[00:00] we're pretty good on disconnected, if I may say so myself [00:00] I saw the story about the renames as first class citizens. And usability seems to be a big driver. [00:00] Cool - thanks for the info! Think I'll give it a try. [00:02] excellent. Just ask here and I'm sure someone will give you a hand, though often weekends are a little quieter than the week [00:05] Thanks. [00:10] lifeless: Do you know what triggers display or hiding of progress bars? [00:12] jelmer: yes [00:12] lifeless: bzr-svn's progress bars are sometimes not shown by bzr, even though I know they'r ebeing created [00:13] do they appear eventuall? [00:13] s/ll/lly/ [00:14] yeah [00:14] ok [00:14] there is debounce protection for small operation [00:14] *operations* [00:18] what can happy is that your first operation takes more than 0.1 seconds [00:18] e.g, many seconds [00:18] without doing a nested child bar or other output [00:19] and this results to the user in no visible output until the first operation finishes [00:19] its possible this is what you are triggering [00:20] hmm, I don't think that's the case. afaik there are progress bars for everything now [00:20] *anything significant [00:22] jelmer: can you trigger this behaviour? [00:23] it appears to be somewhat random to me but I've never spent time looking into it once I hit it [00:23] next time it hits, ctrl-\ and you'll be in pdb [00:23] its the only thing that occurs to me [00:24] I suppose its possible that there is a bug in the bar nesting resulting a nested bar not triggering a show after the MIN_PAUSE [00:24] or many extremely small bars never getting past MIN_PAUSE each [00:28] hmm [01:25] wow [01:25] inventory extraction really blows [01:31] .37 seconds each [01:31] makes indexing...slow [01:42] this is the main bzrlib performance pain point in loggerhead too [01:43] make log -v fast and everyone will be happy === lamont` is now known as lamont [01:55] mwhudson: well [01:55] I need paths_for(revs_to_fileids) to be fast [01:55] which is in principle log -v FILENAME [02:07] to get the xml's only: [02:07] 0.014 [02:29] lifeless, so it's the xml parsing that's slow basically? [02:29] its a significant factor [02:29] anyhow, new feature time in a sec [02:29] I'm on make it work [02:38] thats better: [02:38] :!bzr search COPYING [02:38] COPYING in revision 'robertc@robertcollins.net-20080608052044-s7ktacyyyiib8zi4'. Summary: 'No summaries yet.' [02:39] nice and fast: [02:39] Misc/HISTORY in revision 'svn-v3-trunk1:6015fed2-1504-0410-9fe1-9d1591cc4771:python%2Ftrunk:17190'. Summary: 'No summaries yet.' [02:40] lifeless: yes, the inventory object creation seems like it might be a bit too all or nothing atm [02:41] so, 1) make it work completed [02:41] "No summaries yet."? [02:41] right [02:41] if you do a search that hits a commit message you see [02:42] Revision id 'svn-v3-trunk1:6015fed2-1504-0410-9fe1-9d1591cc4771:python%2Ftrunk:15502'. Summary: 'Fix a problem reported by Oleg Broytmann, who complains that very' [02:42] file hits used to show the fileid, revision id pair [02:42] my most recent push makes it show the PATH and revision id [03:07] but adds 0.3 of a second for every revision indexed === mark1 is now known as markh [04:58] lifeless, around? [05:00] yes [05:00] (must diet sometime) [05:01] I'm finally trying to get Loggerhead to use bzr-search [05:01] cool [05:01] and, btw, it's always very nice to go through your code :) [05:02] anyway, I just want revids for loggerhead, and, well, each type of hit seems to handle that differently [05:03] so [05:03] what do you mean 'just want revids' ? [05:03] the results [05:03] as to what revisions ids match that query [05:03] rather than say specific paths ?:) [05:04] btw, did you see: [05:04] 11:39 < lifeless> Misc/HISTORY in revision 'svn-v3-trunk1:6015fed2-1504-0410-9fe1-9d1591cc4771:python%2Ftrunk:17190'. Summary: 'No summaries yet.' [05:04] current tip, its extremely slow to index in this manner, I'm fixing that now. [05:04] yes, I say, and I pulled, and I jumped right into getting this working :) [05:05] you're