[00:08] mtaylor: if you think you can add it do so; if not let me know and I'll add it now [00:09] lifeless: I looked and wasn't sure where it should be added [00:14] mtaylor: there is a loop across branches right ? [00:14] mtaylor: you need to do the check for missing data on each branch [00:16] if its not stacked then it will quietly do noting [00:25] lifeless: so: [00:25] def upgrade_branch(self, branch_url): [00:25] upgrade(branch_url) [00:25] right before the upgrade(branch_url) ? [00:27] not quite [00:27] change it to do two passes over the branches [00:27] in pass one do the fixup [00:27] in pass two do the upgrades [00:34] lifeless: lp:~mordred/autolp/fix-branch [00:50] jkakar: mtaylor: spm: mtaylors branch looks correct to me. [00:52] i have a problem with a related bug, where someone did a --fixes on bzr commit. the problem is he mistyped the bug number. i mailed him, he removed that relation but now it appeared again! [00:53] spm: (I want to disappear for, oh, 4 hours.. so if you want my help on this please look at it now) [00:53] spm: (drizzle can only upgrade on mondays, because they don't want downtime-like-things during the week) [00:54] lifeless: oki [01:03] spm: I can't tell if that is 'its ok, I don't need you' [01:03] or 'ok I'm looking at it now' [01:03] lifeless: heh. sorry - am looking atit now. [01:03] just about to do a trial [01:08] lifeless: so last time we ran it "AUTOLP_DEBUG=1 ./bin/autolp upgrade-branch-format -p landscape-client" which was going to iterate over all ls-clients; in this case, I gather we don't want to run against 250 ish branches in a single hit. can we limit somehow? [01:08] spm: we don't want landscape-client [01:08] no :-) [01:08] spm: we want drizzled. [01:09] spm: why would we want to skip branches? [01:09] yes, but do we want all 250 branches of it? [01:09] that'll take ~ 5 days to run. [01:09] spm: nonsense. [01:09] spm: most should be stacked. [01:10] lifeless: select count(*) from branch where target_suffix = "drizzle" and stacked_on is NULL; ==> 249 unstacked branches [01:10] grah [01:10] thats ... insane [01:10] mtaylor: ^ [01:11] if we can limit to say to do batches of 5 or so?? [01:11] spm: we can actually skip all non stacked branches except for trunk [01:11] hang a sec [01:12] lifeless: I have no idea why they are not stacked [01:12] oh- perhaps those are the branches we pushed from before when stacking existed? [01:13] we probably don't need to upgrade any branches that haven't been pushed to in a month [01:13] in case that helps [01:14] lp:~lifeless/autolp/onlystacked [01:14] spm: ^ [01:14] mtaylor: if you want to hack that in, go right ahead. [01:14] ta [01:15] however, for now, my little patch will just hard skip the unstacked branches [01:15] so we need to do: [01:15] lifeless: ok. [01:15] autolp upgrade yada yada yada [01:15] bzr upgrade lp:drizzle [01:15] (and repeat for any other series branches you have) [01:16] spm: nice catch there. [01:16] lifeless: not really mine - thumper mentioned it earlier. That's his syntax fail query above :-) [01:16] $credit all around [01:17] spm: eh? [01:23] spm: hows it looking? [01:23] oh. sorry - I thought we were waiting on other stuff being done. I'll kick that off now. the target is 'drizzle' yes? [01:24] spm: pull my branch [01:24] spm: then yes. [01:24] yup, done that. oki. lets make some music! [01:26] lifeless: https://pastebin.canonical.com/25412/ ?? [01:32] mtaylor: branch.bzr_identity != branch :P [01:32] spm: fixing [01:32] np [01:33] k, pull. [01:33] lifeless: heh. sorry - I thought I'd fixed that... [01:33] oh well [01:33] lifeless: thanks! [01:33] spm: thanks [01:33] no worries [01:34] this is the 'throw at the admin to test' lifecycle of software development [01:34] aka spmdo (forever immortalised in my mind) [01:35] spm: Did you quotes page it ? [01:35] no, didn't thnik of that tbh [01:36] ahha http://bit.ly/SPmdo [01:38] so, did that make it happier? [01:38] gah. missed the pull comment. sorry, my bad. [01:41] bleh. perm fail; bzr lp-login, tey #2 [01:41] try! [01:42] lifeless: https://pastebin.canonical.com/25414/ hrm. still go away. [01:42] oh haha. that's on chinstrap. argh. [01:44] spm: (so, I'm doing this, then training to the city; oh 40 minutes back ;P). I will be _very prompt_ on issues. [01:44] spm: that branch is broken. [01:44] let me delete it [01:44] oki [01:45] its a lightweight checkout pointing at *my hard disk* [01:45] IzBZRBug [01:45] lifeless: wha? bzr can haz bugs? [01:45] special ones [01:46] spm: k go [01:47] going [01:47] https://code.edge.launchpad.net/~gregorytorres73/drizzle/gregory may make it hiccup too [01:47] thumper: what is bzr_identity for mirrored branches ? [01:48] lifeless: it is the location on LP [01:48] lp:~foo/bar/baz [01:48] lifeless: progress. https://pastebin.canonical.com/25415/ [01:48] hmm, we'll we'll see what error we guess I suppose ;) [01:49] I blame mtaylor for that one. [01:49] spm: fixed, pushing. [01:49] * lifeless throws it at the sysadmin again [01:49] rev9 up [01:50] heh [01:50] while broken: lifeless.fix();spmdo() :) [01:51] hmm. dunnon what your thoughts there are Rob, but that trailing ':)' would seem to be a necessary part of the syntax. [01:51] :P [01:51] or pep8 style [01:51] : { [01:51] sorry [01:51] boom. bzrlib.errors.NotBranchError: Not a branch: "bzr+ssh://bazaar.launchpad.net/~mordred/drizzle/deve/". [01:51] : P [01:52] mtaylor: ping [01:52] mtaylor: please delete that dud branch [01:52] spm: or you can - see https://code.edge.launchpad.net/~mordred/drizzle/deve [01:52] never pushed to, no content. [01:52] I have the ducky [01:53] and go again [01:53] gone [01:55] where's the kaboom, there is meant to be an earth shattering kaboom. [01:56] my favourite bugsbunny/marvin quote. ever. [01:56] goodness, it's still going! [01:56] we'll be home freeish when it starts doing upgrades [01:57] famous last words? :-) [01:57] mtaylor: @ LCA; I vant beer :P [01:57] mtaylor: and you will want to give me beer - 2a is kumtreyah [01:59] geez, what's with the 10 hour wait time on i386 ppa builds [02:00] lifeless: https://pastebin.canonical.com/25416/ [02:02] spm: run it again [02:02] kk [02:03] Huh, that's a bit surprising. [02:03] spiv: ? [02:04] spm: we don't expect lp's server to fail. [02:04] To get a smart server connection dropped like that. [02:04] spm: but it did. [02:04] really? bugger! [02:04] Especially within the data centre, where I presume the network is pretty reliable :) [02:04] well. yes! [02:05] And the smart server itself generally at least sends some sort of error before blowing up, if it does blow up. But that appears to be a straight drop of the connection. [02:05] spm: I've tossed up a slightly quieter version so if it breaks again do a pull [02:05] oki [02:05] As if the remote side segfaulted or something, which is a pretty odd thing to do in response to a BzrDir.open request. [02:06] tbh, the noise (atm) doesn't worry me. it's a clue that things are still happening. [02:06] spm: shurg. :) [02:07] spm: ok, so here is a noisier one. [02:07] rev11 [02:07] anyhow, how is it going ? [02:08] still doing it's thing [02:08] heh. noisier. :-) [02:09] bzrlib.errors.NotBranchError: Not a branch: "bzr+ssh://bazaar.launchpad.net/~mkindahl/drizzle/drizzle-refactoring/". [02:10] empty branch. should we ignore these in future? or manually delete? [02:10] delete. [02:10] they are crap in the system [02:10] if you pull again we'll get some clearer info about progress [02:11] spm: if it's empty, and has no links to bugs or anything, and they are weeks old, then I don't think anyone will miss them. This one ticks all those boxes. [02:11] oki [02:12] I really need a long white haired cat, whose fur to gently stroke when I do these deletions. Just doesn't feel right without the proper props. [02:13] still chugging ? [02:13] yup [02:17] still going... [02:17] \o/ [02:17] running my admin friendly copy? [02:20] argh. bzrlib.errors.NotBranchError: Not a branch: "bzr+ssh://bazaar.launchpad.net/~mkindahl/drizzle/plugin-fix/". [02:20] not yet; but will