[00:01] hello [00:01] spiv, jam, good morning [00:02] morning [00:03] hi poolie [00:03] hello beuno, lifeless [00:04] https://bugs.edge.launchpad.net/bzr/+bug/256409 [00:04] Launchpad bug 256409 in bzr "inconsistent delta with deleted directories" [Critical,Triaged] [00:04] I'm working on this [00:04] I think its quite serious, I'm trying to understand it enough to create a test and figure out if its just 1.3.1 or if 1.6 is affected too [00:05] so did jelmer ever wake up today? :) [00:06] I haven't seen him [00:06] jelmer won't be around for a bit [00:06] I've been expecting a [00:06] "w00t, bzr-upload and bzr-search are in Debian", but nothing [00:06] lifeless, is he OK? [00:06] beuno: yup all good [00:07] hey poolie, where's the call? [00:09] hi [00:09] i was waiting for you here [00:10] ah, sorry [00:10] I saw the hello [00:10] np [00:10] you usually just call without response :) === edcrypt1 is now known as edcrypt === edcrypt1 is now known as edcrypt [00:44] jam: youhave a reply - and thanks for reading my essay :/ [00:48] jam: I admired your ascii art. [00:51] hardly ascii [00:51] Well, yeah. === mw is now known as mw|out [00:53] back shortly === kiko is now known as kiko-zzz [01:29] hm [01:29] are 'location aliases' documented anywhere? [01:30] are what now? [01:32] things like :this [01:32] (which turned out to be the one i wanted) [02:14] mwhudson: no. I should do that. [02:15] lifeless: I thought bug 256409 was a solved problem! What's up with it? [02:15] Launchpad bug 256409 in bzr "inconsistent delta with deleted directories" [Critical,Triaged] https://launchpad.net/bugs/256409 [02:15] good evening i'm having problems chceking out the ubuntu-docs branch from #lp and getting this error: http://pastebin.ubuntu.com/36701 in regards to a KnitPackRepository error [02:15] abentley: I'm trying to reproduce at the moment [02:15] abentley: I'm leaning towards 'what I found and fixed has just been causing trouble for people' [02:15] abentley: but until I can trigger it with test code in 1.3.1, I can't check it against 1.6/1.7 either [02:16] abentley: its pretty scary regardless of whether its now-fixed or not, because 1.3.1 is widely spread [02:16] abentley: if you want to poke at it too that would be great [02:16] lifeless: Oh, I had this confused with something else. [02:17] lifeless: You remember we had a problem where file kinds weren't being properly updated by TreeTransform.apply()? [02:18] vaguely [02:18] this seems to be a cluster of problems - my note about what we need to do itemises them [02:47] beuno: thanks :) [02:48] emmajane, :) [02:48] beuno: is it not possible to have something on the launchpad page to that effect? [02:49] emmajane, the explanation itself? [02:49] I mean, I'm all about increasing karma points by asking easily answered questions, but... ;) [02:49] yeah [02:49] hahah [02:49] I agree. I haven't done so because vila should be back from his vacation any day now, and we should be able to wrap it up and release a tarball [02:49] that'd be fantastic. [02:50] I know it's going to be discussed at a BoF at the DrupalCon in two weeks. [02:50] so, that and a .deb will supersede the bzr co... [02:50] bzr-upload? [02:50] bzr generally. [02:50] ah, how cool! [02:50] please get us some feedback on whatever comes out of that [02:50] absolutely! [02:50] I'm just looking for the session description. [02:51] http://szeged2008.drupalcon.org/program/sessions/bzr-bazaar-source-revision-control-system [02:52] The workflows page in the documentation is excellent, btw. Not sure who wrote it but my full compliments go to them. [02:53] emmajane, that would be Ian Clatworthy, which I don't see around at the moment [02:53] he's done most of the docs [02:53] The Five Minute intro had a few leaps of logic that I wasn't able to make (I only know CVS and SVN...and basic skills at that) [02:53] emmajane, like? [02:54] This will expose my naivety. ;) [02:54] poolie: see my email on the list, the log performance slowdown is because of the VersionedFiles changes and how it changes how we get at the per-file graph. [02:54] good, that we wat we can learn [02:54] I proposed