even giving the UI some lovin' :p [05:05] yah, sexy is a feature :) [05:05] anyhoo [05:05] I guess I'm saying 'why do you want to limit yourself to just the revision_id' [05:06] well, I guess it's because I already have the rest of the information [05:06] do you ? [05:07] so you have file_id, revision_id -> path mappings already available? [05:07] and, the current interface for loggerhead accepts a list of revids [05:07] yeap [05:08] so [05:08] I think there are a couple of ways to tackle this [05:08] one is to make the Hit interface more pluggable [05:08] so loggerehead can have its own hit types [05:09] this would involving making the constructor interface homogeneous [05:10] yes, although from a glance, it seems to be that if all types had revids and fileids (if applicable) available in the same way, well, that would also cover all the use cases similar to LH [05:10] but I may be over-simplifying things [05:11] another is to make a revision key always be available which lh could interrogate [05:11] but I think that that is possibly/likely overconstraining the interface [05:13] anyhow [05:13] what would you prefer [05:13] do a patch to make it be like that, with enough tests that I can't accidentally change it on you [05:13] and I'm cool [05:14] I'd prefer to "do the right thing", although lazyness is pushing towards the easy fix [05:14] I'm looking at how I can make the contructor interface pluggable [05:14] in a non-hackish kind of way [05:14] I don't think the right thing is entirely clear yet [05:14] so I'd rather do the thing-that-is-easiest-for-you [05:16] I wonder if there are many more use cases for different types [05:18] well [05:18] tags [05:18] threads [05:19] phrases [05:19] I'm considering the logics of working tree indexing [05:19] and taking the very core and making it fully generic [05:20] working tree indexing? [05:20] yeah [05:20] wonder where a line of code is in your working tree [05:21] TAGS is kindof primitive [05:21] can't lookup phrases [05:21] (for instance) [05:21] as in the last revision? [05:21] or uncommitted changes too? [05:22] uncommitted [05:22] I've been also thinking that something *really* cool to add on top of this is something like "who worked on this class/method" [05:22] I suppose it would be language-specific [05:22] last revision in already available, just need an interface to search for revision ids directly (and stash them as terms( [05:23] but that would of been very useful for me many times [05:23] it would be interesting indeed [05:23] the way I would approach is is to add a term for the method to a result which is an author [05:23] and that should a fairly straightforward UI in LH [05:23] "click on the class/method in annotate, or search for it" [05:24] possibly author,fileid,revision_id or something [05:24] hm, yes [05:24] but this is an example where revision_id starts to become less relevant as a mandatory piece of info [05:24] right, I can see that [05:25] anyhow, tell me what works best for you and either do it, or I will for you [05:25] but I'd have to whip a new UI in LH, so I can tailor that around what bzr-search gives me [05:25] I'll take a stab at a patch now [05:26] the shortest-possible-path for you is an isinstance check in LH [05:26] 1) make it work [05:26] :) [05:29] actually, yeah, but being able to get revids in a nicer way will probably tend to break less with newer versions [05:30] anyway, yes, I'll make it work in a hackish-sort-of-way, and then go back :) === _W_ is now known as W_ [05:47] aaaaaaaaaand, it works :) [05:48] oh yay [05:48] I think I have a corrupt test data set [05:48] jelmer: ping [05:49] ah no, rich root breaks my optimised code === mwhudson__ is now known as mwhudson [06:04] much better [06:04] did 100 in 0.0627507519722 each [06:04] thats down from 0.37 [06:05] its also chopped 150M of memory usage :) [06:06] what did you do? [06:06] patience kemosabe [06:06] hahah [06:06] * beuno stis back down [06:06] *sits [06:07] well, LH uses bzr-search [06:07] it and rocks [06:07] revno 34 [06:07] mwhudson, ^ [06:07] and you can