now. [02:22] old. empty; deleting... [02:26] Just saw launchpad.net/builders for the first time due to the build for my package being 13 hours away. After poking around the page a bit, I was wondering if the build farm is strictly Canonical machines or if there's a way to run a build machine as a virtual machine on another system? (The reason I'm asking is that I have a few high-end servers that are idle about 90% of the time and could help build...) [02:27] darkpixel: at themoment its canonical only as we don't have a trusted path to build on machines we don't control [02:27] a hostile admin could compromise packages easily :(. [02:28] wooo. much nicer noisier version. [02:29] lifeless: That's what I figured. That's a tough challenge though--how to secure a machine that's physically accessible to a malicious admin... [02:30] spm: you ask, I deliver. [02:30] lifeless: :-) [02:31] added a branch count [02:31] spm: pull again if it fails agian [02:31] kk [02:31] spm: how many seconds per branch ? [02:31] * lifeless considers that we're going to be hammering the ssh handshakes ... [02:31] ~~~~hand~~~~ 3-10 ? [02:33] fwiw, I got two syntax errors: print >self.outf, "Missing inventories: %r" % needed <== type of thing. File "./autolp/commands.py", line 101 & File "./autolp/commands.py", line 117. i s/ >/>>/ but am unsure if this is supposed to be doing something special? [02:34] mtaylor: I am lovnig the names of some of these branches. "whitespace-uncrustify" is fave so far :-) [02:36] bleh. Not a branch: "bzr+ssh://bazaar.launchpad.net/~mkindahl/drizzle/testing-overhaul/". [02:38] lifeless: oh bother. that one's ^^ linked to a bug. is old tho. ??? [02:38] mtaylor: ^ opine please [02:39] https://code.edge.launchpad.net/~mkindahl/drizzle/testing-overhaul/ <== easier clicky [02:39] > 1 year old [02:39] delete [02:39] Duhlete. [02:39] huh. I read that as '09. lala me. must be monday. [02:40] rev 13 up [02:40] with less syntax errors [02:40] off again [02:41] how many branches does it say its considering [02:41] Branches: 238 [02:42] so, 2400 seconds to check them all [02:42] once they are all checked, we should probably disable that part of the logic. [02:42] that'd be the max; yeah [02:42] --no-check? [02:43] heh. --are-you-really-sure-you-know-what-youre-doing-no-check [02:45] --no-fix pushed. [02:45] ta [02:45] though, that said, if there are 238 branches, most have to be unstacked. [02:46] as you said there were 249 unstacked branches. [02:46] >< [02:46] we haven't deleted 11 branches. [02:46] no... 3-5 at most [02:46] so, I dunno whats up there... [02:47] meh. let thumper figure it out? ;-) [02:48] ok, rev 15 when it breaks next. [02:49] heh. timing. [02:49] Not a branch: "bzr+ssh://bazaar.launchpad.net/~ro4tub/drizzle/main/" [02:50] old. deleting. [02:51] re-running, no-fix [02:51] spm: nono [02:51] ^C [02:51] spm: we want no-fix *after* we completely fix them all and we start getting into actual upgrades. [02:52] Ahhhh. righto. I misunderstood the logic. kk [02:52] because at that point the fix check is waste. I'm adding something now to make it report rather than die on NotBranchError [02:52] ta [02:54] spm: ok, stop it [02:54] stopped. [02:54] spm: pull rev 16 [02:54] start it [02:55] this should print out 'Not a branch' for things that are not-a-branch [02:55] sweet [02:55] so we don't have to do a no-op check on stuff we've checked already; N^2 time sucks. [02:55] +1 on that [03:03] no fails yet; still running [03:05] couple of 'fixed's just now [03:09] lifeless: bzrlib.errors.InvalidURL: Invalid url supplied to transport: "lp:~gregorytorres73/drizzle/gregory": The remote branch at ~gregorytorres73/drizzle/gregory has no URL specified. [03:12] spm: hmm, I think the script should probably skip remote (as opposed to hosted or mirrored) branches. [03:12] so remote is 'not mirrored at all' ? [03:13] lifeless: that's my understanding, e.g. when LP branches were hosted on devpad but had some metadata tracked on LP [03:13] lifeless: they are supposed to have a URL associated