:07] Odd_Bloke: we don't want to support more than one init system in a release if we can help it [11:07] Odd_Bloke: that said, for vivid we might need to support upstart on the phone and systemd on desktop/server/cloud [11:07] Odd_Bloke: it depends on getting the remaining porting done: http://pad.ubuntu.com/systemd-porting-sprint [11:08] Odd_Bloke: https://blueprints.launchpad.net/ubuntu/+spec/core-1411-systemd-migration is the current roadmap [11:11] Odd_Bloke: pitti: imho we need machine-id under upstart for the benefit of dropping dbus-machine-id and making the rest of things work correctly. E.g. doesn't logind under upstart need it? and things like that. [11:11] however I do hope that juju would be complete and we will be able to switch to systemd across the board [11:12] and possibly even the phone as well. [11:14] abhinav_: It's far easier if you go for something that already interests you. === _salem is now known as salem_ [11:20] jpds : and i m doing the same [11:28] Odd_Bloke, do you want to assign the machine-id bug somewhere else other than nova? === kickinz1 is now known as kickinz1|lunch [11:29] jamespage: Just posted a comment; nova should _potentially_ treat an empty machine-id the same way it does a missing one. [11:30] jamespage: https://bugs.launchpad.net/ubuntu/+source/nova/+bug/1413293/comments/3 [11:30] Launchpad bug 1413293 in nova (Ubuntu) "Unable to start instances, empty /etc/machine-id file." [High,Triaged] === barry` is now known as barry [11:45] @pilot in === udevbot changed the topic of #ubuntu-devel to: Archive: open (alpha 2 freeze) | Devel of Ubuntu (not support or app devel) | build failures -> http://qa.ubuntuwire.com/ftbfs/ | #ubuntu for support and discussion of lucid -> utopic | #ubuntu-app-devel for app development on Ubuntu http://wiki.ubuntu.com/UbuntuDevelopment | See #ubuntu-bugs for http://bit.ly/lv8soi | Patch Pilots: didrocks === MacSlow is now known as MacSlow|lunch === kickinz1|lunch is now known as kickinz1 [12:49] dholbach, FYI usually to create USB sticks I run dd [12:50] the startup usb creator tool gives me always dbus errors [12:50] :) [12:50] right [12:50] I like to keep it simple ;) === MacSlow|lunch is now known as MacSlow === Laney changed the topic of #ubuntu-devel to: Archive: open | Devel of Ubuntu (not support or app devel) | build failures -> http://qa.ubuntuwire.com/ftbfs/ | #ubuntu for support and discussion of lucid -> utopic | #ubuntu-app-devel for app development on Ubuntu http://wiki.ubuntu.com/UbuntuDevelopment | See #ubuntu-bugs for http://bit.ly/lv8soi | Patch Pilots: didrocks === smoser_ is now known as smoser [14:45] didrocks: thanks a lot, didrocks! yeah, actually i did file a second MR but since brainwash did his MR on top of mine, both ended up in one... [14:47] ochosi: yw ;) yeah, I saw that seb128 approved it, but didn't merge into the mainline [14:47] indeed :) [14:47] ochosi: thanks to you! uploaded now [14:47] phew, now we can move one step further and go for SRU [14:48] took us ages to realize how that problem of the screensaver timeout being reset happens and then to come up with the fix [14:48] i guess 14.04 users will be happy about this... [14:49] yeah, seems to have been complex to untangle :) [14:49] nice work! [14:49] thanks - and thanks again for the support! [14:49] yw [15:02] pitti, please let me know if you have troubles in the merge [15:02] I'll open a bug instead :) [15:03] LocutusOfBorg1: I don't see it on http://reqorts.qa.ubuntu.com/reports/sponsoring/index.html ? [15:03] nope I sent a mail privately [15:03] LocutusOfBorg1: hm, I didn't get any [15:03] Subject: debdiff samba [15:03] Folder: spam [15:03] LocutusOfBorg1: *blush* sorry ^ [15:04] + + debian/patches/CVE-2014-8143.patch fix CVE-2014-8143. [15:04] Samba 4.0.x before 4.0.24, 4.1.x before 4.1.16, and 4.2.x before 4.2rc4, when an Active Directory Domain Controller (AD DC) is configured, allows remote authenticated users to set the LDB userAccountControl