/srv/irclogs.ubuntu.com/2009/05/04/#bzr.txt

mwhudsoni'm thinking the linux kernel is probably a little ambitious for the moment...00:00
RAOFmwhudson: Banshee would work.  That's ~20Mb or so of git.  Or is that not medium enough?00:05
mwhudsonRAOF: let's try it00:05
mwhudsonRAOF: url me up?00:05
RAOFAlso has the advantage that I know bzr-git successfully branches it.00:05
RAOFgit://git.gnome.org/banshee00:05
mwhudsonoh right, all the gnome imports00:06
RAOFThere's _lots_ of them available :).  You don't want one of them?00:06
RAOFlibdrm would be another one.00:06
mwhudsonRAOF: no, i'd just forgotten00:07
=== UdontKnow is now known as \0
lifelessmwhudson: linux should work well with dev600:09
mwhudsonlifeless: until --default-rich-root is dev6 or similar...00:09
mwhudson(actually i think it will import into 1.9-rich-root now)00:09
=== \0 is now known as root
=== root is now known as KnightWhoSaysNi
lifelessspiv: I'm a little worn on repeated 'cut a round trip off' day days00:15
lifelessspiv: I had wanted to pair today, but my desktop blew a sprocket on saturday00:15
lifelessspiv: so perhaps tomorrow?00:15
spivlifeless: sure, tomorrow sounds good to me.00:16
jonnydee1hi :) I recently stumbled over the following bug and reported it: https://bugs.launchpad.net/bzr/+bug/370551  Is there a chance that a fix will make it in a 1.14 maintenance release, or in 1.15, at least?00:16
ubottuLaunchpad bug 370551 in bzr ""bzr mv --auto" provokes traceback when auto detecting files that were moved to unversioned directories" [Medium,Confirmed]00:16
jonnydee1bazaar cannot handle files that were moved to unversioned directories. but this makes the auto-detect feature more or less useless (for me, at least)00:17
lifelessjonnydee1: thats a dupe actually00:18
lifelessjonnydee1: no, 1.14 won't get a fix; its not a regression. You can just 'bzr add' the new directory before you move files into it, to avoid the bug.00:18
lifelessbug 36762800:20
ubottuLaunchpad bug 367628 in bzr "bzr mv <stuff> unversioneddir/ fails" [High,New] https://launchpad.net/bugs/36762800:20
lifelessthats the root cause00:20
jonnydee1lifeless: I know I can add the directory. But I need auto-detection when I refactor code using an ide. and there it happens often enough that files are moved to newly created directories....00:20
lifelessjonnydee1: in your example you manually move a file to the unversioned dir first00:20
jonnydee1lifeless: and how do I recover from the stack trace if I occasionally got trapped by this bug? Will it suffice to 'bzr add' the unversioned directory, afterwards?00:27
lifelesshttps://bugs.edge.launchpad.net/bzr/+bug/367628 is the main bug for 'bzr mv foo unversioned/'00:27
ubottuLaunchpad bug 367628 in bzr "bzr mv <stuff> unversioneddir/ fails" [High,New]00:27
jonnydee1The problem is that 'bzr st' does not work anymore, which makes it hard to figure out what directories are needed to be versioned...00:27
=== abentley2 is now known as abentley
jonnydee1lifeless: ok, should I mark the other bug report as a duplicate?00:28
lifelessno00:28
lifelessit may be slightly different00:28
lifelessI'm going to note that in the bugs00:28
lifelessmy browser just stopped working for a minute, doing that now00:29
jonnydee1ok, thank you, lifeless, for your help :)00:29
lifelessno problem00:31
lifelesspoolie: http://igotgenes.blogspot.com/2009/04/how-symbolic-on-removing-symlinks-in.html00:34
pooliehm that's good i guess00:35
pooliewbn if somebody fixed the bugs though :)00:36
pygilifeless, poolie : will be nice to see you at UDS :)00:36
poolieyou too00:39
=== spm_ is now known as spm
=== abentley1 is now known as abentley
=== abentley1 is now known as abentley
mwhudsonspiv: wow, bug 250451 gets nastier and nastier01:54
ubottuLaunchpad bug 250451 in bzr "bzr suggests wrong URL for break-lock with a LP hosted branch" [High,Confirmed] https://launchpad.net/bugs/25045101:54
spivmwhudson: yeah01:54
spivmwhudson: I suppose LP could monkeypatch sys.stdout to autocorrect that message ;)01:54
mwhudsonspiv: we _should_ translate paths in some exceptions01:55
mwhudsonbut i don't think that will help here01:55
rockyugh, how many times do i merge stuff from another branch, forget to commit, do a bunch of work, and realize now i have my merge changes all messed up with my new work02:09
rockyis there an easy way to fix that ? ^^02:09
thumpershelve?02:12
rockyi've done a bunch of work, shelve will take me all night :(02:12
mwhudsonshelve, redo the merge, commit, unshelve02:13
mwhudson?02:13
rockyoh you mean shelve everything ?02:13
AfCWhat's with the alternating '>' and '<' in the pretty display when communicating with a remote server?02:18
spivAfC: traffic direction02:18
mwhudsonrocky: right02:20
mwhudsonrocky: you'll might have to fix some conflicts when you unmerge, but hopefully not too bad02:20
mwhudsons/unmerge/unshelve/02:20
AfCspiv: oh02:22
AfCIs there a way to turn off the progress bar and just have the textual information?02:22
rockymwhudson: wow, conflicts are huge (like whole files)02:23
spivAfC: not that I know of.02:26
mwhudsonrocky: :(02:27
rockyactually it wasn't so bad02:27
rockyit was just one file like that and the fix was easy enough02:27
rockyso yeah i'm all set now. .. thanks mwhudson! :)02:27
AfCspiv: I recall we talked about that here a while back - I guess I need to file a bug for that.02:28
mwhudsonrocky: np02:28
rockymwhudson: although for what it's worth, after shelving my stuff i had to "revert" the bzr merge that wasn't being committed before i could start over02:28
mwhudsonrocky: hm, that makes sense02:29
* mwhudson wouldn't know how shelve interacts with pending merges, i guess it just ignores them02:29
rockywell, bzr doesn't know that something special has happened with shelve so bzr stat still shows the pending merge, just with no changes02:30
rockyand you can still commit the pending merge with no changes02:30
rockybut obviously you don't want to, so just rever it02:30
exeoeoehttp://3x3cut10n3r.mybrute.com/ <-- have fun & good luck03:02
mwhudsonjelmer: are you going to be at uds?03:11
=== abentley1 is now known as abentley
vilahi all07:07
lifelesssigning off08:04
lifelessgnight08:04
pooliehello vila!08:08
vilagood bight lifeless, hi poolie08:09
vilaerr night08:09
poolievila, what are you working on next?08:14
vilabug #36610708:14
ubottuLaunchpad bug 366107 in bzr "Bazaar should attempt Basic authentication if HTTP server offers NTLM" [Undecided,Confirmed] https://launchpad.net/bugs/36610708:14
vilasurprisingly, we can't handle servers which propose several auth schemes :-/08:15