a "possible" fix. [02:54] jam, i just saw it [02:54] thankyou! [02:54] Though it probably needs some discussion with lifeless [02:54] beuno: the --push felt like it was the equivalent (mirror) of a checkout but in the opposite direction. This is not the case at all. [02:55] * beuno looks [02:55] an extra sentence which read, "This will upload only the .bzr directory." would have helped me immensely. [02:56] I think of a branch as being a set of files that I download to work on in CVS. [02:56] which is an incorrect interpretation of that word, if I've correctly understood the first little bitof the full intro. [02:57] emmajane, is this the part you are reffering to? http://doc.bazaar-vcs.org/latest/en/mini-tutorial/index.html#publishing-your-branch-with-sftp [02:57] hello emmajane [02:57] beuno: that's the part, yes. [02:58] * emmajane waves at poolie [02:58] emmajane, I see, it would make sense to explain that it only uploads the repository [02:59] as a side note, if you push locally, you actually do push the full working tree (the actual files) as well [02:59] but not if you're going over the 'net? [03:00] emmajane, not if it's something remote (sftp/ftp/ssh) [03:00] * emmajane nods [03:00] I think the five minute intro would also benefit from having a link to the workflow page. It seems as though the setups are different depending on what you want to set up (go figure). [03:00] emmajane, are you subscribed to our mailing list? [03:01] not at this point. [03:01] if you'd like to (it's fairly high traffic though), and maybe send in your views so we can discuss them, I can put together a patch to fix it [03:02] I can fix it anyway, it would just be more interesting if you could put it all together, and we can ping-pong with everyone else to see what else we can improve [03:02] I'd be happy to once I've actually figured out where I think the documentation is unclear. :) Sometimes it's just my fault that I'm skimming. :) [03:02] I'm keeping notes on what I'm doing to set up my environment though. [03:02] emmajane, that information is still intersting, because we can change things so they get cought while skimming too :) [03:03] heh [03:07] emmajane, yes if you send a mail with it all in one place that would be good [03:09] * emmajane nods. I'm keeping very careful notes at this piont, but I thought it would be better to get things running the way I want and then go back and evaluate my notes and the original documentation to see what would have really helped. [03:26] I'm getting a strange error when trying to branch from an smb mounted share with bzr.dev: "ERROR: Transport error: [Errno 12] Cannot allocate memory.../filename.rix". Is that a known problem? [03:27] markh: no [03:29] beuno: Actually, I wrote the workflows page, and Ian polished it. [03:29] its a little strange - a branch seemed to fail, but a checkout worked. Then I did "bzr up" on the checkout, annd it said it applied all (1) changes, *then* gave the error. [03:29] I'll look in the log once the test suite finishes [03:29] markh: we have another bug open with SMB shares [03:29] * markh decided he should try to test his patches on Linux ;) [03:30] abentley, my apologies then. emmajane ^ [03:30] And I forget who did the images. [03:30] abentley: The workflows page is /fantastic/ [03:30] emmajane: Glad you like. [03:32] markh: bug 255656 [03:32] Launchpad bug 255656 in bzr ""bzr: ERROR: [Errno 22] Invalid argument" when "bzr pack" is executed manually or when "autopack" is triggered on a repository located on a windows network share" [Undecided,New] https://launchpad.net/bugs/255656 [03:33] markh: my guess is there is a commonality of some sort [03:33] (beyond 'windows sucks') [03:34] :) [03:40] markh: if you could look into that it would be really good [03:56] mornin' all [03:57] evening Verterok [03:59] Hi beuno, g'evening [04:04] * beuno is off to bed [04:33] lifeless: File "/home/skip/src/bazaar/bzr.dev/bzrlib/osutils.py", line 598, in sha_file_by_name [04:33] f = os.open(fname, os.O_RDONLY | O_BINARY) [04:33] OSError: [Errno 12] Cannot allocate memory: '/home/skip/o/src/bazaar/bzr.work.tests.blackbox//bzrlib/bugtracker.py' [04:34] but