UF_SERVER_TRUST_ACCOUNT bit, and consequently gain privileges, by leveraging delegation of authority for user-account or computer-account creation. (http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2014-8143) [15:04] this is the last changelog line ;) [15:04] just to be sure it isn't the same as yesterday! [15:04] no problem pitti my fault [15:04] I should write more verbose mail, to avoid spam [15:05] BAYES_80 and DNS_FROM_AHBL_RHSBL, that killed them, sorry [15:09] oh, the first second of vivid has been released :) [15:09] pitti: He corrected it in the u-d-a version. :) [15:09] infinity: ah, so the first was on u-release@ only, good [15:10] pitti: Note the subject line said A1 as well. [15:10] yeah, I wondered [15:10] pitti: Friends don't let friends email drunk. [15:10] he doesn't seem to be on IRC, for saying congrats? [15:10] pitti: He's in #-release [15:22] hi, can somebody with a recent ubuntu installation, encrypted root fs and upstart as init system do me a favour: please post the output of '/sbin/status cryptdisks-udev DEVNAME=""'. [15:22] I'm trying to fix a bug in cryptsetup but I don't have a ubuntu+upstart system at hand right now :-/ [15:50] bug 1383794 [15:50] bug 1383794 in cloud-init (Ubuntu Utopic) "[SRU] GCE datasource should use the short hostname" [Undecided,New] https://launchpad.net/bugs/1383794 [15:53] pitti, dholbach do you think one day I'll be ready to apply for becoming an ubuntu developer? [15:53] https://nm.debian.org/public/process/locutusofborg [15:53] the debian process is already started [15:53] sure, you could already start putting together an application [15:53] and ask your sponsors for some comments [15:54] it'll give you an idea how ready you are [15:54] should I become firstly a DD? [15:54] to make the process easier? [15:55] LocutusOfBorg1: do you want me to fix little errors in your merge myself, or let you do another round for the learning experience? [15:55] (looking at samba) [15:56] pitti, feel free to send a mail with the problems ;) [15:57] or tell them here, as you wish === mvo_ is now known as mvo [16:01] LocutusOfBorg1: sent [16:01] thanks [16:01] unfortunately the page doesn't mention DM, just DDs [16:02] so I don't know how to best apply for a Per package upload [16:07] Some Breaks/Replaces/Conflicts help please. I've read https://wiki.debian.org/PackageTransition carefully but not sure about this case. [16:08] mysql-common-5.6 ships only /etc/mysql/conf.d/my5.6.cnf, and apart from comments it's actually empty (but may have user modifications). [16:08] In my next upload, I'm getting rid of this binary package. mysql-common will supply /etc/mysql/conf.d/, and mysql-server-5.6 will supply server-specific configuration, including for 5.6, but not at the previous path. [16:09] Right now, in my testing, everything is fine and mysql-common-5.6 is suggested by apt for autoremoval. [16:09] Is this fine, or should the new mysql-common declare Breaks against mysql-common-5.6 (<< new) to force mysql-common-5.6 to be removed? === doko_ is now known as doko [16:10] Not that this makes much difference, because the old file is a conffile and so will remain until purge, but it might help to indicate that it's no longer necessary. [16:10] Alternatively, maybe mysql-common Breaks+Replaces+Conflicts mysql-common-5.6 (<< new) to actually force the old package to disappear? I don't see this usage in PackageTransition though. [16:10] EOquesiton [16:12] thanks pitti mail sent :) [16:14] rbasak: unless m-common actually ships a /etc/mysql/conf.d/my5.6.cnf there shouldn't be an acutal file conflict, or am I missing something? [16:15] rbasak: what you could do is to add a .maintscript snippet to clean up the obsolete conffile on upgrades (to m-common) [16:15] pitti: yeah there's no actual file conflict. It'd be nice to clean up the old package though. [16:15] rbasak: as it's comment-only it's not like it would actually hurt, but if you want to be clean :) [16:16] rbasak: oh, for that; yes, Conflicts:/Provides: is your friend there [16:16] pitti: I'm not sure this counts as an obsolete conffile though. It just needs a purge of mysql-common-5.6, which the user can choose to do (or not). [16:16] rbasak: Conflicts: for removing it, and Provides: for nudging apt that it's okay to do that instead of holding back the upgrade [16:16] rbasak: yes, true that, so ignore that bit [16:17] rbasak: but Breaks: mysql-common-5.6 (<< new) wouldn't be right as there is no "new" [16:17] rbasak: if you actually want to replace an entire package, then Conflicts: is right (as per policy) [16:17] rbasak: https://www.debian.org/doc/debian-policy/ch-relationships.html#s7.6.2 [16:17] pitti: so Conflicts: mysql-common-5.6 and Provides: mysql-common-5.6? [16:18] rbasak: ah, and I forgot Replaces: [16:18] pitti: OK. So all three, unversioned? [16:19] rbasak: yes, that's the standard approach; Provides: *might* not be necessary, but often it is and it doesn't hurt [16:19] pitti: OK. Thanks! [16:19] (wrt. nudging apt) [16:21] pitti: so I have a similar problem in that the mysql-server package that previously depended on mysql-server-5.5 now depends on mysql-server-5.6, and apt wants to resolve the dist-upgrade by removing mysql-server. [16:22] pitti: is this another case that I can fix by having mysql-server-5.6 provide+conflict+replace mysql-server-5.5? We'll be removing 5.5 this cycle if that's relevant. [16:22] rbasak: if -5.5 goes away, then that sounds right [16:23] rbasak: assuming that -5.6 just takes over an existing database? [16:23] rbasak: but if that happens that means that -5.6 conflicts with -5.5 already? [16:23] * rbasak checks [16:23] Replaces: mysql-server-5.5, virtual-mysql-server [16:23] Provides: virtual-mysql-server [16:24] Breaks: mysql-server-5.5, virtual-mysql-server [16:24] Yeah [16:24] yeah, that doesn't look quite right; unversioned Breaks: are not well defined [16:24] rbasak: so changing that to C/R/P: -5.5 should work better [16:24] pitti: OK. Thanks! [16:24] rbasak: if semantically -5.6 takes over an existing db and just continues working, of course [16:24] Yes, that's correct. [16:24] rbasak: if that's not the case, then the upgrade should just be aborted [16:24] ok, good [16:25] Thank you for your help. I'll try this and test. [16:25] cheers [16:26] rbasak: and yes, the above is exactly the kind of nudging that apt needs :) [16:26] rbasak: it's rather hesitant to uninstall packages; there must be several packages depending on the replacing one until their score becomes higher than the removal penalty [16:27] therefore Provides:, which removes that [16:27] * pitti is sure that mvo can explain this a lot better [16:28] rbasak: OOI, do they actually never change their on-disk format, or do newer versions just convert it on the fly? [16:28] pitti: yeah - the dependency tree is a little altered now. I think that's swinging the balance a little the wrong way. Hopefully C/R/P will fix it. [16:28] IOW, once you ran 5.6, could you go back to 5.5? [16:28] I'm not sure. I'd need to check with upstream. [16:29] I think it'll refuse to let you and fail the maintainer script after a critical debconf prompt. [16:29] Or at least that's what we resolved it should do. [16:33] @pilot out === udevbot changed the topic of #ubuntu-devel to: Archive: open | Devel of Ubuntu (not support or app devel) | build failures -> http://qa.ubuntuwire.com/ftbfs/ | #ubuntu for support and discussion of lucid -> utopic | #ubuntu-app-devel for app development on Ubuntu http://wiki.ubuntu.com/UbuntuDevelopment | See #ubuntu-bugs for http://bit.ly/lv8soi | Patch Pilots: [16:42] LocutusOfBorg1: hm, now you entirely dropped the logrotate changelog entry? [16:43] pitti: I'm curious about the decision not to run package hooks on the phone. Was it because Errors didn't accept all fields or for performance reasons? [16:44] yes, sorry [16:44] my bad [16:45] LocutusOfBorg1: replied [16:46] thanks [16:48] random