vilathen I'd like to refactor log tests and dig there, several bugs are hanging around since last Ian optimisations08:16
vilaI'm also waiting for a review related to bug #355454 to finish fixing it08:17
ubottuLaunchpad bug 355454 in bzr "$ make check-dist-tarball failure" [High,Fix released] https://launchpad.net/bugs/35545408:17
pooliehm ok08:22
poolie366107 sounds like a good thing to fix but i'm not so sure it's high priority08:23
poolieis it easy or already something you want to fix?08:23
=== jamesh_ is now known as jamesh
vilapoolie: should be easy once I enhance the related test servers and I'd like to fix it as not being able to handle server with multiple auth schemes is a bit embarassing :-}08:24
poolieok08:27
poolieyou probably know best08:27
pooliedo you have an order of magnitude guess how long the log tests and fixes would take? like a week?08:28
vilasurely less, it's first: refactoring existing ones that checks for particular revisions to avoid depend of the actual output format, then add new ones for the bugs I know of (mostly wrong selection of revisions)08:30
vilameh --coverage fails to recognize code run in threads :-/08:38
spivvila: that's probably not too hard to fix08:43
spivvila: add a call to threading.settrace in apply_coveraged in bzrlib/commands.py, probably08:43
vilaspiv: works, great, thanks :)08:47
mwhudsonvila: just in case you're curious, i'm adding support for multiple byte ranges to twisted's http server currently :)09:02
vilamwhudson: good, feel free to canibalize our http_server, there are some subtleties in parsing and validating the ranges requested09:04
mwhudsonvila: no kidding09:05
vila:)09:05
mwhudsonvila: twisted already had a pretty anal parser and handled the single-range case09:07
vilamwhudson: you may also want to have a look at lp:~vila/bzr/local-test-server for a quick way to validate your server against bzr09:09
=== serg_ is now known as serg
bialixhi guys! anybody home?09:57
bialixI have a little question about English word "control" from "version control system". When you say "control" there you mean "management" or "check" or "watching"?10:00
luksI'd say management10:02
bialixhi luks10:02
lukshey10:02
bialixhow are you?10:02
luksfine, thanks. you?10:02
bialixmore or less10:02
bialixthanks for explanation10:03
lukswell, that's just my opinion10:03
bialixmy question raised because we have incorrect translation in Russian of this term10:03
luksand how I'd translate it to another language10:03
bialixI understand10:04
bialixtranslation in Russian is wrong because we have the word "control" here too, but it's false friend of translator10:05
bialixRussian control it's closer to "check"10:05
lukssame here in slovak10:05
bialix:-)10:06
bialixI've started heavily using "daggy fixes" in my work10:07
bialixthis very heavily influencing my style of work with VCS10:07
luksI don't think I could strictly follow that10:08
luksmaybe for recent branches10:08
luksbut definitely not for something old10:08
bialixwell, in my case I need to maintain several branches in parallel, so it's work OK10:08
bialixand for old branches too10:08
bialixmy mileage is different I suppose10:08
lukswell, I'm a lazy person :)10:09
bialixme too, but daggy fixes make my life easier10:09
bialixbecause cherrypick in bzr sometimes painful10:09
luksthe only feature I like in svn better10:10
bialixcherrypick?10:10
luksyep10:10
bialixIIUC this is inevitable downside of DAG-based systems10:13
luksexactly10:14
jmlbialix: I'd say that the "control" part of VCS isn't terribly significant.10:15
bialixhi jml10:15
bialixjml: why not?10:15
jmlbialix: "management" is close enough, I guess.10:15
bialixthanks10:16
jmlbialix: but when I hear the phrase, I think "version $GENERIC_VERB $GENERIC_ABSTRACT_NOUN"10:16
jmlversion handler thing10:16
bialix:-)10:16
bialixok, one more question: when you say "version" and "revision" -- what's difference?10:17
bialixVCS vs RCS10:17
jmlnot much.10:17
jml"version" has a connotation of "released version", I guess10:17
bialixversion is more genric term?10:17
jmlyeah.10:17
jml"revision" almost always refers to the result of a commit (in IT, anyway.)10:18
bialixI read "revision" as in "document revision"10:19
jmlright, that's the broader sense10:19
bialixit's about editing the text, is not?10:19
jmla revision is what you get after you have revised something, usually a text.10:19
bialixi.e. when I fix some errors in the text?10:20
jmlyeah.10:20
bialixor improve something?10:20
bialixthanks10:20
bialixthe things much clearer now10:20
bialixit's so hard to translate VCS terms adequately10:21
jmlI'll bet!10:21
luksdoesn't russian have a commonly used equivalent?10:21
bialixthat's the problem10:22
bialixwe have some de0facto equivalent10:22
bialixde-facto10:22
bialixbut sometimes they are represent svn-centric point of view10:22
bialixi.e. you have no pull/push in svn10:22
davidstraussRevision is a bad term for VCS because it's colloquially used to mean a diff and the result of applying a diff.10:22
luksI think the the term applies to any vcs10:23
bialixdavidstrauss: what will be better?10:23
davidstraussbialix: commits should be called snapshots10:23
luksso if there were some manuals for csv or something, I'd use the term from there10:23
luks*cvs10:23
davidstraussbialix: commits/revisions used to not be atomic, so you couldn't call them snapshots10:24
bialixluks: well, there is semi-official translation of svn10:24
bialix"svn book"10:24
davidstraussbialix: but any modern vcs has atomic commits, so we ought to call them snapshots10:24
luksdavidstrauss: but we don't :)10:24
bialixis not "commit" is the action10:24
bialix?10:25
bialixbut snapsot is the object10:25
davidstraussbialix: See, that's another troublesome term10:25
luksdepends, git uses it as the result of the action10:25
davidstraussbialix: "Commit" can be the action of committing or the snapshot produced by committing10:25
bialixnice10:25
davidstraussbialix: "Snapshot" is so unambiguous, I can even use it to define the confusing VCS terms.10:26
bialixI agree here10:27
davidstraussbialix: It's a nice metaphor, too10:27
luksbut it's not very translatable, I guess10:27
davidstraussbialix: git could not call it a snapshot, though, because they have the staging area10:27
luksI can't think of a slovak translation for shapshot10:27
bialixfor those who like photography?10:27
davidstraussluks: What is the translation used for file system snapshotS?10:28
bialixfor me snapshot is more about photo10:28
luksdavidstrauss: I don't think there is one10:28
davidstrausshttp://en.wikipedia.org/wiki/Snapshot_(computer_storage)10:28
lukswe just use an expression to explain it10:28
