[00:00] now, if we could only get it to stop crashing... [00:00] mwhudson, wow, pretty big [00:00] are you testing on diff? [00:00] beuno: yeah, definitely a surprise [00:00] morning [00:00] beuno: yes [00:01] mwhudson, interesting... I wonder if that will suffice to make it stop crashing [00:02] mornin' igc [00:02] morning beuno! [00:02] I like the templating layout for zpt much better, seems cleaner [00:05] it's a lot more verbose [00:05] mwhudson, did you update the branch, or as-is? [00:06] so after rendering a 30002 line diff richly with kid loggerhead is using 711m virt, 493m resident [00:07] after rendering the same diff with zpt-templating, it's 375m, 152m [00:07] beuno: i merged trunk into it and fixed the text conflicts [00:08] (oh and 50 s vs 16s) [00:08] heh, very interesting [00:08] yes [00:09] I've been thinking how much we could reduce our memory usage if, instead of looping into a variable all the lines, we somehow start building the HTML as we loop [00:09] moin [00:11] I think it may worth testing, should scale better [00:11] hey lifeless [00:13] mmf, two concurrent requests to that diff page still chew a lot of ram [00:13] well, about twice as much, no real surprise there [00:15] mwhudson, I'm going away for a few hours, but I'll probably work for a few when I come back. Anything specific you want me to look in to? [00:15] beuno: not that springs to mind immediately [00:15] beuno: i'll send you email if anything comes up [00:16] mwhudson, cool. If not, I'll play around with some of the ideas I have and see if we can get a significant memory usage reduction [00:16] cool [00:16] later! [00:17] lifeless: hi! [00:17] lifeless: feeling any better? [00:24] the fever seems broken [00:24] so yes, but not 'well' [00:28] spiv: ping [00:30] lifeless: pong [00:30] I'm just pushing up a new snapshot of versioned_files [00:31] poolie said you were interested in network edition of same [00:32] Yeah, definitely. I want to use the new streaming stuff. [00:33] there are two layers that need doing [00:33] a 'fetch' level layer, which replaces repository.get_data_stream [00:34] I have some cheap hacks that implement a couple of pack specific RPCs (to e.g. autopack) that get significant improvements, but the long term solution is to use your stuff. [00:35] and a text/inventory/revision/signature level layer, for use by e.g. diff [00:35] there are some open questions [00:35] tree building for instance commonly needs all the texts for one revision [00:36] so it could use the repository level api to minimise the request size [00:37] OTOH when we have even a couple of local commits and the rest is on the network we won't need *all* the texts for one revision to cross the wire, but sending a whole inventory of .oO over the wire to ask for its tree probably outweighs the retransmission cost [00:38] and that doesn't even address the conceptual layering that would be needed to do that [00:38] Using saved location: sftp://rookery/~/public_html/baz2.0/versioned_files [00:38] Now on revision 3375. [00:38] 'punt' [00:38] :) [00:38] Thanks! [00:39] I'm going to try to work today [00:39] there is much to do [00:39] note that that branch is currently in the middle of eliminating get_inventory_weave [00:40] specifically bundles are bust while I adapt the multiparent code [00:40] Ok [00:41] hmm, how best to collaborate here; I think if you have a look at the code [00:41] and get a feel for it [00:42] then we can chat on IRC? [00:44] I think so. I'm currently focusing on getting my various push acceleration work so far to the list. Digging into using your branch would be next on the list, though. [00:45] cool [00:46] also for the network apis we may likely want to use the single key space aaron advocates, because for the smart server there is no arbitrary split between indices (in some regards ideally no index at all) [00:46] That makes sense to