immediately after that, Python can do a normal open(fname, "rb") on that file [04:34] maybe samba's just running out of file handles. [04:36] interesting [04:36] try a gc.collect() before that call ? [04:36] ok [04:39] seems to work! removing it again to make sure [04:39] so [04:39] I speculate its a HANDLE leak [04:39] we do a lot of readvs [04:39] probably the same one causing test suite failures all over windows :) [04:40] both LockContentions and warning messages about loads of test dirs being left behind [04:40] I can't make it fail again. Did bzr pack the remote branch? [04:41] it failed when it said "Finishing Pack" on the progress bar. Now it completes every time [04:44] hi jonnyde1 [04:55] hi lifeless :) [04:55] just got an email about your new bug comment message [04:56] I will try your suggestions when I'm back at work, today... [04:58] need to have my breakfast now and then go to work... best regards and many thanks for your help :) [05:03] bug 256409 [05:03] Launchpad bug 256409 in bzr "inconsistent delta with deleted directories" [Critical,Triaged] https://launchpad.net/bugs/256409 [05:33] morning [05:35] i'm advocating usage of LP to GNUmed project...how is it possible to report bugs via email? does one configures separate email-list (i did something similar with roundup) or something else? [05:35] https://help.launchpad.net/BugTrackerEmailInterface [05:36] thanks a lot [05:43] i'd like to be able to stick stuff on the end of location aliases [05:44] bzr push :parent/../other [05:48] mwhudson, that would be good [05:49] i guess it's even not that hard [05:49] just a bit more code in AliasDirectory.look_up [06:10] lifeless, i have a clue [06:10] yay! [06:10] alexander graham bells invention time? [06:15] poolie: whats your clue? [06:15] see mail [06:15] (phone) === mark1 is now known as markh [07:16] today is bazaar-1.6 release day? [07:33] gour: I don't think so [07:33] poolie: do you have a release day for 1.6 specified? [07:35] running "bzr upgrade" on a dirstate-tags branch prints a deprecation warning telling you that the branch is in ann old format and that you should run "bzr upgrade" ... [07:35] (that's with 1.6rc1) [07:36] hmm, someone told me the other day that something is gonna to be released today... [07:38] hrm - merge.py's do_merge takes 3 locks, then sets up a finally to release them all. Shouldn't there be a finally per lock? [07:40] jamesh: yeah, it's a known bug that "bzr upgrade" does that (it's not specific to this particular format transition). [07:41] markh: probably that would be more correct, although the risk of the two unlocks for the lock_reads failing is pretty slim. [07:41] I'm thinking the test suite as much as anything... [07:42] markh: so, gc.collect() worked ? [07:42] lifeless: yeah, I replied. After that though I can't repro it by taking it out. Is that because bzr would have packed the remote branch? [07:42] markh: packing occurs on write operations only [07:43] markh: you can check your bzr.log to see if a pack occured [07:43] and gc.collect() fixes failure to remove test directories in some cases, so I'm wondering if cases like that do_merge might be responsible... [07:46] so, we do have a high volume of file operations [07:46] is there some way to tell python to suck less? [07:47] so now I can't repro it again :( [07:48] hi lifeless, sorry, i don't remember what exactly i did back then :/ [07:48] robsta: ah hi [07:49] robsta: so, the branch is a bit of a mess, I've spent all day trying to figure out WTF happened [07:49] I'm probably your number #1 spammer in your INBOX as a result :) [07:49] lifeless: "bzr mv" i did , that's for sure [07:49] what was revno6 about by the way [07:50] you said 'repaired' ... [07:52] lifeless: i did something that broke my local repo, therefore cloned again and copied over changes [07:52] when you say broke [07:52] what do you mean [07:52] oh, a stale lock [07:52] oh [07:52] ok [07:52] 'bzr break-lock' [07:53] oh (: [07:53] we avoid os-locks for our core data - the don't work well on e.g. FTP :) [07:56] lifeless: so can i do anything apart from