peek [06:07] beuno: fantastic [06:07] beuno: screen shot! or demo site! [06:07] I would love to show bkor it running :) [06:07] beuno: nice! [06:08] let's see if the router will let me redirect 8080 [06:08] 34 is pushed FWIW [06:09] loggerhead could probably steal that paths_from_ids code [06:09] I don't know if it is faster when looking at a full revision [06:10] but generally we aren't looking at a full revision in the search code [06:10] (even though we need full xml to process things) [06:12] lifeless, http://200.127.6.219:8080/bazaar/bzr_garbage/changes [06:12] search away [06:14] cool [06:14] so, you're not showing FileTextHits any different to revision hits? [06:15] not yet :) [06:15] no worries [06:15] the sql cache got in the way and I lost a bunch of time on that [06:15] (it turns out it doesn't like duplicate hits) [06:16] I also want to show hwo many results there are, highlight it (text if in commit message, file if in file) [06:17] s/hwo/how [06:17] excellent [06:17] well, feel free to send me patches to make bzr-search nicer for you [06:17] but I should probably have some way to index the repos from LH first :p [06:17] I *like* patches [06:18] * lifeless goes hunting a gnome sysadmin [06:18] heh, will do [06:19] I think a limit is the next thing [06:19] to be able to say 'no more than 1000 revs' or whatever [06:20] yes, that's probably sensible [06:20] and, an "order by relevance" [06:20] I'm thinking for LH actually [06:20] well, relevance is actually the hard thing about searching :P [06:21] relevance talks about precision, if you want to read up on IR theory [06:21] yeah, start with something basic, like most amount of hits, prioritize commit messages maybe [06:22] mwhudson: actually, I think ids_to_paths should be on Repository [06:22] mwhudson: but, backward compatibility etc [06:22] something basic will be much better than seemingly random order [06:22] lol small deltas: [06:22] did 100 in 0.00272021055222 each [06:23] beuno: there is no weighting in the core at the moment though [06:23] beuno: so any order is equivalent as far as the system is concerned [06:24] who beuno [06:24] ehm, Martin? :p [06:24] getting your last name right [06:25] to do a demo to Lynne of the web search [06:25] she was asking how the code was going :) [06:25] who Lynne? :p [06:26] erm, my Fiance! :) [06:27] ah! you're enganged. I somehow don't know that [06:27] hehe, enganged :) [06:28] it would be nice to show more details, I get back 1594.2.17 when searching on your name [06:28] but can't figure out why :) [06:28] NEWS? [06:28] oh, no [06:29] could you run 'bzr search Albisetti' on that tree? [06:29] yes, one sec [06:32] no progress bar is shitting me [06:32] time to fix that [06:32] lifeless, bzr search doesn't return that revid [06:32] LH's cache must be screwing with me again [06:32] beuno: so, first bug found :) [06:33] yes, thank you :) [06:34] I'm going to rip out most of the cache next week, so if it is a cache bug, it should be squashed with that [06:36] mmm [06:36] would be nice to be sure of it/be able to show people this :) [06:36] oh, yes, I won't defer it, just that I won't fix the cache if it is [06:37] I'll do some cleanups, like, not hardcode '/home/beuno/bzr_devel/bzr.garbage' into the code, and push it [06:38] sweet [06:40] mwhudson_: welcome back [06:40] python-packs$ time (bzr search socket inet_pton | head -n1) [06:40] Mac/Include/pyconfig.h in revision 'svn-v3-trunk1:6015fed2-1504-0410-9fe1-9d1591cc4771:python%2Ftrunk:22230'. Summary: 'No summaries yet.' [06:40] bzr: broken pipe [06:40] real 0m0.726s [06:40] user 0m0.592s [06:40] sys 0m0.084s [06:41] that's with the inventory tweaks? [06:41] they only affect index time [06:41] search is orthogonal [06:42] ah, diffs didn't tell me *where* that change was :p [06:43] I thought the commit log would :P [06:43] well, if we had a bzr pull --show-log... [06:44] (yes, I did ignore the commit message, it's just a good time to plug in a feature I want) [07:48] ugh, I'm stuck and sleepy. I'm off to bed [07:48] g'night [07:49] gnight [08:21] abentley: abentley@panoramicfeedback.com-20070410141425-86yh4m1cb6wwnwwm seems to have a corrupt inventory [08:21] abentley: oh, nvm, it's a format behaviour I have not accomodated [08:22] You mean the fact that it has a unique root? [08:23] I didn't remember what format 5 looked like [08:23] I've written a custom path-from-xml extractor thats 5 times faster than deserialising to an Inventory [08:24] lifeless: But we're still on format 5... [08:24] on average that is, for processing deltas when you don't need all-paths [08:24] abentley: right, format-5 rich roots were not in my test suite [08:24] I tried format-5 plain, and rich-root-pack in my test suite [08:25] lifeless: that's quite nice. [08:26] It seems to have worked out well. I'd like to fit it into Repository or some such [08:26] its not optimised yet either [08:26] but 5 times faster was sufficient to stop at for now [08:38] I'm hoping that some of what I develop may go into the core packs as faster lookups for common things [08:38] this just came together as an interesting project :) [08:55] with bzr-svn is it possible to work out which svn revision number I've just merged? [08:58] I don't know [09:16] If you find a way, I'd like to know. [09:17] Could tags be used? Does bzr scale well with thousands of tags? [09:22] Peng: I don't think you'd want one tag per Subversion revision. [09:22] johnf: if you have a branch which is just a straight pull of a svn+ssh:// URL, then I think the Bazaar revno == Subversion's. I could be wrong, but that was the impression I got. Let me test that hypothesis some. [09:23] AfC: Why not? [09:25] Peng: signal to noise ratio, of one thing: it would make other uses of tagging a little less palatable to use. [09:25] Yeah.. [09:26] I'm really curious if it would still perform well though. (But not curious enough to test it myself.) [09:28] Hm. bzr[-svn] revno = 219 ; subversion revno = 249 [09:28] Guess not. [09:28] Peng: they don't [09:28] I would check the svn revision for a property [09:28] jelmer may have stashed it there [09:29] AfC: svn revisions are repo-wide; bzr revisions are branch-wide. [09:38] svn should be thrown away...they still have problems after 11 RCs with 1.5.0 [09:44] hmm, I should add text hit summaries before blogging [09:44] * lifeless goes to do that [10:06] aha bzr version-info has it [10:06] revision-id: svn-v3-trunk0:90e61fa5-4541-0410-a685-e5b9dba3c764:trunk:74 [10:06] svn revision 74 === pbor|out is now known as pbor [10:20] man [10:21] crying tiger is _hot_ tonight [10:26] hi, is there a way to only show the commit messages from the 'main-line' with 'bzr log' [10:26] not the messages of the merged changes [10:29] log --line [10:29] or --short I think too === VK5FOSS is now known as Kamping_Kaiser [10:50] thanks lifeless, --short looks good === Mez|DPC is now known as mez === mez is now known as Mez|DPC === Mez|DPC is now known as mez === mez is now known as Mez|DPC [13:27] james_w: ping [13:27] Does anyone know how to commit to an sf.net svn repo with bzr-svn? (or have a link to the relevant documentation?) [13:27] johnf, ping [13:27] hi jelmer [13:27] james_w: debian bug 472543 - does that patch look sensible to you? [13:27] Debian bug 472543 in bzr-builddeb "bzr-builddeb: FTBFS: Unknown command "test-builddeb"" [Serious,Open] http://bugs.debian.org/472543 [13:27] jelmer: pong [13:28] johnf: The subversion revision number is sometimes part of the bazaar revision id [13:28] jelmer: yep eveuntually worked that out and even blogged about it :) [13:28] jelmer: That should have been included in the last upload, maybe I messed up the changelog, let me look. [13:29] james_w: The BTS doesn't contain any claims it's been fixed [13:29] johnf: It's not something that's guaranteed though - if the commit was pushed from bzr, it could be a completely different revno [13:30] rexium: Simply use it like any svn repo with bzr-svn [13:30] jelmer, I am getting authentication issues [13:30] rexium, see the FAQ entry about authentication and http [13:31] jelmer: ah, my mistake, let me close it. [13:31] jelmer: thanks for