question: is it possible to configure a system to have something init at boot time *after* some other process? [16:48] s/process/process does/ [16:48] rbasak: It shouldn't provide 5.5, that's a lie. Just C/R. [16:49] rbasak: C/R is the magic that tells apt "prefer this one over the other one". [16:49] rbasak: Which you do with virtual-mysql-server [16:50] pitti, "debian/samba.logrotate: use systemd to restart smbd service." [16:50] do you like it? [16:50] s/a system/an init script for a given package/ [16:50] LocutusOfBorg1: no, not systemd -- "service" :) [16:50] * teward fails at typing apparently. [16:50] + debian/samba.logrotate: use service to restart smbd. [16:51] maybe "service" [16:51] LocutusOfBorg1: I'd keep xnox' original rationale, too: such that it works under both upstart and systemd. [16:51] LocutusOfBorg1: the point of the changelog is not to say *what* changed -- that's obvious from the diff; it should give a rationale why it changed [16:51] pitti: MySQL upstream changes the formats occasionally, but newer versions are smart enough to be able to upgrade formats on the fly during a DB fsck, which happens on first run. [16:51] + + debian/samba-common.config: [16:51] + - Do not change prioritiy to high if dhclient3 is installed. [16:52] e. g. that -- that still doesn't give a rationale, just descibes the patch in English [16:52] infinity: ah, so you couldn't go back anyway, and indeed the packages shoudln't be co-installable or have a Provides: [16:52] pitti: Right. [16:53] pitti: Well, you might be able to downgrade from 5.6 to 5.5, I don't know, but it's not a given that you can. [16:54] jamespage: last week you'd suggested rejecting heat from the utopic -proposed queue but were going to double check IIRC. Should I reject it now? [16:55] damn pitti now I got the problem [16:55] I was looking at the wrong line [16:55] :( [16:55] sorry [16:56] infinity: http://paste.ubuntu.com/9821906/ is what I have right now, and it doesn't work. Test in ppa:mysql-ubuntu/collab. Installing mysql-server from the archive, and then dist-upgrade with the PPA enabled, and apt wants to remove mysql-server. [16:56] + debian/samba.logrotate: use service command to reload (send SIGHUP) the main [16:56] processes such that it works under both upstart and systemd. [16:56] infinity: the only difference I see is that instead of Conflicts, I have Breaks. [16:56] rbasak: Those Breaks don't belong. [16:57] rbasak: Breaks/Replaces means "overwrite some files, yo". Conflicts/Replaces means "I'm better than this package, nya nya". [16:57] rbasak: If your Breaks/Replaces is unversioned, you know you've done something wrong. [16:57] infinity: OK, so I'll switch a bunch of the breaks to conflicts. Thans. [16:59] rbasak: Basically, think of it like this: [16:59] rbasak: Breaks: I don't get along with some versions of this package (should always be versioned) [16:59] rbasak: Conflicts: I don't get along with this package at all, and probably never will, it's a jerk. [16:59] rbasak: Replaces: Overwrite files from the other package. [17:00] rbasak: Conflicts/Replaces: the logical union of the two, ie: overwrite the whole package, prefer me, I'm SPECIAL! [17:00] rbasak: Breaks/Replaces (versioned pair): overwrite files that belong to the package version I don't like, but leave new ones alone. [17:01] infinity: thanks. [17:01] rbasak: And P/C/R is a logical union of C/R and P, in that you claim to basically BE the same package entirely. P/C/R sets are usually found mutually between packages, but sometimes also used to transition between names. [17:01] I more or less got that, but some of the special cases aren't completely obvious to me on the wiki [17:02] rbasak: In your case, mysql-5.6 really isn't mysql-5.5, so a provides would be a lie (on the off chance someone really does need to depend on the exact version), but C/R seems sane, and should hint apt to DTRT. [17:02] rbasak: If it doesn't, yell, cause I'm curious. [17:02]