[00:06] chandlerc, do you use any fancy compiler flags on Gentoo? [00:26] I'm getting a revision_id from a file w/in a branch that isn't in said branches revision_history()... I'm assuming this is a submerge of some sort, but how should I go about tracing it back to a rev in the branches revision_history? [00:28] jelmer: not really [00:29] jelmer: -O3 -fomit-frame-pointer.. i really only aim for inlining and reducing register pressure [00:30] jelmer: and bzr-svn is compiled with nothing, and lots of debugging, i don't install it via gentoo tools [00:42] chandlerc[g], what about the compile flags for python itself? [00:47] i reproduced it with a hand built, debugging enabled python [00:48] was still using the system installed subversion libs and bindings tho [02:06] I repository A, which has been in use for a long time. Now recently I've created repository B to hold some related work, and I've committed many revisions to it. I would like to import repo B in to a sub-directory of repo A, including all revision history.. is this possible? [02:06] *I have [02:10] sure [02:12] bob2: OK, how? :) [02:13] cd A/somewhere ; for branch in `bzr branches ~/B` ; bzr branch ~/B/$branch $branch ; done [02:13] but that breaks for branches in nested dirs - I think you'd need to use push --create-prefix [02:15] cd ~/b ; for branch in $(bzr branches) ; do (cd $branch ; echo bzr --create-prefix push ~/a/somewhere/$branch) ; done [02:16] bob2: thanks, I'll experiment [02:16] repo-push would be ideal. but I don't think it can push into a subdir of a repository [02:17] Well, I would read the question as asking rather about merge... [02:37] bob2: "bzr branches" just prints a single newline when run on either of my repos [02:38] bob2: and besides.. I don't wish to have a separate branch existing within a subdirectory of another branch... I wish to merge revisions from repo B in to A, except I want the files in repo B to appear inside a subdir of repo A [02:39] I'm going to delete repo B once this is done [02:39] I think we're conflating branches and repositories [02:39] ok, probably so :) [02:41] in bzr, repositories are just places where branches live and share storage. you can not ever use a repository, and it'll only cost you some disk (and maybe bandwidth). [02:42] ok [02:42] are A and B branches (ie contain files you've commited etc)? [02:42] yes [02:43] so yeah, I was using the wrong terminology [02:45] are A and B related at all (ie is one a branch of the other)? [02:45] no [02:45] but I've decided that the content is logically related, and so I've decided to merge the branches [02:46] s/merge/combine/ [02:46] and I have no idea what that means in terms of BZR [02:47] I think the trick would be: (in B) "bzr mkdir subdirwhereyouwantthefilestoappearinA ; bzr mv * subdirwhereyouwantthefilestoappearinA" (in A) "bzr merge -r0..-1 ../b ; bzr commit -m 'merge in B.'" [02:47] er, commiting in B after the mv [02:47] fullermd: right? [02:48] sounds like we're getting there... the thought that I would have to "mirror" the directory structure occured to me [02:48] (you could do the mv after the merge, but it'd be fiddler if any of the names collided) [02:51] Pretty much, yeah. With rich roots, 'bzr join' should be able to automate some of that. Ditto for the 'merge-into' plugin. But it's not that hard to do it manually. [02:51] (don't forget to commit after the mv in B before you merge) [02:53] bob2: OK, well yes what you suggested did work. [02:53] looking at 'bzr log' in A it seems that all the revisions from B are now sub-revisions of the revision I committed to A after merging [02:53] and I suppose when I push to SVN it will only go through as a single commit [02:54] so Trac won't have my actual history... [02:54] is this always how bzr merges work? [02:55] Well, when you try and shove it into a round hole like SVN... ;) === sdboyer is now known as sdboyer|zzz === beaumonta is now known as abeaumont [13:36] how can i rename files? [13:39] dereine: is "bzr mv" what you are after? [13:43] thx i tought mv would be do easy ^^ === fta_ is now known as fta === lamont` is now known as lamont === sdboyer|zzz is now known as sdboyer [23:02] spiv: call? [23:02] lifeless: sure. I'm on skype if you want to use that. [23:03] its trying ... [23:07] fooding [23:32] lifeless, spiv, anyhow today i was going to finish setting up kerguelen to the point it can at least manually build correct packages [23:32] and then do 1.8final and 1.8rc1 [23:33] poolie: morning! hey quick q... I did mv of a version controlled file (without bzr mv), then mistakenly did bzr commit but aborted it, renamed back, then bzr mv but then bzr says it's not under revision control. I think that something does get committed at least as far as the live state is concerned? [23:33] I can run through it in a fake repo to reproduce ... this is 1.7 I think [23:34] if you can write a reproduction script that would be good [23:34] on it. no worries. [23:34] poolie: cool [23:34] when you say "aborted" do you mean you uncommitted, or you pressed ^C during the commit? [23:34] poolie: I'm working on fetching inventories faster [23:35] if the latter, it's possible that you caught it after the workingtree was updated but before the branch was updated [23:35] since it is not precisely atomic across those objects [23:35] lifeless: and doing something on