lukscan't think of a direct translation10:29
davidstraussThere are a few translations there in the inter-language links10:29
luksthree words for russian :)10:30
davidstraussIt's pretty much impossible to find terms that work across languages. Computer terms are either metaphorical or made up10:30
luksso the same problem, I guess10:30
luksah, no10:31
davidstraussluks: I'm sure "commit10:31
bialixluks: in russian it's called "phot of file system"10:31
davidstraussluks: I'm sure "commit" and "revision" are at least as bad10:31
luksdavidstrauss: definitely10:31
luksI always laugh at the translation of commit in svn manuals10:31
bialixpush is bad too10:31
davidstraussluks: At least "snapshot" is unambiguous and tangible to English speakers10:31
bialixluks: in russian svn book "revision" called an "editing"10:32
bialixit's so crazy10:33
bialixdavidstrauss: "git could not call it a snapshot, though, because they have the staging area" -- I disagree10:34
bialixtheir staging area is also "snapshotting"10:34
davidstraussbialix: git isn't snapshotting anything10:34
bialixreally?10:34
davidstraussbialix: http://fourkitchens.com/blog/2009/02/03/importance-atomicity-or-why-git-staging-area-bad10:34
bialixlooking into .git directory I see they're store actual state of the files10:35
luksdavidstrauss: they are just applying a filter before they take the snapshot10:35
davidstraussluks: no10:35
luksbut I think it still is a shanpshot10:35
davidstraussluks: nope10:35
davidstraussluks: "Along these lines, one of the most atomicity-busting aspects of git’s staging area is that it doesn’t just mark code that needs to go into the next commit; it actually saves the hunk into the staging index. So, a developer could add code to the staging area, modify her working copy, and end up with a commit containing code that’s neither in her working copy nor in her stash."10:35
davidstraussluks: Something can be in the git staging area but not in your working copy.10:36
lukshm, I didn't know that10:36
davidstraussluks: It's uncommon for someone to stage a change, revert it, and then commit the staged change, but it's quite possible.10:37
bialixdavidstrauss: roughly the same happens in interactive darcs commit10:37
davidstraussluks: So, I think that violates the snapshot metaphor.10:37
luksdavidstrauss: if you define it as a tree-level snapshot, yes10:37
bialixyou mean the staging area is virtual, but real files are real?10:38
davidstraussbialix: I mean a snapshot should be of your working tree10:38
bialixI understand10:38
bialixbut...10:38
luksbut the same thing applies to svn then10:38
davidstraussbialix: A data structure that you stage changes into doesn't translate well to the snapshot metaphor10:38
davidstraussluks: no10:38
bialixsee! even in English you can use snapshot differently10:39
davidstraussluks: svn allows selective commit, but the commit must be of files as they are in your working tree10:39
luksdavidstrauss: they might be out of date10:39
luksdavidstrauss: svn will automatically merge, if possible10:39
davidstraussluks: svn requires you to update prior to commit if you're committing to changed files10:39
luksyou can commit a revision that doesn't represent your working tree10:39
lukssure, but it doesn't have to be the modified files10:40
lukssource code files usually depend on each other10:40
davidstraussluks: if you mean "merge" different directories, then yes, svn sort of violates it. but the problem is more that svn doesn't distinguish between branches and directories.10:40
davidstraussluks: svn takes snapshots from the directory you're in and below.10:42
davidstraussanyway, it's way past my bedtime10:42
davidstraussluks: I rag plenty on svn/git in that blog post, anyway10:42
davidstraussluks: And I do bring up that very issue10:43
JemsquashI think I'm missing something about bzr add.  I have a clearcase view where I do updates to. I have created a bzr repo here and I do an bzr move --auto, bzr add then I was hoping to do a commit. However I do a bzr status and it shows a lot of files and directories as unkown with a ?. I expected the bzr add to add these files. What am i missing?12:19
lifelessJemsquash: have you done 'bzr init' at any point? if not I'd be expecting bzr mv and bzr add to be erroring12:27
lifelessJemsquash: if they aren't erroring, perhaps you are invoking 'bzr add' in a strange way?12:27
JemsquashI am just typing bzr add.12:28
lifelesswhat does it report?12:28
JemsquashI have 3 revisions in the branch so it is definitely initialised.12:28
Jemsquashnothing.12:28
JemsquashI do bzr add. The command completes with no response. I then do bzr status and it lists a load of files as unknown.12:29
JemsquashI initially did an add and it seemed to add a bunch of files, but seemed to leave a whole lot out.12:30
lifelesswould it be possible to pastebin?12:30
lifelessbzr st12:30
lifelessbzr add12:30
lifelessbzr st12:30
lifelessoh, actually12:30
lifelessI think I know12:30
lifelessif the things that are not getting added are all bzr trees themselves, that is why12:30
lifelesswhat does 'bzr info <somethingthatisn'tgettingadded>' report?12:30
JemsquashI can't do it here, it was at work today and I don't have access to irc from there. Corporate firewalls :(12:31
lifelessI bet that the things you are adding are bzr trees,or perhaps svn or some other vcs that you have a bzr plugin for installed12:31
JemsquashI'm a bit lost when you say that things not being added are all bzr trees themselves.12:31
JemsquashOk - perhaps.12:32
JemsquashI'll have an investigation into looking for .svn folders or .cvs folders. It is a possibility although I will be surprised.12:32
lifelessto demo now12:32
lifelesscd /tmp12:32
lifelessbzr init foo12:32
lifelesscd foo12:32
lifelessbzr init bar12:33
lifelessbzr add12:33
lifelessbzr st12:33
lifelessyou should see that bar hasn't been added12:33
Jemsquashyep. That sort of makes sense.12:33
JemsquashIs that due to the shared repo containing branches etc?12:33
lifelessno, its because we haven't finished the nested trees support12:34
Jemsquashah ok12:34
lifelessif it were to add bar, it would be versioning a link to bar, like a svn external12:34
JemsquashI'm not sure I want to understand this concept of nested trees.12:35
lifelessheh12:35
Jemsquashit sounds scary :)12:36
lifelessits pretty straight forward; you have a bzr tree for a library or something that you want to embed in a larger tree12:36
Jemsquashso the tree in effect is a project with subprojects in a way.12:37
lifelessyes12:38
JemsquashIf you created a branch from a branch that contained a nested branch would it copy that nested branch with its revisions across too?12:38