starting over in a new repo? [07:56] because we don't use os-locks, its possible to get stale locks, and thats what happend; probably you had some external event, or possibly triggered some actual assertion [07:56] robsta: I've written several test cases to try and reproduce and I have failed to date [07:57] robsta: can you attach your ~/.bzr.log* to the bug report [07:59] done [08:00] thanks, looking [08:03] I'm going to study this [08:04] until I can reproduce it its pretty hard to say how-long the piece of string is [08:04] I think yes, copy the whole tree, rm -rf .bzr, init and start over :( [08:04] I'm not happy about this advice [08:05] can i do something to fix the repo on gnome's bzr-playground too, or would that need manual intervention? [08:05] that will probably be fine if you do push --overwrite [08:06] it will have cruft that can be garbage collected but unreferenced data is not fetched [08:06] or propogated [08:06] oh, good [08:07] ok, interestingly you used mkdir cbd [08:07] I'm going to use this log to try to reproduce again [08:08] possibly i used "bzr mv" with wildcards a few times [08:13] Could someone do me a small favor and "bzr upgrade lp:~bzr/bzr-push-and-update/trunk"? It's still using knits. [08:16] beuno: ^ [08:21] Peng_: done. [08:22] jml: Thanks. :) [08:38] is there a simple howto for using bzr's api to controll workdirs? [08:38] theres some stuff on the wiki [08:38] (i need to rewrite pidas bzr integration, currently its a ugly hack with subprocess)) [08:39] robsta: good, news I seem to have reproduction [08:40] lifeless: it's easing my pain a bit to know the suffering was not in vein :P [08:40] lifeless, yay [08:40] robsta: I have duplicated the failure at revision 11 [08:41] robsta: I suspect the others will be dups, reflecting how you used the tool [08:42] lifeless: oh, what's triggering it? [08:42] robsta: I am reducing the size of test at the moment [08:42] robsta: still a lot of variables [08:56] damn [08:56] missed a change in my test, still can't reproduce [08:57] * lifeless tries real-old-fashioned reproduction [08:58] That's a wonderful out-of-context quote :) [09:04] RAOF: heh [09:04] robsta: have you been using the same bzr version the whole time ? [09:05] lifeless: think so, unless ubuntu changed it [09:05] that is, no dist upgrade [09:05] I'm tracking down LockContention on Windows and I don't understand how it is supposed to work. I've instrumented 'do_merge' and can prove that in some cases, self.this_tree._dirstate._filename == self.base_tree._dirstate._filename [09:06] markh: things like revert could do that [09:06] then that method takes a WriteLock on the lhs, and attempts a ReadLock on the rhs - the readlock fails - which seems to be correct. But it works on Linux (and the paths compare true there too) [09:07] markh: yes, oslocks are different on windows and linux [09:07] heh - no wonder then :) Is that by design? [09:08] ie, what do we do about the test failures? [09:08] markh, it's the documented behaviour of locks [09:08] no; see my mail to the list about a simple change to reduce the issues here [09:08] i would say we made a poor choice (in hindsight) to use them [09:08] so we just skip those tests on Windows? [09:09] (on phone to lifeless) [09:09] i think there is a 'known failure' for the fact that they're not exclusive on linux [09:10] i _think_ the general approach is meant to be that we should detect that they're the same object and handle that at a higher level [09:11] i don't have your context here [09:13] many tests on Windows are failing with a LockContention error - as the locks are exclusive on Windows. [09:14] so if Linux had exclusive locks, I guess all those tests would fail there too? [09:15] best I can tell, do_merge, in some cases at least, is taking a read and write lock on the same dirstate (and failing to take the read lock) [09:16] I hit 54 such cases [09:16] s/I/the test suite/ [09:20] heh - which is more than 1/2 the total remaining errors ;) [09:30] hi, I'm just starting to develop using Bazzar. I've been merging back and forwards between two branches. If I merge again I've just merged and