catching it. [13:32] james_w: Thanks [13:32] * jelmer is chasing down RC bugs at a BSP atm === Mez|DPC is now known as mez === mez is now known as Mez|DPC [16:24] anyone want to review my first attempt at a bzr screencast? [16:24] http://jameswestby.net/scratch/introducing_yourself.ogg [16:27] james_w: you speak a little bit too fast, I think. [16:27] * Leonidas watches it to the end now [16:28] james_w: i think the same - too fast and too long sentences. otherwise it's very nice [16:28] * gour watched till the end [16:29] * Leonidas too [16:30] Yeah, the text in the console is even pretty much readable. [16:30] thanks, I'll try and slow down for my next one. [16:30] james_w: great, but I agree previous comments about speed of talking [16:31] james_w: what do you use for casting? [16:31] heh, recordmydesktop to capture [16:31] james_w: use 'comma' in your speech ;) [16:31] but it gets pretty messy from then to get to the finished product. [16:32] gour: heh, what are they? [16:32] heh, atm i'm doing some video-editing in cinelerra [16:32] I tried that, but I couldn't work it out at all. [16:32] james_w: look for this ',' sign on your keyboard :-) [16:33] * gour thinks video-editing on linux still sucks. hopefully lumiera will fix it, but the project needs help [16:34] james_w: btw, did you see the announcement of the extremadura vcs-pkg meeting? [16:34] jelmer: yeah, I never registered my interest. [16:40] james_w: the podcast is good, the audio could do with some work though [16:40] the music needs to fade in/out before/after you finish talking [16:41] plus your voice could probably do with being a bit loud [16:41] *louder [16:41] I like having a bit of overlap, but I could cut the music a bit further. [16:42] oh, a bit of overlap is fine, it just seems a bit too much overlap, and seems to drown out your voice a bit [16:42] My mic picks up a lot of noise on my laptop, so I can't have it too much louder, but I'll put my new soundcard in my other machine and see if it is any better there. [16:42] thanks for the feedback though. [16:47] james_w: Ah, ok [16:47] james_w: Thanks for merging that patch btw === bigdo2 is now known as bigdog [18:20] Is there a bzr plugin that will help me create debs from a bzr repository? [18:26] bzr-builddeb? [18:27] rockstar: https://launchpad.net/bzr-builddeb [18:27] Awesome, thanks. I knew there was something, but bzr-deb doesn't google well... [18:28] :) [18:28] That should belong to the bazaar project group, methinks [18:55] hi rockstar [18:55] give me a shout if you have any trouble with it. [18:56] james_w, well, I'm not a great packager. Ideally what I would like is a tool that just creates a dummy debian/ folder with default files in it... [18:57] rockstar: dh-make will do that for you [18:58] rockstar: I've added it to the bazaar group, thanks. [18:58] james_w, well, not without needing a .orig.tar.gz first [18:59] rockstar: ah true, though it probably has a "native" package mode that will avoid the need for that. [19:00] rockstar: yup, dh_make -n [19:00] * rockstar wipes egg off his face [19:13] hey guys, i am trying to push via ssh. Isn't the url syntax similar to that of sftp? sftp works, ssh i think is spitting a command not found. [19:14] scorpioxy: you need to have bzr installed on the server to use bzr+ssh:// [19:14] james_w: i think i do. Issuing bzr does the right thing. [19:15] is it installed in the $PATH? [19:16] james_w: yes [19:17] and "ssh server bzr" doesn't complain? [19:18] james_w: umm...what is that? do you mean bzr serve? [19:19] james_w: oh oh...never mind [19:20] james_w: it says command not found...but bzr is in my $PATH...this is a shared host, does it have anything to do with that? [19:22] scorpioxy: is it in a part of the $PATH that is set in .bashrc or similar? [19:22] I've got to go, but I will point you to $BZR_REMOTE_PATH, and suggest that you hang around, because someone else might be able to help. [19:22] james_w: alright, thanks [19:25] jelmer: on the bug tracker, I saw you say something along the lines of: [19:26] "branching schemes are