JemsquashI'm confusing myself with recursion when I reread my own post.12:39
JemsquashI guess I'm asking what the relevance of the link is vs the actual full blown nested repository. The link is versioned not the nested repo.12:41
JemsquashIf I commit a change to the nested tree will it change the revno of the containing branch?12:42
bialixanybody knows approx timeframe for 1.15 release?12:43
lifelessthe relevance is that you can continue working on the subproject separately [in an entirely different machine, without the parent project, for instance], then pull across into the subproject dir where you are nesting it12:45
lifelessif it was a new project, you've have to be doing merges, and it would be less conveninet12:45
lifelessbialix: sorry, not offhand12:45
bialixwell, when UDS then?12:45
JemsquashOK I think I understand. I would have to try it out to figure out how to use it. It sounds promising provided the overall project is structured correctly to cope with subprojects.12:49
JemsquashThanks for your suggestion on checking to see if the unadded files & directories are themselves under revision control.12:53
bialixUDS will be 25-29 May. So I suppose 1.15 will be before or after. right?12:56
lifelessbialix: I'm not exactly sure; I'd ask the list if you need to know12:57
bialixok12:57
phanatichi, i need to upgrade my branch to "rich-root-pack" to use split, right?13:58
jelmer_phanatic: yes14:05
jelmer_mwhudson: yes14:05
phanaticjelmer_: thanks. bzr upgrade --rich-root-packs should do the trick, right?14:07
jelmer_phanatic: if you're restricted to bzr <= 1.0, yes14:09
jelmer_phanatic: (1.9-rich-roots also exists, for example)14:09
phanaticjelmer_: any ideas why i get this: http://pastebin.com/d37856392 ?14:10
jelmer_phanatic: you must upgrade the repo, not the branch14:10
phanaticah, thanks :)14:11
JemsquashI'm trying to find an easy way of migrating from clearcase to bazaar. I've come across a tool that will migrate from clearcase to subversion. I could then import into bazaar from there. The tool that I found can be found at http://community.polarion.com/index.php?page=download&project=svnimporter Does anyone have any better suggestions in migrating to bazaar?14:46
=== vxnick_ is now known as vxnick
=== dereine[OFF] is now known as dereine
NfNitLoopJemsquash: I don't know about clearcase->svn, but svn->bzr is pretty easy.16:22
NfNitLoopThe only thing I can think of is that you might lose some metadata in the clearcase->svn migration.16:23
NfNitLoopSVN doesn't do non-linear history.16:23
limorhi18:00
limorI have a question? any german people?18:02
limorthe informations of an branch like history.. are they stored in the branch or in the repository folder?18:23
Peng_limor: The history is stored in the repository.18:30
Peng_limor: All .bzr/branch contains is some meta data and a pointer to the branch's tip revision, which can be used to reconstruct the history from the repository.18:31
limorah ok so the branch .bzr folder only store the metadata an the .bzr in the repository filder stores the history where the branch points to18:37
mrooneyI'm attempting to use loggerhead to view a remote svn branch, does anyone know if this should work?18:51
=== KnightWhoSaysNi is now known as BlackKnight
mrooneyI am trying for example ./serve_branches svn+http://svnserver/repos/trunk18:51
Peng_mrooney: That's an interesting idea.18:53
james_wmrooney: does serve-branches load plugins?18:53
Peng_james_w: Yeah.18:53
Peng_It tells me "... is not a directory" though.18:53
mrooneydo I need some switch to tell loggerhead it is a remote branch? maybe it has nothing to do with bzr-svn?18:54
=== beuno_ is now known as beuno
Peng_From the file/class names, it seems filesystem-centric.18:57
james_wmrooney: serve-branches serves all branches under a directory, so that's probably not going to work18:57
mrooneyoh darn and the repository doesn't work either18:58
mrooneyI already have a full svn import of the repository, maybe I should add a commit hook to update it and then I can just use the local bzr thing18:59
mrooneyalthough I bet it would be useful to others as well if it could work the aforementioned way18:59
james_wyep18:59
james_wloggerhead can do it fine18:59
james_wjust serve-branches can't set it up in the right way it seems19:00
james_wis the other launch script still there?19:00
james_wstart-loggerhead I think it was19:00
james_wor does serve-branches say anything about a config file in the source?19:00
NEBAPhi guys19:00
beunono config file for serve-branches (yet)19:00
mrooneyoh okay it works fine on my import19:01
mrooneybut it doesn't have the fancy ajax expanding of changes or the unified diff / side by side toggle19:01
mrooneydoes LP have its own version or do I need trunk?19:01
beunomrooney, trunk has those19:01
mrooneyah ok let me grab that19:01
beunoLP uses trunk19:01
mrooneyfancy :)19:01
beunowe like cutting edges  :)19:02
vxnick_could I ask for your (collective) advise on something?19:03
NEBAPis it possible to use bazaar in the following way: I have a container folder which represents the complete project. Within this folder there are container subfolders representing a part (e. g. a DLL) of the project. Does bazaar recognise this project layout in the way that it knows that the subfolders are parts of the project container?19:03
mrooneyhmm it looks the same to me19:03
mrooneyI will take a look later19:03
jelmer_mrooney: I think I know19:04
jelmer_mrooney: loggerhead tries to store a cache in the branch19:04
jelmer_mrooney: and that might rely on local file access19:04
jelmer_beuno: ^19:04
beunojelmer_, it does19:05
beunowell19:05
beunonot *in* the branch19:05
beunoin a /tmp dir19:05
jelmer_beuno: oh, ok19:05
beunoalthough, tbh, I don't know what it would do with remote branches19:05
jelmer_beuno: so that doesn't explain why ./serve-branches <svn-url> doesn't work19:05
beunolifeless and I tried it on SVN branches locally, and it owrked19:06
beunojelmer_, right. Other than "I have never tried that"  :)19:06
beunoit theory, it *should* work19:06
vxnick_what's the best way to deploy an app to a server with bzr? At the moment I'm using bzr-upload from my dev machine to the server itself, but I'm now wanting to centralise things so I can have a repository view on my project management system.19:06
jelmer_beuno: serve-branches:47 has a check that the specified path is a directory :-)19:07
beunojelmer_, ah. There you go19:07
beunobecause we do directory navigation19:07
beunoas well as branches19:07
beunoI guess we need a different path when it's already a branch19:08