me. [00:53] poolie: talking is still unpleasant; if you want to catch up IRC would be massively preferable [01:11] Jc2k: ping [01:14] beuno, sorry about that, the server I'm running it on appears to've been assigned a new IP again === cammoblammo` is now known as cammoblammo === kiko-afk is now known as kiko [01:42] poolie: ping in fact === mw is now known as mw|out [01:57] abentley: can you check BB please? It appears to be down again [01:59] igc: Should be back in a minute. [01:59] thanks [01:59] lifeless: sorry you were minimized :) [02:00] Who is it that updates the downloads page on the bzr site? [02:00] Pilky: whoever is making the release; for the last one that was jam [02:00] poolie: must be offscreen rendering, I hadn't noticed [02:00] lifeless: igc was wanting to know if there was anything he can do to help with merging stacks [02:00] ok [02:01] poolie: updating it to bzr.dev would be good [02:01] in that case, jam the OS X 10.5 installer link on the download page still points to 1.4 not 1.5 [02:01] let me make sure its recorded and pushed [02:03] ssh: connect to host bazaar.launchpad.net port 22: Connection refused [02:03] bzr: ERROR: Unable to connect to SSH host bazaar.launchpad.net; EOF during negotiation [02:03] ? [02:04] jml ^^ [02:04] oh yeah, downtime [02:04] oh [02:04] 2 hour window [02:04] I would so love to fix that [02:45] igc: poolie: sftp://rookery/~/public_html/baz2.0/shallow-branch updated === kiko is now known as kiko-zzz [02:46] thanks lifeless. I'll take a look after I finish reviewing jam's patches [02:47] it needs bzr.dev merged into the bottom thread and propogated up [02:47] yup - shall do [03:08] jam, in your mail about merge annotations (if you're still here) you say "merge node gets annotate (current)" - do you mean that's what we do now? [03:08] iirc it is [03:08] I believe thats what it got updated to do [03:08] I would really like to see multi-ancestor annotations [03:09] but thats not trivial :) [03:11] * RAOF queues up a strange bzr issue for later. [03:14] shoot [03:14] no need to queue :P [03:16] Actually, it's queuing on _me_. [03:16] But public acknoledgement makes it easier for me to remember :) [03:18] RAOF.pop() [03:27] poolie: (current) means it is what we do now, yes [03:27] the 'last modified' for that line is set to the merge revision id [03:29] so [03:29] ok [03:29] replying now [03:29] I seem to have a bug with left_matching_block stuff [03:29] btw is there a bug number for the current failure? [03:29] I'd like to run this past someone :P [03:29] a left_matching_blocks of: [03:30] [(0,0,1), (1,1,0)] [03:30] means from 0, to 0, for 1 line [03:30] and from 1, to 1, for 0 lines [03:30] yes? [03:39] from memory yes [03:39] that seems odd [03:40] i vaguely recall there is always a match anchored at the end even if it's 0 lines [03:40] ^^ this may be completely wrong [03:40] lifeless: ^^ [03:40] this is for ['line'] to ['line'] [03:40] which is why it smells excruciatingly odd to me [03:41] its being generated from multiparentversionedfile though, which is unfamiliar code to me [03:41] and I've got cold sweats right now; I think I'm going to call it in a few minutes [03:43] jam, replied to your annotation mail, hope that helps [03:45] igc, speaking of faster local branch, it would be nice to show better progress there too [03:46] poolie: progress as in UI, or progress as in 'faster before 1.6 ships'? [03:47] heh [03:47] :) [03:47] i meant progress bars; it's just wishlist though [03:47] my patch cuts the phases from 2 to 1 so ui progress is better IMO [03:47] oh nice [03:50] * igc lunch [04:05] * poolie lunches [04:25] I'm calling it a day [04:25] fading too much [04:26] I have at least tracked down the failure [04:26] we were generating a non-normalised line fulltext [04:26] due to a bug in annotation merging [04:26] it may well be the cause of some reports in launchpad [04:39] get