unecessary. Kill, kill, ex-ter-mi-na-te" [19:26] could you elaborate on how they are unecessary? [19:27] ddaa: hi [19:28] ddaa: Yes - they do not need to be part of the revision id [19:28] ddaa: Since we can just follow the history of the path we're branching [19:28] uh [19:29] sorry, dvcs neurons a bit rusty [19:29] the idea would be to replace branching schemes with a somewhat similar thing called 'repository layouts' that is less intrusive [19:29] I'd love to hear about it. [19:29] e.g. just used to determine what a branch is when converting the whole repository, but not when analysing history [19:30] let me think aloud for a bit [19:31] ATM, branching scheme are not used to "guess" merges right? [19:31] no, merges are never guessed - only tracked explicitly using file properties [19:32] The only way you get merge in bzr-svn history is (currently) through revisions created by bzr, svk and eventually svn>=1.5. [19:32] correct [19:34] the thing is, those merges involve other paths in the svn repo [19:34] and that's where branching schemes get involved [19:35] why would that involve branching schemes? We just allow all paths in the repo as branch paths [19:35] sorry, I do not think I'll be able to make much sense, this stuff has just gone too hazy :( [19:36] my concern comes from the fact I needed to create a new branching scheme to hack around a limitation in bzr-svn [19:36] i.e. when pushing a commit, it tried to inspect some branches that I do not have access to, and failed 403 [19:37] It really should not need to look at those branches (such as "/toolkit/trunk") when pushing to my branch (such as "/language/trunk") [19:38] that's the bit that's going to be replaced by repository layouts [19:38] it's calling Repository.has_revision() to figure out which of the revisions it will push are already present somewhere in the repository [19:38] but it was easier for me we write a RestrictedTrunkBranchingScheme that just looks at /foo/trunk, /foo/branches/*, and /foo/tags/* [19:40] jelmer: the current behaviour is meant to allow things like "commit on A, merge A into B and commit B, push B"? [19:40] (where A and B are svn branches) [19:41] the current behaviour is meant to support something like this: [19:42] pushing large branch foo to /branches/foo [19:42] then pushing large branch bar which descends from foo to branches/bar [19:42] the latter would do a copy from /branches/foo to /branches/bar and then push the extra revisions not yet present in foo [19:42] I see. [19:43] anyway, that bit is not going away, it's jsut going to be more flexible [19:43] so it would be possible for the user to specify how hard bzr-svn would have to look for revisions to reuse [19:44] are you actually working on this (or planning to, at least), or is that a "someone would need to siphon out jelmer's brain and take over" kind of plan? [19:45] I'm working on this bit by bit [19:45] The initial work (now done) was adding support for having more than one mapping scheme present in the same version of bzr-svn [19:45] so it's now possible to support revisions created with and without branching schemes without having to use a different version of bzr-svn [19:46] most of the branching-scheme-less mapping also already works and I've done ad-hoc testing of it [19:47] it knocked my socks off when I saw that if I pushed revisions from with my custom branching scheme, and then switched back to the standard trunk1 scheme, it preserved the shape of the ancestry graph. [19:48] jelmer: which kind are you? Motivated by thank you or by complains? [19:50] ddaa: yes [19:50] ddaa: I mean, both [19:50] I can do both :) [19:52] So: Thanks, at work they use svn (thanks god!), now I have something which will allow me to make merges without having to reimplement gnu-arch-era star-merge in scripts... [19:52] also, making bzr-svn to work gave me the motivation to install ubuntu on my office workstation (thanks WUBI) [19:54] (between newline translation on windows native and issues recompiling the python bindings on cygwin, I could not make it work on windows) [19:55]