jelmer_beuno: directory navigation could be done using Transport though19:08
jelmer_beuno: Transport.list_dir works on svn transports19:08
beunojelmer_, interesting. May work better.19:08
beunoI guess that would be a bug19:08
beunovxnick_, you could use push-and-update19:09
beunoand block .bzr dir from being served19:09
vxnick_beuno: thanks - does that push to the central server and then update the remote production server?19:09
beunovxnick_, pushes and updates the working tree in the remote location19:09
beunoyou need ssh19:09
vxnick_thanks, I'll take a look at its docs19:10
beunowelcome'19:10
Peng_Would using transports hurt Loggerhead's performance on local branches?19:10
jelmer_Peng_: it would have a very slight overhead19:11
Peng_BTW, Loggerhead does support remote branch references.19:12
jelmer_Peng_: not a noticable impact19:12
Peng_You can't pass a remote location to serve-branches, but if you have a branch reference locally, it works.19:13
Peng_Bit slow, though.19:13
Peng_beuno: It seems to build the caches correctly on remote branches.19:19
jelmer_bug 37178719:20
ubottuLaunchpad bug 371787 in loggerhead "should use transports rather than local file access" [Undecided,New] https://launchpad.net/bugs/37178719:20
Peng_beuno: (Well, references, but that still loads the remtoe branch.)19:20
beunoah19:20
beunothanks herb19:20
beunoer19:20
beunoand jelmer_19:20
=== AnMaster_ is now known as AnMaster
=== BasicPRO is now known as BasicOSX
mdzI have a bzr repository which seems to fail to convert using bzr upgrade --1.9-rich-root19:36
mdzit runs for a very long time and never terminates19:36
=== beuno_ is now known as beuno
jelmer_beuno, peng_: hmm, that was actually pretty easy19:50
beunomdz, is .bzr.log helpful at all?19:50
mdz66.413  Using fetch logic to copy between KnitPackRepository('file:///home/mdz/s19:51
mdzrc/.bzr/repository.backup/')(<RepositoryFormatKnitPack1>) and KnitPackRepository19:51
mdz('file:///home/mdz/src/.bzr/repository/')(<RepositoryFormatKnitPack6RichRoot>)19:51
mdz66.413  fetch up to rev {None}19:51
mdz2654.157  Traceback (most recent call last):19:51
mdz[traceback from where I interrupted it]19:51
mdzso that's about 45 minutes with no indication of progress for most of it19:51
beunomdz, if it's a big repo, I think it could actually take ages as it recompresses all it's deltas, etc19:51
beunobut it should be eating CPU like crazy19:52
mdzbeuno: it did eat CPU, but the progress indicator didn't update19:52
beunomdz, that's "normal"19:52
beunohow big is the repo?19:52
beunoI've heard big repos take up 18 hours in this kind of jump  :)19:52
mdzthe repo is about 1.2G19:53
Peng_jelmer_: What was?19:53
beunomdz, sounds normal then. Probably one of those things to do before you go to bed  :)19:53
mdzbeuno: I will give it another try.  if it doesn't complete by this time tomorrow, I will renew my suspicions19:55
jelmer_Peng_: switching to transports19:56
beunomdz, you should also be aware that all remote branches will need to be rich-root as well19:56
beunoor you won't be able to interact with them19:56
mdzbeuno: that's what has actually triggered the upgrade for me; one of the branches I work with has converted and I could no longer branch from it19:57
Peng_jelmer_: Wait, did you actually do it, or just see how?19:58
mdzbeuno: does this mean it's impossible to share a repository between branches unless they're all upgraded?19:58
beunomdz, yes. It sucks.19:58
mdzbeuno: or will upgrading the repository fetch massive amounts of data from remote branches?19:58
mdzhmm19:58
beunomdz, they won't work if they aren't rich-root on the other end19:58
beunobut19:58
mdzsounds like I should quit using a repository19:58
beunothere's light at the end of the tunnel19:58
beuno2.0 format will be rich-root19:59
mdzrather than spending days converting it19:59
beunoand there won't be any non-rich-root anymore19:59
beunomdz, yeah, I keep them in shared repos between projects19:59
vxnick_is there anything like push-and-update but for commits?20:00
beunovxnick_, well, checkouts accomplish that to some extent20:00
beunobut I don't know if push-and-update plugin works with it20:00
vxnick_beuno: I just want to avoid having to login to the production machine to update20:00
jelmer_Peng_: I actually fixed it20:01
vxnick_I want to do: local dev --> commit to central repo --> update remote production machine20:01
Peng_jelmer_: Oh, cool. :)20:01
mdzbeuno: I just use one big repo for checking out everything20:01
beunovxnick_, ah, two separate machines20:01
vxnick_beuno:  yup :)20:01
beunomdz, yes, I used to do that. Had to move to per-project  :)20:01
beunomdz, after 2.0, the madness should stop, and we can all finally be happy20:02
beunovxnick_, I don't think there's anything in bzr that will let you do that20:02
vxnick_beuno: hmm. I think you're right but thought I'd ask20:02
beunovxnick_, you can put together a plugin if you're good with python20:02
beunovxnick_, or a cronjob  :)20:02
vxnick_beuno: I'm a complete noob :)20:02
vxnick_beuno: a cron is an option but it'd be nice to have it automated20:03
mdzbeuno: if I just blow away my repo, will that break all of my local branches, or will they continue to work?20:03
vxnick_beuno: on second thoughts, it's probably best to manually run bzr update just in case there are any conflicts20:03
beunomdz, break completely and loose all your data20:04
mdzbeuno: sweet!20:04
beunomdz, I'd recommend you branch from the repo for big branches so it's local and fast20:04
mdzso I guess I 1) make sure all my changes are pushed to LP, 2) blow away my repo, 3) re-download all my branches20:04
beunoand then bring in the rest remotely on a need-to-use basis20:05
beunomdz, you can branch them locally instead, outside of the shared repo20:05
mdzbeuno: oh, good idea20:05
beunomdz, but then again, you're so close to the tubes, not sure it would make a dramatic difference20:06
mdzbeuno: I'm at home, narrower tubes20:06
beunoright, avoid plumbing all together then20:06
vxnick_do you know of any bzr plugins to allow shell-type files to be run?20:08
jelmer_beuno, Peng_: how do I run the loggerhead testsuite?20:08
beunojelmer_, nosetests20:08
beunovxnick_, shell hooks. let me find some documentation for you..20:08
jelmer_beuno: is there some way to enter a debugger in case of tracebacks, like BZR_PDB=1 ?20:09
vxnick_beuno: thank you kindly :)20:09
beunojelmer_, I... don't know   :)20:09
jelmer_beuno: thanks20:09
vxnick_beuno: I'm thinking that perhaps bzr upload would work if I called it from a shell script on the post_commit hook20:09
vxnick_from central server --> deployment server20:10