well soon lifeless [04:45] lifeless: see ya [04:54] mwhudson, still around? [04:55] beuno: yep [04:56] so, I'm about to jump into code, but I wanted to run by you what I'm getting into [04:57] in the revision view (diff), we might have multiple files [04:57] and now, we show the diff for all of them [04:57] so, I'd like to try the following: [04:58] by default, just show what files where changed, and add a "View Diff" to each [04:58] i'm writing a long-ish mail to the bazaar list currently [04:59] we would get the diff through ajax to make it nicer for the user and the server, and, we generate *just* the diff HTML on the server, most probably skipping the templating engine all together [04:59] as it should be fairly straightforward to generate [04:59] on LP specifically, that should reduce resource consumption a *lot* [05:00] as you wouldn't get every single diff every time, even when the user doesn't want it [05:01] yeah, i think that's probably a sensible idea [05:02] and, if we find out that getting the same diff multiple times is expensive, we can just cache the diff itself [05:02] which will be fairly space efficient [05:03] beuno: i sent the mail, see what you think of it [05:03] * beuno reads [05:07] mwhudson, wonderful. Now we take a 2 week vacation and come back with that solved, right? :) [05:07] beuno: heh, i doubt it [05:08] mwhudson, sums it up very well, can't think of anything I'd add [05:08] cool [05:08] now i think i'm going to work on bringing zpt-templating up to date [05:08] ok, so I'm going to dive into what I explained above, on the zpt branch, which seems like the most promising [05:09] ah, perfect [05:09] can you push whatever you have so far? [05:11] ok, it's not much [05:12] it's just so I diverge as little as possible, not expecting anything new other than resolved conflicts :) [05:19] beuno: ~mwhudson/loggerhead/zpt-templating/ is up to date with what i have [05:23] aargh, i just want to stab things :/ [05:24] * RAOF .pop() [05:24] So, sometime in the last couple of weeks I've lost the ability to push a bzr tree to my vfat-formatted USB stick. [05:25] bzr push will error out with errno 1: operation not permitted. [05:26] I'll pastebin the actual error, but it looks like somewhere around bzr 1.4/1.5 this broke; it previously worked. [05:29] server seems to have died :/ [05:46] * igc pick up kids [05:48] mwhudson, I can't seem to get the zpt branch working, I've narrowed it down to: http://paste.ubuntu.com/15471/ [05:48] any ideas? === beuno_ is now known as beuno [05:49] beuno: ah right, you need zope.pagetemplate installed [05:50] ah, it's in zope-replacesupport [05:50] thanks :) [05:50] what an obvious package name [05:51] yes, I thought so [05:52] as in this package? http://packages.ubuntu.com/hardy/zope-replacesupport [05:52] and the fact that it depends on zope 2.10 makes me think something is going to go wrong [05:52] mwhudson, yeap [05:52] i find this a little hard to believe [05:52] that's what apt-cache tells me [05:52] ah, it lied to me [05:53] beuno@beuno-laptop:~/bzr_devel/loggerhead.zpt$ apt-cache search pagetemplate [05:53] zope-replacesupport - Add search and replace functionality to TTW Zope objects [05:53] i think my money is more on zope3 [05:54] I already have that :) [05:54] then you should have pagetemplates [05:55] ah, perhaps you're running with python 2.5 ? [05:56] yes, turbogears explodes with 2.5 [05:56] er, 2.4 [05:56] that is definitely not zpt [05:56] beuno: not for me it doesn't [05:56] it only works for 2.4? [05:57] zope3 only works for 2.4 yes [05:57] i suspect page templates would be fine with 2.5 [05:58] gah, why is everything horrible [05:58] easy_install-2.5ing zope.tal should work [05:59] mwhudson: why is everything horrible? 