committed when there's nothing to be done, it seems to create a "pending merge" in the destination. Am I doing something wrong? [09:32] i commited and pushed a change that contains a bad log entry; is there way to fix the log for that revision? should i uncommit my last revision and push it again? i've search in bazaar-vcs.org docs with no luck, so i'd appreciate some help or advise, whatever you like [09:32] Stumbles: no, its normal behaviour (though not very useful) [09:32] Stumbles: I plan to enahnce the 'nothing to do' check to catch that as well [09:32] alex_muntada: if you really care about it, uncommit and commit and push --overwrite [09:32] lifeless: so I should just `bzr revert` to back out of that? [09:33] Stumbles: yup [09:33] will he need to --forget-pending-merges too? [09:34] lifeless: it's not a typo and it could be misinterpreted, so thanks for the tip :-) [09:34] markh: no [09:36] robsta: so [09:36] * robsta perks up [09:36] robsta: I suspect something is going on that is outside the variables I can meaningfully choose to examine here [09:36] robsta: I'd like to ask you to do a few things for me, if I may [09:37] lifeless: cheers [09:37] Stumbles: happy to help [09:37] robsta: you're using a deb package right? [09:37] robsta: can you do dpkg -l bzr? [09:38] lifeless: yes and dpkg -l bzr [09:38] Desired=Unknown/Install/Remove/Purge/Hold [09:38] | Status=Not/Installed/Config-f/Unpacked/Failed-cfg/Half-inst/t-aWait/T-pend [09:38] |/ Err?=(none)/Hold/Reinst-required/X=both-problems (Status,Err: uppercase=bad) [09:38] ||/ Name Version Description [09:38] +++-=========================-=========================-================================================================== [09:38] ii bzr 1.3.1-1ubuntu0.1 easy to use distributed version control system [09:38] ok cool [09:39] I did grab that packages source and test with it myself to no avail [09:39] robsta: are you using NFS or anything other than a local filesystem (for the branch that glitched) [09:39] no [09:39] have any other branches done this sort of thing to you ? [09:40] no, this is my only bzr project ATM [09:40] ok [09:41] can you, in a terminal somewhere, run 'bzr selftest --noplugins' [09:41] bzr: ERROR: no such option: --noplugins [09:41] --no-plugins [09:41] this will take a bit of time [09:42] but if it fails in an unexpected manner it may help explain things [09:42] also, I'd like you to take a copy of the corrupt branch [09:42] and in that copy: [09:42] rm -rf .bzr/checkout [09:43] bzr branch -r 10 . ../FOO [09:43] cd ../FOO [09:44] bzr mkdir cbd [09:44] bzr mv ../src/css-border.c ../src/css-border.h . [09:44] (oh, cd cbd first :) [09:44] bzr mv ../src/css-color.h . [09:44] bzr mv ../src/css-gtk.c ../src/css-gtk.h . [09:44] bzr mv ../src/css.h . [09:44] bzr commit -m "test commands" [09:45] bzr check [09:45] (and post the output) [09:45] "bzr mv ../src/css-border.c ../src/css-border.h ." fails, the paths are wrong [09:45] yeah cd to cbd first [09:45] I forgot to mention that cause its not in .bzr.log ;) [09:46] lifeless: sorry, you even wrote it above [09:46] didn't read that far [09:46] thats ok :) [09:47] here goes [09:47] bzr check [09:47] checked branch file:///home/rstaudinger/Desktop/Devel/FOO/ format Bazaar Branch Format 6 (bzr 0.15) [09:47] checked repository format [09:47] 11 revisions [09:47] 32 file-ids [09:47] 64 unique file texts [09:47] 179 repeated file texts [09:47] 0 unreferenced text versions [09:49] one final check [09:49] cd .. [09:49] bzr branch -r 10 FOO BAR [09:49] cd BAR [09:49] bzr pull ../FOO [09:50] from "FOO/cbd" i need to "cd .." twice, right ? [09:50] just trying to make sure everything is exactly right [09:50] ah yeah [09:51] "All changes applied successfully. [09:51] Now on revision 11. [09:51] " [09:51] the self test is thru too [09:52] Ran 10755 tests in 511.042s [09:52] OK (known_failures=9) [09:52] 663 tests skipped [09:52] Missing feature 'FTPServer' skipped 76 tests. [09:52] Missing feature 'Internally performed glob expansion' skipped 5 tests. [09:52] Missing feature '_winreg' skipped 2 tests. [09:52] Missing feature 'case-insensitive filesystem' skipped 