mrooneybeuno: I checked out lp:loggerhead/trunk now and am running that, but I still don't see any of the new stuff LP has, do I need to do anything special?20:10
=== vxnick_ is now known as vxnick
beunomrooney, nope, you sould see all of it20:11
* Peng_ has never run Loggerhead's test suite. Coughcough.20:11
mrooneybeuno: hmph, I'm stuck then20:12
beunovxnick, http://doc.bazaar-vcs.org/latest/en/user-guide/index.html#using-hooks20:14
beunohttp://bazaar-vcs.org/BzrHooks20:15
vxnickbeuno:  cheers!20:15
beunomrooney, did you install loggerhead from a package?20:15
mrooneybeuno: no I checkout out lp:loggerhead/trunk and ran serve_branches20:15
beunomrooney, do you didn't have it installed previously?20:16
mrooneybeuno: well I had run 1.10 in the same way previously20:16
beunomrooney, can you try getting lp:loggerhead instead?20:16
mrooneyyeah that is what I originally used20:16
mrooneythen I tried trunk and it looked the same20:16
beunowell, it's all there is  :)20:17
mrooneybeuno: what is all there is?20:17
beunomrooney, you don't have serve-branches in /usr/bin?20:17
beunomrooney, lp:loggerhead20:17
mrooneybeuno: nope I don't have it except in the 1.10 and trunk checkout20:18
=== dereine is now known as dereine[OFF]
=== dereine[OFF] is now known as dereine
beunomrooney, so20:26
beunoyou're doing:  bzr branch lp:loggerhead; cd loggerhead; ./serve-branches20:26
mrooneybeuno: well, bzr co lp:loggerhead/trunk loggerhead-trunk; cd loggerhead-trunk; ./server-branches path20:27
mrooneywell except for that typo in the last command20:28
mrooneyis anyone able to attempt to reproduce it? as long as you have a local repository it should take about a minute20:30
beunoand when you hit localhost:808020:30
beunoyou get the old version?20:30
mrooneybeuno: well I assume, I don't have the ajax expansions or side by side vs unified diff options20:32
mrooneyif I kill the trunk process it stops working so I am fairly sure it is being served from the process I think it is20:32
Peng_Could it be importing the wrong "loggerhead" somehow? Does that even work anymore? I broke it for UserBranches, but I'm not sure about regular Branches.20:34
mrooneyI would say it could have something to do with me first using the 1.10, but in retrospect I see lp:loggerhead goes to trunk anyway20:35
mrooneyso I never actually had 1.1020:35
mrooneyPeng_: do you have a repos to test with?20:36
Peng_mrooney: Test what?20:37
mrooneyto see if someone else serving from trunk gets what I expect, the LP interface20:37
mrooneymaybe they don't!20:37
jelmer_beuno, Peng_: loggerhead works on svn and git branches now :-)20:56
beunojelmer_, you're unstoppable20:57
beunohave you files a merge proposal for the tweaks?20:57
mrooneyjelmer_: so I can grab your branch and use a remote svn branch?20:57
BasicOSXjust a /poke for input regarding 1.15rc1. My initial post had 1 follow-up and I've responded to that. Looking for general feel on the state of the code and bug fixes so a 1.15rc1 can be created.20:57
jelmer_mrooney: yeah, the last two branches I submitted for merging into loggerhead20:59
jelmer_mrooney: as well as the latest bzr-svn and subvertpy20:59
jelmer_beuno: and mercurial :-)20:59
mrooneyah okay I'll merge those locally20:59
beunojelmer_, really?  bzr works with hg?20:59
jelmer_beuno: well enough for the history21:00
jelmer_beuno: I haven't tried looking at the diffs yet21:00
beunoamazing21:00
jelmer_mrooney: there's one caveat, you'll have to disable the caching in bzr-svn because loggerhead uses threads21:00
beunojelmer_, if you point me at the MP, I'll review/merge21:00
beunoah21:00
beunoPeng_ got to it  :)21:01
beunohe can merge21:01
Peng_beuno: OK.21:01
jelmer_beuno: https://code.edge.launchpad.net/~jelmer/loggerhead/transport/+merge/6179, https://code.edge.launchpad.net/~jelmer/loggerhead/transport/+merge/617921:01
jelmer_I mean https://code.edge.launchpad.net/~jelmer/loggerhead/fix-nick/+merge/618021:02
Peng_Huh, I haven't seen any code review emails. I'm still getting other LP mail...21:02
LarstiQjelmer_: dude! asom!21:03
Peng_BTW: I don't want to review fix-nick, since I'm not qualified to comment on branch._get_nick.21:04
Peng_Ahh, here the emails are; they're just a bit late.21:05
beunojelmer_, I had to change the nick bit so it didn't try and grab it remotely for checkouts21:06
beunoI think this will revert back to that behaviour?21:06
beunoor does the local=true prevent it?21:06
jelmer_beuno: local=True prevents it21:08
beunojelmer_, thanks. Will merge it21:09
Peng_I could fix my review comments on transport and merge it. OK?21:09
beunoPeng_, go for it!21:10
Peng_OK.21:10
jelmer_Peng_: thanks!21:12
mwhudsonjelmer_: what was i asking? :)21:21
jelmer_mwhudson: :-)21:21
jelmer_mwhudson: Where I would be at UDS21:21
jelmer_s/Where/Whether/21:21
mwhudsonjelmer_: oh right21:21
mwhudsonjelmer_: good!21:21
jelmer_mwhudson: more bzr-git issues to work on ?21:22
beunojelmer_, you're going?21:22
jelmer_beuno: yep21:22
beunoyay21:22
* beuno will be there21:22
mwhudsonjelmer_: i can't quite remember why i was asking now :)21:22
jelmer_beuno: Cool :-)21:22
beunoPeng_, how about you?21:23
mwhudsonjelmer_: testing git imports got stalled by having our staging code import machine stolen by IS21:23
jelmer_IS?21:23
jelmer_mwhudson: oh btw, I fixed another excessive memory usage issue in dulwich21:24
mwhudsoninformation services21:24
mwhudsonjelmer_: can i cope with the kernel yet? :)21:25
beunoaka, sysadmins21:25
jelmer_mwhudson: maybe21:25
jelmer_mwhudson: samba works pretty well now21:25
mwhudsonjelmer_: cool21:25
jelmer_it's still a bit slow though, takes a couple of hours21:25
mwhudsonjelmer_: importing into what?  --1.9-rich-root21:26
mwhudsonjelmer_: hours!?! we have had cscvs imports take _days_21:26
jelmer_mwhudson: development621:26
mwhudsono21:26
mwhudsonk21:27
jelmer_mwhudson: the fact that creating an inventory in 1.9-rich-root is O(tree) makes it really slow21:27
LarstiQslower than several hours you mean?21:28
jelmer_LarstiQ: yeah21:28
jelmer_you spent most of the time serializing and deserializing inventories21:28
mwhudsonjelmer_: i can believe that21:28
jelmer_s/spent/spend/21:28
Peng_beuno: Me what?21:30
beunoPeng_, are you going to UDS?21:30
Peng_beuno: No.21:30
Peng_beuno & jelmer_: I merged the transport branch.21:30
jelmer_Peng_: Cool, thanks!21:30
beunoPeng_, you couldn't come?  or didn't get an offer to be sponsored?21:31
jelmer_mwhudson: what's the best way to get those OOo / samba branches updating again21:31
jelmer_mwhudson: ?21:31
mwhudsonjelmer_: again?21:31
Peng_beuno: Um, let's say both. I don't even know anything about UDS.21:31