usertest? writing more scripts? [23:35] lifeless: btw i looked yesterday and orcadas was still running the same job [23:35] and we have a hit. ok wraping up the script [23:35] since it runs bzr branch several times [23:38] you can reproduce it? [23:41] poolie: https://bugs.launchpad.net/bzr/+bug/282402 [23:41] Launchpad bug 282402 in bzr "bzr losing track of file in aborted commit of rename" [Undecided,New] [23:41] ye that ;-) [23:41] * arjenAU forgot about the bot [23:41] poolie: no, I mentioned usertest to andrew in the context of 'usertest is still fetching' [23:41] is that meant to be an 'uncommit'? [23:42] nop [23:42] poolie: it only clones the full repo once [23:42] arjenAU: ?? [23:42] # abort the following commit, because we realise we should've done the rename through bzr [23:42] bzr commit [23:42] what does that mean? [23:42] poolie: exit out of the changelog editor without saving, so it does not commit [23:43] oh i see [23:43] except SOMETHING is stored anyway... [23:43] yea. and veyr common possibly, it's so easy for the fingers to type mv and late realise that it was wrong - you realise when you see the changelog in commit, so you back out, ....blahdiblah and there you are, a murdered history [23:44] I haven't tracked what is actually mucked up where, but since you guys will fix this I won't bother ;) [23:44] :) [23:44] arjenAU: so you've done 'mv foo bar' [23:45] * arjenAU hugs all bzr devs - great work everybody, thanks! sooooo easy. kinda feels like bk when I first got to use it years ago [23:45] i answered it [23:45] arjenAU: and "bzr commit -m ''" loses the fact foo existed [23:45] ? [23:45] I bet its the implicit delete code [23:45] yes, it is [23:46] 6ok cool [23:46] never use -m [23:46] -> repository code [23:46] lifeless: i remember you have an opinion about that but i can't remember which way it points :) [23:46] against or in favor of implicit delete? [23:46] poolie: I want auto-add and auto-delete in commit, for symettry [23:46] poolie: I wrote a patch to do auto-add in commit, it was .. contentious [23:47] ok [23:47] oh hmm... well commit doesn't auto-add things, so I don't want it to auto-rm things either [23:47] but yeah, I'm fine with the functionality [23:47] i agree they should be symmetrical but i suspect it should default off [23:47] but it should be either or none [23:47] you know you could make it even more nifty [23:47] poolie: the thread is there on the list :> [23:48] if a file vanishes, grab a hash of the alst revision and see if any of thenew files is the same? [23:48] suggesting a rename! [23:48] won't work if it was also edited, but... [23:48] i think there's a bzr-autorename plugin that does this? [23:48] plugins, plugins... [23:50] poolie: I'm really quite happy with the behaviour of 'bzr rm' now [23:50] poolie: I'm thinking of going back to it and removing some of the extra knobs [23:51] ok so anyway you guys have an angle on what's going on. that's good, then I can move on and see the update soon! [23:52] arjenAU: you do know you can 'bzr revert foo' too, to restore bzr's tracking of it after it was renamed [23:52] lifeless: so it's kinda half committed? [23:52] lifeless: I'm just trying to get my head round what bzr is thinking there [23:52] arjenAU: no; the commit is writing an update to the staging area [23:53] arjenAU: there is a thing called dirstate that lists what is versioned in your current tree [23:53] ye, so autoremoves shouldn't be written there before the commi is confirmed, then eh? [23:53] arjenAU: commit has two major tasks: to add a new revision to the repository, and to update the dirstate to record what was committed so it doesn't show as changed anymore [23:54] arjenAU: and its just unlocking rather than rolling back the change to the dirstate [23:54] gotcha. so it's kinda half committed [23:54] arjenAU: so its precisely equivalent to 'mv foo bar' && 'bzr rm' [23:54] ye ok. so with bzr revert I can undestroy my world. but you're going to fix it anyway ;-) [23:55] arjenAU: to restore the data for 'foo', you can do 'bzr revert foo' (which will restore the dirstate entry) and then 'mv bar foo' [23:55] foo bar. yea [23:55] oh sure, its a bug, but its kinda cosmetic compared to some of the performance and functionality stuff [23:55] lifeless: dude this is serious, I keep tripping over it [23:55] (no, bar foo, to put your copy of bar back) [23:55] Does bzr work with Python 2.6 yet? [23:55] MattCampbell: there is a patch, I don't know if its landed yet [23:56] MattCampbell: it is landed, it should work in 1.8rc1 [23:57] bug reports welcome [23:57] Hi ppl! After all the assistance on Friday to reset the svn hard-wiring in my brain I'm back with another question. One that will make ArjenAU very happy [23:58] I've successfully switched a few svn repos to bzr by doing something like: "bzr branch file:///home/jon/temp/svn/blah blah" (with the bzr-svn plugin installed) [23:58] jonoxer: go for it, grasshopper. your world will be better [23:58] But it runs out of memory about 3 hours and 11,000 commits into converting a 21,000 commit, 4.4G repo [23:58] I get a traceback from bzr [23:59] This is on a machine with 2G RAM, 4G swap, and it doesn't ever fill the swap [23:59] jonoxer: well, whats the traceback [23:59] Is there any recommended way to convert a large repo? [23:59] lifeless: I'll pastebin it, gimme a sec...