2 tests. [09:52] tests passed [09:52] ok [09:52] the good news, is that doing what you appear to have done previously, has worked [09:52] and there isn't anything obviously wrong system wide [09:53] what plugins do you have installed (bzr plugins) [09:53] bzr plugins [09:53] launchpad [09:53] Launchpad.net integration plugin for Bazaar. [09:53] svn 0.4.9 [09:53] Support for Subversion branches [09:53] yeah, fine [09:55] so the bad news then is that if this is representative of your system, bzr behaved correctly previously, so something must have caused the confusion subsequently [09:55] now you said you'd had a locking problem [09:55] has that happened more than once? [09:57] no, and it was my fault, closed the terminal while bzr was prompting for commit msg [09:57] heh [09:57] that works in svn [09:57] * robsta ducks [09:57] mmm, wonder what signal bzr got, perhaps it could handle it better [09:57] feel free to file a bug [09:59] not that important [09:59] Use vim, it ignored ctrl-c [09:59] It chastises you and makes you use :q! [10:01] awilkins: closing the terminal is a little harsher [10:01] robsta: still, be a nice bit of polish [10:01] robsta: I'll file it [10:01] done [10:01] awilkins: $EDITOR ignores it too, nano or whatever is default [10:02] now, the icing on the cake [10:02] is that there is only one missing key [10:02] the one for the new directory [10:09] lifeless: out of interest, have you ever regretted writing bzr in python? [10:09] for lack of compiler checks and whatnot [10:10] nope [10:10] compiler checks catch such a small number of genuine bugs [10:11] by which I mean code that would be syntactically correct, pass selftests but a compiler can tell you about [10:11] Hmm ; I find checked languages to be easier going in a good IDE [10:12] If nothing else the IDE usually helps you out more [10:12] lifeless: yeah, that was my next question; figured it might be an advantage being forced to write tests for just about everything [10:13] When I'm hacking bzr I am often stymied by what "interface" a given class is supposed to support (and where in the hierarchy it gets it from) [10:13] Hunting around in text files is the usualy result [10:13] awilkins: static analysis is reall quite orthogonal [10:13] awilkins: consider erlang for instance [10:13] or ocaml [10:14] I'm too tired to think right now ; was up all night installing Windows (a thousand curses to Dell for fiddly BIOSes that only boot USB with legacy support on) [10:15] And another thousand curses to IT beurocracies that insist on Windows-only full-disk encryption tools [10:15] awilkins: heh, ouch [10:15] And a small curse to my wife for giving in to them [10:15] awilkins: just install windows in a linux VM :) [10:15] I was all prepared to reinstall Ubuntu on dm-crypt partitions for her [10:16] robsta: I'll look at possible corruptors tomorrow [10:16] I could have left that running and gone to bed [10:16] robsta: I did find *one* code path that could cause it, but it doesn't look like it could possibly have been involved with the parameters you used [10:17] lifeless: well, thanks for the awesome support [10:17] uhm, i have a merge problem, i try to recouncile two divergeant branches [10:17] it tells me that i must precise a reivsion base, so i tell it which reivisions to apply [10:18] robsta: and I've sent a patch up for that which will be in trunk tomorrow [10:18] the merge occurs succesfully [10:18] then i try to push, but then, it tells me again that the branches diverge [10:18] robsta: this is the first seriously borked repository we've ever had that wasn't quite obviously a NFS FAIL or memory bit-errors or the like [10:18] robsta: so I'm taking it very seriously [10:23] No one for an idea ? :) [10:24] kiorky: are you sure they are diverged and not just completely unrelated? [10:24] lifeless: they are unreleated [10:25] lifeless: why i thought why something like that : i merge from the other branches, what is different [10:25] then i pushed back the changes [10:25] lifeless: thing is that i update my bzr-svn repo, more nicely, and i try to merge back a locally branched thign. [10:25]