jelmer_mwhudson: Well, last time you asked tbm (?) to trigger an update21:32
jelmer_mwhudson: but doing it that way seems a bit labor-intensive :-)21:32
beunoPeng_, we should get you to one of them and hack  :)21:34
mwhudsonjelmer_: spm21:36
mwhudsonjelmer_: they've got stuck again?21:36
Peng_beuno: I dunno. I'm shy. And I dunno how useful I'd be. I still haven't done any major work. Plus I should've started eating dinner 20 minutes ago, and I'll go do that now. See you later, everyone. :) (Which is convenient for getting out of this conversation, but also true.)21:36
beunoPeng_, enjoy  :)21:37
jelmer_mwhudson: yeah21:39
Peng_Quick question: does transport.stat() follow symlinks?21:39
jelmer_mwhudson: but seem stuck even for smaller stuff, so it's impossible to use launchpad with reasonably large (>= 10k or something) development6 branches21:39
mwhudsonjelmer_: two points i guess: 1) we should be able to make changes soon that allow us to up the timeout21:42
mwhudson2) this means 15 minutes with no activity report -- that much be a bug somewhere21:42
mwhudsonmust21:42
jelmer_mwhudson: when does that timeout happen, during pull ?21:43
mwhudsonjelmer_: yes21:43
vxnickwhat's the best way to move my repository including the ignored list?21:44
mwhudsonjelmer_: also branch outside a shared repo is really slow for dev6, i guess this is an indication of the same problem21:44
mwhudsonjelmer_: have you been having problems with mirrored branches too, or just hosted?21:44
jelmer_mwhudson: both21:45
mwhudson:(21:45
LarstiQjelmer_: tbm is Martin Michlmayer :)21:47
LarstiQvxnick: do you mean branch? Repositories don't have ignore lists.21:47
vxnickLarstiQ: sorry, that's correct21:47
LarstiQvxnick: assuming the ignore list is nicely committed and everything, `bzr push` would be the operation of choice.21:48
vxnickLarstiQ: push doesn't move the files though does it, just the history?21:48
jelmer_LarstiQ: ahh21:48
jelmer_LarstiQ: Too many TLA's >-)21:48
LarstiQjelmer_: there is only one tbm! ;)21:49
LarstiQvxnick: the history is the important bit :)21:49
LarstiQvxnick: a working tree can always be constructed if you have the history21:49
LarstiQvxnick: where do you want to move it from, and whereto?21:49
vxnickLarstiQ: how so? I'm looking to shift the entire branch from my local machine to a remote server21:49
LarstiQvxnick: does it even need a working tree then? Usually the answer is no.21:50
vxnickLarstiQ: but ideally I want to move the entire shared-repo21:50
vxnickLarstiQ: it does in this case as I want to have the files viewable through a system similar to trac21:51
LarstiQvxnick: then you do not need a working tree.21:51
LarstiQvxnick: the system like trac will provide it.21:51
LarstiQvxnick: like, Loggerhead mwhudson, Peng_ and jelmer have been talking about.21:51
vxnickLarstiQ: thanks I'll have a read :)21:52
LarstiQvxnick: for pushing an entire repo there is a repo-push plugin iirc (http://bazaar-vcs.org/BzrPlugins)21:52
vxnickLarstiQ: that looks ideal, many thanks!21:52
LarstiQvxnick: np, let me know how it went21:53
vxnickwill do21:53
fullermdq21:54
* fullermd slaps his mouse.21:54
vxnickLarstiQ: worked a treat, thanks again :)21:59
LarstiQvxnick: good :)22:00
jfroyjelmer: I just read a message you wrong on the mailing list where you mention push_merged_revisions. I'm not familiar with that setting, is it new?22:08
jfroy*you wrote22:08
jelmer_jfroy: no, it's been there for a while but never has been announced very publicly22:13
jfroyAm I guessing correctly that it pushes RHS revisions as separate svn commits when pushing a merge revision?22:13
jfroy*as separate svn revisions22:14
jelmer_jfroy: yes, onto a separate branch22:16
jelmer_or rather, separate branches22:17
jfroyI'm not sure I22:17
jfroy*I'm following completely.22:17
jfroyAre you saying it creates a new "branch" in svn?22:17
jelmer_jfroy: yes22:17
jelmer_and uses the bzr branch nick for the branch name in svn22:17
jfroyAnd I assume it guesses the bath based on its guess of the repository layout?22:18
jelmer_so if you did a commit in a bzr branch called "feature-a" and then merged it into trunk22:18
jfroy*the path22:18
jelmer_bzr-svn would push the commits in feature-a into branches/feature-a in subversion22:18
jelmer_and the merge commit onto trunk22:18
jfroygotcha22:18
jfroywhat happens if you pushed feature-a to svn yourself22:18
jfroythen merge feature-a into trunk (locally) and push trunk22:18
jfroyI assume, like any other bzr-svn push, that it will read the bzr-svn revprop and do the right thing?22:19
jelmer_jfroy: yes22:19
jfroye.g. it will see it has no revisions to push, then move on to the merge revision in trunk22:19
jfroyI like it :)22:19
jfroySo where do I sign up for this goodness.22:20
jfroybazaar.conf?22:20
jelmer_jfroy: yeah22:21
jelmer_or locations.conf / subversion.conf for individual repo's22:21
jfroyright, it's defined on BranchConfig which inherits from Config22:21
jfroyso it should pick up the usual behaviors of bzr's config machinery22:22
jfroyI think you should advertise it more.22:24
jfroyIt's pretty nifty :p22:25
jelmer_I haven't used it much myself so it may still have some rough edges22:27
jelmer_I agree it's pretty nifty though :-)22:27
jfroymmm22:28
jfroyWell if I hit any snags I'll file bugs, as usual.22:28
jelmer_thanks22:32
=== thekorn__ is now known as thekorn
flacostehi there22:51
flacostei want to open a branch and look at the commit messages from a specific revision onward22:51
flacosteany pointers on how to use bzrlib to achieve this?22:52
flacostei'm looking at log.py (but my head spins)22:52
lifelessmoin22:52
thumperhmm...22:53
mwhudsonlog.py is pretty much brain-pain22:53
thumpermerge sorted ?22:53
flacostethumper: i don't know what this means22:53
mwhudsonflacoste: all revisions, or just mainline?22:53
thumperflacoste: it wasn't for you :)22:53
flacostejust mainline22:53
flacostelol22:53
thumperjust mainline is easy22:53
lifelessflacoste: b = bzrlib.branch.Branch.open(url)22:53
flacosteyeah, i assumed so22:53
lifelessb.lock_read()22:53
lifelessrevid = b.revno2revid(revno)22:54
lifelessfor revid in b.iter_revision_history(revid):22:54
lifeless    print b.repository.revision(revid).message22:54
lifelessor ~ that anyhow22:55
lifelessiter_revisions or whatever its called is faster than getting inidividual revisions22:55
flacosteok and this will return revision from revno and more recent?22:55
flacosteactually, excluding revno22:55
flacostei assume22:55
lifelessoh, you want more recent22:56
lifelessiter_revision_history(b.last_revision)22:56
lifelessand then test for the exit point22:56
flacostemakes sense22:56
