[05:10] jelmer: hi === Mkaysi_ is now known as Mkaysi [10:46] emacs diff-mode is very interesting, thanks vila [10:56] rehi [11:03] jelmer: thanks for working on fixing the lp:gcc issues! Any ideas why it's broken? [11:03] is it just the hugeness of the tree that makes life hard for bzr? [11:04] or something deeper? [11:06] jimis: it's the number of tags (and the fact that they are not on mainline revisions) in particular [11:07] interesting :-) [11:07] how many tags are we talking about? [11:08] jimis: I think it was a couple of thousand [11:08] :-o [11:09] doing even basic work with this tree had given me trouble in the past [11:09] but all is well now with 2.4, at least for me [11:10] ofcourse it could be faster, but at least it's fast enough to do some basic work :-) [11:11] yeah, jam did a lot of work to improve the performance on trees that are gcc-sized [13:46] Hi [13:47] hi Anteru [13:47] I'm trying to write a plugin, and I'm somehow can't get started. [13:47] I have created a plugins/binstore folder in my home [13:47] and put an __init__.py as wel as binstore.py in there [13:47] Which are completely bare-bones [13:48] bzr plugins -v tells me [13:48] No module named binstore [13:48] Unable to load plugin u'binstore' from u'C:/Users/Anteru/AppData/Roaming/bazaar/ [13:48] 2.0/plugins' [13:48] err, that last line-break shouldn't be there [13:48] __init__.py is: http://pastie.org/2263571 [13:48] Anteru: what's in the __init__.py file? [13:48] ah [13:48] binstore.py: http://pastie.org/2263572 [13:49] Anteru: that looks like incorrect python syntax [13:49] what does "python __init__.py" say? [13:49] I have only python 3.2 installed, let me try anyway ... [13:50] ah I have 2.6 too [13:50] (the import on line 6 of __init__.py seems incorrectly formatted to me) [13:50] .bzr.log should also have some details [13:51] indeed [13:51] yeah, the import was wrong, now python __init__.py complains about missing bzrlib [13:51] Hm, bzr plugins -v doesn't work though [13:52] Anteru: any details in ~/.bzr.log ? [13:52] how does bzr plugins blow up? [13:53] just the error message [13:53] bzr log: http://pastie.org/2263596 [13:53] encoding issue? [13:56] I don't understand what exactly is missing? binstore is a symlink under /plugins/, but that should be transparent for bzr ... [13:57] I'm on Win7/bzr 2.3.3 fwiw [13:57] it looks like you have invalid information in your version_info [13:58] hm? [13:58] Anteru: can you comment out version_info and try again? [13:59] new __init__.py: http://pastie.org/2263612 bzr.log: http://pastie.org/2263615 [14:02] I also tried to replace binstore with blahstore to check if the from bzrlib.plugins. ... line is the problem, but it fails earlier [14:09] Anteru: have you tried with an empty __init__.py ? [14:09] hm, doesn't work either -- something is fishy here [14:09] Do I have to do anything besides putting a folder into my plugins directory? [14:10] putting it into the plugins directory should be sufficient [14:10] that said, I don't really have a lot of experience doing development on windows [14:11] are you using the python-based installer or the standalone installer? [14:12] I used the normal standalone installer [14:13] Everything "stock" as much as possible [14:14] Anteru: I think you might have to use the python based installer if you want to add plugins [14:14] IMBW [14:17] hm putting keywords there works [14:18] ah [14:18] ah? [14:18] hm [14:18] Well, I did a branch of lp:bzr-keywords to plugins and that works [14:19] after removing setup.py from there, it works still [14:19] if I copy and paste the __init__.py, I get the same error [14:21] So it's not my bazaar installation, but something specific to my binstore plugin [14:21] Ah [14:21] It does not work if binstore is a symlink [14:22] I.e. if I symlink something from elsewhere to plugins, bazaar can't find it [14:24] Uah, that's weird at best. [14:24] (me wonder what the heck bzr does that it breaks on symlinks on Windows) [14:24] Anteru: what do you mean by symlinks here? [14:25] Anteru: NTFS symbolic links? [14:25] the folder plugins/binstore was a symlink to V:/Dev/Current/Testing/binstore [14:25] Yes, NTFS symlink [14:25] Anteru: bzr just uses python's import [14:25] so presumably that has troubles with symlinks [14:26] yes but it somehow sees the folder "binstore" but fails afterwards. [14:26] Oh well, first hour of plugin development spent :) [14:32] Thanks! [14:34] Anteru: yeah, that's annoying [14:34] Anteru: are you working on a big file store of some sort? [14:38] Yeah, the one I wrote on the mailing list [14:38] I'll be back in ~ 1.5 hours, gotta go [14:38] how can I see a what changes a "bzr merge" will do, but without merging yet? [14:39] or how can I "bzr diff" current branch with another branch? [14:41] jimis: bzr diff should work [14:43] jelmer, I was expecting the following to diff "mytrunk" to current branch: [14:43] bzr diff --old=mytrunk [14:43] it outputs nothing [14:45] jimis: bzr diff -rbranch:mytrunk [14:45] though --old should do the same thing [14:45] it seems odd that mytrunk would be a subdirectory of the current branch though [14:46] thanks jelmer [14:47] -rbranch:mytrunk immediately reported that there is no such branch [14:47] --old=mytrunk just returned silently, with 0 return value [14:47] jelmer: Indeed mytrunk is a subdirectory of parent dir, but I thought it would find it :-) [15:20] let's say I want to merge 5 commits from another branch, but I want to commit them separately using the same commit messages used in that branch. [15:20] How can I do that? [15:24] jimis: bzr merge will preserve the commit messages [15:27] jelmer: thanks, I understand now. The message that "bzr ci" requests after that, does not erase the separate messages [15:34] (if you merge and commit, it will show in e.g. "bzr qlog" or "bzr log -n0") [16:37] re [16:37] jelmer: Yes, I want to do this "store large files on the server only" thing [16:44] bye [17:39] how can I commit selectively some of the changes I've made? [17:44] jimis: "bzr shelve" the changes you don't want [17:50] jelmer: then the question is how to shelve selectively? :-p [17:50] for example I have one file, changed in two places that I want to commit separately [17:52] jimis: shelve is interactive - it'l prompt you for what chunks to shelve [17:52] hah! I just read the wishlist: http://wiki.bazaar.canonical.com/BzrWishlist [17:52] bzr commit --interactive should run shelf's hunk-selection script. [17:53] jelmer: that will do thanks [17:57] should I submit some bug report? I'd be very interested in interactive commit feature [17:58] jimis: please file a bug about it [17:58] will do, thanks [18:05] was already there: https://bugs.launchpad.net/bzr/+bug/307285 [18:05] Ubuntu bug 307285 in Bazaar "Support interactive commit" [Medium,Confirmed] [18:05] jimis: in that case, please use the "This affects me too" thing [18:07] jelmer: I did, and posted a comment too :-p === Mkaysi is now known as Mkaysi_ === Mkaysi_ is now known as Mkaysi [18:40] hm, the content filter gets a chunk, where's the documentaton of what a chunk is? [18:40] can't find it in the bzrlib documentation [18:40] Anteru: it gets chunks [18:41] Anteru: which are basically strings which contain a part of the file. the size of each chunk varies (it depends on the way the file was stored) [18:41] oh ok, but its for one file? [18:41] in other words, "".join(chunks) will get you the full file content as a string [18:42] hm, that doesn't sound too good, as the return value is chunks, too. Which means I must have the whole file in memory once there if I want to have it in a single file. [18:42] Ok, I'll try using two files first then [18:42] Anteru: that's true - this is a limitation of the current API [18:43] i.e. foo.meta stores the hash, and when retrieved, it produces a foo file with the actual contents [18:43] and one of the harder challenges in adding support in bzr core for large files [18:43] which will require manually touching of foo.meta if you update foo.bin [18:43] you can have a post-commit hook of some sort that does that automatically [18:43] uhm, sorry, a pre-commit hook [18:43] and a pre-commit too which checks [18:44] yeah I wonder if that's not easier than a content filter? [18:44] i.e. have a pre-commit/post-commit hook which checks the file stamps and be done with it? [18:44] you'll need the content filter to create the file on checkout [18:44] pre-commit post-update [18:44] I'm not sure if there is a post-tree-update hook at the moment [18:44] other than the content filter stuff [18:45] but I wouldn't be surprised if there was one [18:45] hmpf [18:45] implementation wise, it seems pretty simple, but getting into actually into bzr is going to be fun :) [18:51] how to run selftest in Ubuntu? it complains module testtools is missing [18:52] aptitude install python-testtools [18:53] mgz: ok, but shouldn't it come with bzr or bzrtools? [18:53] mgz: it'll be a standalone installation now, related as orphaned package [18:55] it's like a build-depend, as you don't actually need it to *use* bazaar, just to run the test suite, it's not in the dependancy list [18:56] ok [18:56] thanks [18:58] mgz: hi [18:58] hey jelmer! [18:58] mgz: did you see my bug report about some of the other failures I was seeing? [18:58] I've just commented on one that looked like it was about jml's latest testtools changes [18:59] your first failure from the mp still mystifies me, it might help with that if you can set BZR_TEST_PDB and go up untill you get to the KnownFailure/_ExpectedFailure instance and have a look at it [18:59] ah, I must've missed that [18:59] mgz: the MP is about some different issues [19:00] it was just this second, you may have not got the mail yet [19:16] mgz: still haven't seen it [19:16] * jelmer looks at ze web [20:23] is there any version of bzr-svn available that works with bzr 2.3.x? the newest i've been able to find only works with 2.2.x [20:37] whitelynx: only snapshots [20:38] aah ok... could you point me at the bzr branch that has the latest? the one used by the bzr-svn-bzr package in AUR seems older than 1.0.4 [20:39] lp:bzr-svn [20:39] cool, thanks [20:41] I get 'bzr: ERROR: Unsupported protocol for url "lp:bzr-svn"' when i try branching that... does that require an LP ID to branch? [20:41] no [20:41] but it does need the launchpad protocol [20:41] s/protocol/plugin/ [20:41] aah, ok... so --no-plugins is the issue [20:47] sweet, it works now! [20:47] jelmer, lifeless: thanks for the help :) === yofel_ is now known as yofel [23:59] Good morning folks.