with them, though, so it sounds like maybe this remote branch is just broken anyway? [03:13] spiv: rev 17 [03:13] spm too? :-) [03:14] I suspect so :) [03:14] blarh [03:14] yes [03:15] spiv: I find it most curious that not only the similarities in our nic's, but that we're both devilishly handsome as well. Shrug. Just one of lifes co-incidences I guess. [03:17] Hah [03:26] spm: how is it looking ? [03:26] good! [03:26] two exceptions that were happily passed over [03:27] another iteration if it dies again [03:27] (I mean, I've pushed again :P) [03:28] kk [03:28] spm: how far through it is, do you think ? [03:28] 208 ish. +~ 10 now? [03:29] actually remove a few. some double lines [03:29] but ~ 3/4's I guess [03:32] awesome [03:34] wow. it just started on the upgrades! [03:35] ok [03:35] so we're committed now :) [03:35] at this point --no-fix if we have to start over [03:35] :-) [03:35] kk [03:35] and you'll want my rev 18 in that case, it handles the same cases that the check was dying on before [03:35] how fast is it chugging through? [03:36] kaboom [03:36] ok, thats not that fast :P [03:37] lifeless: https://pastebin.canonical.com/25417/ [03:37] pastify! [03:37] rev 18 pulled [03:37] oh, I see [03:37] fixing [03:37] we need nosmart+ [03:38] btw; I'll need to afk for the school run and lunch in ~ 20mins. [03:38] I think [03:38] ok [03:38] :-) [03:38] spiv: ^ your thoughts? [03:39] spiv: can you backout the upgrade of add-more-cpplint ? [03:39] spm: ^ [03:40] spm: that is, if you look inside the branch, you'll see backup.bzr, and that du -sh on .bzr shows it to be smaller than backup.bzr [03:40] that's rename the backup.bzr dir? [03:40] ok [03:40] mv .bzr .bzr.broken [03:40] mv backup.bzr .bzr [03:40] kk [03:40] then pull my rev 19 and try again [03:40] this time with --no-fix [03:41] what bzr version do you have? [03:42] Bazaar (bzr) 2.0.2 <== chinstrap [03:42] ok [03:42] so I think I know whats going to happen. [03:43] lifeless: no particular thoughts; I thought the case (upgrading when the stacked-on location is already upgraded) was supposed to work. [03:43] we don't test this particular case - its going to upgrade the repo successfully, then try to upgrade the branch and barf [03:43] spiv: its tested [03:43] /srv/bazaar.launchpad.net/push-branches/00/03/e4/a3 fwiw [03:43] spiv: but I think the test uses the same branch object [03:43] format [03:43] revno 19; trying again [03:44] spm: hang on. [03:44] hanging [03:44] spm: may have to abort and fix some bzrlibby stuff [03:44] bleh [03:44] and I'm nearly 3 hours overdue for crepe :) [03:44] I was going to say :-D [03:44] crepê ? [03:46] lifeless: :/ [03:46] bear with me just a little longer. [03:46] spiv: does that make sense? [03:46] lifeless: it sounds plausible, at least. [03:46] spiv: so, I propose we: [03:46] a) back up bzr+ssh://bazaar.launchpad.net/~drizzle-developers/drizzle/development via spmdo [03:47] b) upgrade that branch itself [03:47] c) resume upgrading [03:47] this will work, because when the branch is upgraded the stacked on branch will already be upgraded. [03:47] spiv: Ack/hate ? [03:48] spm: how long is the schrool run ? [03:48] ~ 20-25 mins. but I need lunch at some point too. [03:49] spm: for sure. [03:49] actually, my rev 21. [03:49] * lifeless rings spiv [03:49] pulling [03:49] lifeless: that sounds sane, if it workds [03:50] spiv: "working" is a fair def'n of sanity to my mind ;-) [03:53] spm: ok [03:53] see the plan above? please start on it [03:53] kk [03:53] spm: lifeless' plan sounds like it will work to me [03:54] spm: if things still fall over, I'll see what I can do to fix [03:54] I will stay until the add-more-cpplint branch has been upgraded successfully (run with --no-fix) [03:54] huh. there's already a backup.bzr dir in here [03:55] spm: recent? [03:55] 2009-03-18 [03:56] If you're paranoid I guess you can rename it to backup.bzr.2009-03-18 [03:56] wfm [03:56] (quicker than deleting, too ;) [03:56] drwxr-xr-x 5 codehost codehost 4096 2009-03-18 21:30 backup.bzr.2009-03-18 [03:56] drwxr-xr-x 5 codehost codehost 4096 2009-03-18 22:04 backup.bzr.for2a [03:56] drwxr-xr-x 5 codehost codehost 4096 2009-03-18 22:04 .bzr [03:56] for the record. [03:57] lifeless: " b) upgrade that branch itself" just "bzr upgrade" in this dir ^^ ? [03:59] * spm holds for 30secs more than really needs to go [04:01] spm: yes [04:02] where 'this dir' is the drizzle trunk dir [04:24] coolio; starting now. [04:25] spm: thankks [04:25] fyi. /srv/bazaar.launchpad.net/push-branches/00/00/3c/e5 for the record. and now, lunch. [04:27] spiv: how will you feel if I bail now? [04:32] lifeless: fine, please bail :) [04:36] * lifeless bails [05:10] that upgrade is finished! [05:10] resume step (c) [05:13] spm: cool [05:13] spm: hmm, you may need to trigger a mirror of that branch [05:14] spiv: oki, fwiw I was doing that in push-branches [05:14] spm: right [05:15] spm: yeah, I think you need to manually trigger a mirror, the webapp still thinks it's 1.9 format. [05:15] now where does mwh hide that these days.... [05:15] spm: most clients will use the mirrored version, not the hosted (push-branches) version [05:15] spm: jml probably knows, and I think he's around [05:17] bleh. wrong. https://code.edge.launchpad.net/~vcs-imports/drizzle/development [05:17] spm: (possibly we'll want to trigger mirrors for all the branches once they're upgraded...) [05:17] ~vcs-imports? [05:18] gargh. imports. wtf am I doing. [05:18] damn. I remember this stuff now. it's sql based. [05:19] spm: you don't get a button in the web UI because you're a rubber duck? :( [05:19] I wish [05:19] spm: I'm pretty sure branch owners do! [05:19] or I'm looking in the wrong place? [05:20] * spiv hmms [05:20] I'm here: https://code.edge.launchpad.net/~drizzle-developers/drizzle/development where abouts would it hide? [05:20] Hmm, maybe it only appears for mirrored, not hosted branches :/ [05:21] blurgh [05:21] mwhudson: jml: thumper: ^^ help? [05:21] There's an API, but that might not be any easier than SQL. [05:21] um. no. :-) [05:22] hat [05:22] spm: what's the problem? [05:22] spm: what are you trying to do? [05:22] mwhudson, jml: he's upgraded a hosted branch directly on disk, and now we want to trigger a mirror of that format change [05:23] we've manually upgraded drizzle https://code.edge.launchpad.net/~drizzle-developers/drizzle/development to --2a on crowberry. but the web interface is still showing the wrong format [05:23] what spiv said [05:23] ah, two sides to this [05:23] (and we may want to do this for many more branches shortly) [05:23] 1) get the puller to run to copy the branch to the mirrored area [05:23] you can do this a number of ways [05:24] mwhudson: for future, it'd be good if admins could have a 'request mirror' button in the web ui for all hosted and mirrored branches. [05:24] with spm around update branch set next_mirror_time = CURRENT_TIMESTAMP AT TIME ZONE 'UTC' where unique_name = $foo; [05:24] spiv, I've already written a *patch to bzr* to provide a convenient plugin for this [05:24] is probably easiest [05:24] spiv: ya think? [05:25] spiv, but it was rejected [05:25] however this still won't update the web ui [05:25] spiv, writing code is way too hard. [05:25] jml: dependencies suck :( [05:25] basically you need to push a new revision to make the scanner look at the branch again [05:25] mwhudson: in this case unqiue name will be 'drizzle-developers/drizzle/development' ? [05:25] which totally sucks [05:25] (rockstar is fixing this) [05:25] spm: with a ~ in front [05:25] mwhudson: For us, just making the puller run is enough, I think. [05:25] oki [05:26] spm: ^ just puller is adequate, I think. [05:26] spiv, jml: http://pastebin.ubuntu.com/336320/ [05:26] * rockstar reads backchat to see what he's committed to fixing [05:26] (oh! the suffering!) [05:26] mwhudson: whee :) [05:27] update done [05:27] rockstar: driving the scanner from a Job [05:27] mwhudson: I'd have called the command "frob", personally ;) [05:27] mwhudson, ah yes. I are doing that. [05:27] spiv: spm-not-do? [05:29]