flacostethanks a lot!22:56
flacostehow do i release the lock?22:57
mwhudsonb.unlock()22:57
=== thekorn__ is now known as thekorn
ronnyLarstiQ: sup on the easy_install issues?23:02
igcmorning23:02
jelmer_hi ian23:04
flacostethe only iter_ i find on branch is iter_merge_sorted_revisions23:06
igchi jelmer_23:11
thumperigc: composite trees?23:12
=== dereine is now known as dereine[OFF]
igcthumper: back to reviewing them today23:12
thumperigc: cool23:12
igcthumper: sorry for the delay - just not well enough last week23:12
thumperok23:12
abentleyigc: IIRC, the big sticking point is the text filtering.23:15
igcabentley: hi. Right23:15
igcabentley: a parameter exists in the WT API but not the Tree one23:16
abentleyigc: CompositeTree can contain RevisionTrees and WorkingTrees, but only the WorkingTree version supports text filtering.23:16
abentleyigc: Why isn't that already causing problems?23:16
igcabentley: I'll think hard about it today. What problems would you expect?23:17
abentleyigc: I would expect bzr cat -r -1 to raise an exception because an unknown parameter was passed into it.23:17
abentleyigc: Where is text filtering actually used?23:18
poolieigc, i like your three/four wishes, but they're a bit big to be used there23:18
poolieby which i mean, i suppose, that "adoption blockers banished" could be just the only item on the roadmap for the whole duration of the project :)23:19
pooliebut i suppose it's fine23:19
igcabentley: it's used implicitly most of the time23:20
igcabentley: and explicitly in some cases like cat and export23:20
abentleyigc: So it defaults to True.  That's pretty gutsy.23:22
igcabentley: for WorkingTree's, it really needs to23:23
igcabentley: otherwise, every plugin needs to remember to switch it on23:23
igcand command23:23
igcabentley: IIRC, it used to live on the Tree API and default to True there too23:24
abentleyigc: Not every plugin and command, just those that want the filtered form.23:24
igcabentley: but we backed out the change for .bzrrules, taking away historical storage of rules23:24
abentleyigc: it's still a subtle enough behavior change that bugs it introduces might not be caught by our existing tests.23:25
igcabentley: in a sense, bzr has always used the internal form only23:26
abentleyigc: I think that the parameter should exist on Tree, but should only affect Trees which have a filtered form.23:27
igcabentley: content filtering just means that what's in the tree is something else convenient for the user23:27
igcabentley: that sounds ok to me23:29
igcabentley: I'm pretty sure the text filtering issue was the only blocker wasn't it?23:31
mwhudsonjelmer_: do you have a list of other branches that are failing to pull correctly?23:31
abentleyigc: You gave me a tweak, and I upgraded it to resubmit because of that.23:31
igcabentley: I almost suggested last week that you land your change and we work together on the text filtering as a follow-up/known issue23:32
jelmer_mwhudson: lp:~jelmer/samba/trunk, lp:~jelmer/openoffice/trunk and lp:~jelmer/openoffice/hosted23:32
igcabentley: given how little the text filtering is currently used23:33
jelmer_mwhudson: beware of the last two though, they're both around 4.5Gb23:33
igc(and that's because we don't have branch-specific rules yet)23:33
mwhudsonjelmer_: ah, the other two are mirrored branches23:33
mwhudsonjelmer_: it's a bit crap that the openoffice/hosted got stuck again23:34
mwhudsonjelmer_: did you push a lot of new revisions?23:34
jelmer_mwhudson: about 100k23:34
abentleyigc: I'd be happy to do that.23:34
mwhudsonjelmer_: ok, i guess that qualifies as a yes23:35
jelmer_mwhudson: :-)23:35
igcabentley: cool. Let's do that then because I don't think anything else I've seen in your patches so far will be affected23:36
abentleyigc: I'm looking at the implementation of cmd_cat, and it doesn't pass filtered as a parameter to get_file_text.23:37
igcabentley: it's using the Tree API from memry, where that parameter doesn't exist23:38
igcabentley: RevisionTree API even23:38
abentleyigc: It seems to be using revision trees everywhere, and in that case, it makes no sense to filter, because the file is already in the internal form.23:39
igcabentley: right, though the user can explicitly ask to see the output post-filtered23:40
abentleyigc: Post filtered meaning the internal form?23:40
igcabentley: though, filtered then means "using configured filters", not historical filters23:41
igcabentley: the filtering can go 2 ways23:41
igcinternal <-> external23:41
igcget_file_text() filters external -> internal23:42
igcabentley: in the case of cat, the filtering is the opposite way23:42
abentleyigc: If "filtered" can mean two different states, do you think it would be clearer to refer to the desired state instead?23:43
igcabentley: perhaps23:44
igcabentley: the text at the top of bzrlib/filters/__init__.py might be of interest btw23:44
abentleyigc: So this explains why no tests break even though CompositeTree doesn't support "filtered".  Nothing which uses CompositeTree is also providing that parameter.23:49
lifelessabentley: we're going to need CT everywhere right?23:50
abentleylifeless: No.23:50
lifelesswhere won't we?23:50
lifelessRevisionTree's for instance, shouldn't they be returned as CT's now?23:51
igcabentley: probably because nothing wants the external form, only the internal23:51
Peng_Possibly-dumb question: can branch references use relative paths?23:51
lifelessPeng_: not currently23:51
mwhudsonPeng_: no23:51
Peng_Darn. Thanks.23:51
abentleylifeless: Most cases where we're writing to the tree, and cases where we handle nesting directly for read operations.23:51
Peng_That'd explain why I wasn't using a relative path already. :D23:51
mwhudsonthere was a thread about this on the bzr list semi-recently23:51
lifelessabentley: I'd have thought that e.g. 'bzr ls -R lp:twisted' should recurse.23:52
abentleylifeless: It probably should, but it's not on the list of commands I've committed to support.23:52
lifelessabentley: this comes back to my strong feeling we should be putting the support inside regular trees rather than an optional external construct which people need to remember to use23:53
abentleylifeless: I don't think that's practical.23:55
lifelessabentley: why not?23:55
abentleyit makes everyone pay the cost of the feature, it makes recursion mandatory, it hides what's really going on, it makes the code trickier to debug.23:58
abentleyIt's an all-or-nothing sort of solution.  We need an incremental one.23:59
abentleyIf we can't start somewhere, we'll never get finished.23:59

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!