'zope' [06:00] lifeless: it has help [06:00] harsh! [06:00] whats the relationship between kid and genshi? [06:00] mwhudson, well, hardy seems to come with turbokid 1.0.1, and loggerhead requieres 1.0.4 [06:00] pkg_resources.VersionConflict: (TurboKid 1.0.1 (/var/lib/python-support/python2.4), Requirement.parse('TurboKid>=1.0.4')) [06:01] RAOF: please do pastebin the error :P [06:01] on the other hand, it does work with 2.5 :( [06:01] lifeless: they're more or less equivalent, they're both templating engines [06:01] beuno: eh? it all works ok for me [06:02] mwhudson: at EP some guy whose name escapes me right now - the one that did the input locale which uses brackets not whitespace - was saying that kid was terribly slow and genshii rocked [06:02] meh, I'll easy_install [06:03] lifeless: i went to the genshi talk at ep, but i think i was still drunk [06:03] mwhudson: this was over lunch [06:03] SteveA was there [06:03] we had pizza [06:03] genshi is much nicer than kid, but it doesn't seem to be much faster or less of a pig resource-wise [06:04] lifeless: Sorry; http://pastebin.com/m69b6b5ea [06:04] mwhudson: ah, you've tested I presume ? [06:04] whereas my zpt branch is 2-3x faster and uses 2-3x less ram [06:04] RAOF: note the temp file name [06:04] jml has pointed me at https://bugs.edge.launchpad.net/bzr/+bug/59424 , which isn't my bug but may well be the cause; this seems to be a regression from 1.3 [06:04] Launchpad bug 59424 in bzr "repository_implementations.test_vfat* is broken." [Medium,Confirmed] [06:05] it's possible the genshi tests were flawed somehow, we haven't spent very long on them [06:05] RAOF: one possible cause the multiple dots [06:05] mwhudson, sorry to be such a pain, but I can't seem to get it working... :( Am I missing something else? http://paste.ubuntu.com/15473/ [06:05] RAOF: but essentially we need to trap and diagnose the bug [06:06] beuno: oh yeah, only the annotate and revision views really work now :) [06:06] beuno: i'm fixing [06:06] RAOF: run the command again with -Derror [06:06] hahah [06:06] alright [06:06] RAOF: that'll give you a full traceback [06:06] RAOF: the bug is orthogonal [06:06] RAOF: you are not even creating your repo [06:06] * beuno breaths [06:07] RAOF: have you tried an old bzr to see if it is better [06:07] lifeless: It works in 1.3 [06:07] lifeless: I'm not sure whether it works in 1.4, but this is 1.5 [06:07] RAOF: if you have the time, it would be great if you could bisect down the mainline of bzr.dev to find the death point [06:08] mwhudson, cool, works, thanks. I am a bit worried that it doesn't work with python 2.5 by default though (although the speed difference justifies it) [06:08] RAOF: I wager it will be a change in tmp file usage, mode management, or object creation [06:08] lifeless: Yeah, that's what I was thinking. I don't have the time right now, but I'll check out the bzr-bisect plugin. :) [06:09] RAOF: the plugin just does the math :P. all you need to do is: bzr branch lp:bzr foo; foo/bzr push [06:09] beuno: it does, just not with ubuntu's zope3 package [06:09] bzr revert -r -400 foo [06:09] foo/bzr push [06:09] beuno: yeah, that part hadn't occurred to me [06:09] bzr revert -r -200 foo [06:09] etc [06:09] bob2: oh good [06:09] well, how would we package it then? [06:10] mwhudson: afaik the only part of zope3 that doesn't run under 2.5 is some zope2 thing [06:10] lifeless: I know. But I liked git's automated bisect (essentially the only git feature I liked). [06:10] beuno: later [06:10] :) [06:10] let's make it work first [06:10] this sounds like a job for buildout [06:11] mwhudson, right, fair enough. Just thought I'd throw it out there. Back to work. [06:11] RAOF: AIUI bzr-bisect is feature equivalent; I haven't used it enough to say [06:11] RAOF: I have my own ideas on what would be nice; time FTL [06:12] lifeless: I'm not sure what 'time FTL' expands to :) [06:12] abentley: bb is down.... [06:13] poolie: working on it... [06:14] RAOF: "but I am very busy" [06:14] load averag is 7.42 [06:14] bob2: Thanks. For some reason I was translating as 'Faster Than Light (travel)' :) [06:33] Can anyone point me toward where to read about how to implement a new transport? [06:33] Or which source files to start reading from? [06:37] Slant: bzrlib/transport/__init__.py [06:37] And other files in that directory. [06:43] poolie: back. === jamesh_ is now known as jamesh [06:49] alright, time to get some sleep [06:50] mwhudson, hopfully I'll have a proof-of-concept branch by tomorrow, based off zpt [06:50] beuno: cool [06:50] cya tomorrow :) [06:57] how do I undo a commit? [06:58] sili: for the last commit, you can do: bzr uncommit [07:00] if you want to undo an earlier commit, but keep everything that came after that, you should do a "reverse merge": bzr merge -r 10..9 (if I rembember the syntax correctly) [07:00] ah just found it. thanks [07:00] cool [07:03] beuno: my zpt-templating branch has all views seeming to work again now [07:03] beuno: some have formatting issues [07:53] jelmer: bzr-launchpad??? [07:53] jelmer: can you explain to me, wtf? [07:53] (why the split, I mean) [08:12] jelmer: pong [08:12] jelmer: i'll be at work in an hour, and so available for about 7 hours.. [08:53] * igc dinner [08:59] jelmer: oooh, nm.. just saw my email :D === AfC is now known as AfC|wine === weigon_ is now known as weigon [11:18] anyone can suggest something in favor of bzr for http://article.gmane.org/gmane.emacs.dvc.devel/2164 ? [11:23] "Merges look terrible in the bzr log output." -- I, on the other hand, find them beautiful and informative [11:28] gour: and the branch nick defaults to the dirname of a branch, but you can set it with `bzr nick nickname`. [11:28] gour: Like Aaron does with "Aaron's mergeable stuff" for example [11:29] heh, it would be nice if someone can post something. DVC uses bzr atm, but there is quite a bashing of bzr in favour of git [11:30] i replied recently showing how tailor can be used for git <--> bzr [11:30] I suspect that bashing is not going to stop just because someone injects facts or opposing opinions [11:31] well, facts are facts [11:32] in advocacy, facts are always debatable :) [11:32] I just wonder how did he manage to get that 'git log' output, if it's a 1:1 copy of the bzr branch === yacc_ is now known as yacc [12:21] re: git/DVC post i) Doesn't --short cut indented merge info out of the log ii) alias merge to merge --pull if you want that behaviour iii) nick is useful information if it's used right - an informative name tells you what the branch was being used for [12:21] But I think all that's already been said [13:17] * Mez wonders how badly he can kill bzr running bzr gannotate on a file in a svn repository [13:19] Mez, you can't, since annotate doesn't work on svn stuff atm :-) [13:20] jelmer - well - it TRIES to do stuff.. [13:20] and then finally fails with a "database locked" [13:22] 2nd try just finishes up with a locked X window ;) [13:22] Exception exceptions.KeyboardInterrupt in >> ignored [13:23] jelmer, I must say - checking out a svn branch with bzr is a helluva lot slower than with svn [13:23] Mez: It fetches all history [13:24] Mez: "bzr co --lightweight" is more like a real svn checkout [13:24] will that make me something I can work with though (and branch) [13:25] yes, but it will contact the remote server a lot more often [13:25] (it's similar to a bzr lightweight checkout) [13:39] dato: to allow people who don't want to have the launchpad plugin to not have it installed [13:40] dato: it also makes sense since the launchpad plugin is totally separate from the rest of the code [14:05] jelmer: I think it's totally a waste an an overkill [14:16] dato: What, fetching full history from SVN repo? [14:16] It would be nice to see history horizons make an appearance. [14:24] is there a bisect command like git [14:24] https://code.launchpad.net/bzr-bisect [14:27] awilkins: uh, no. creating a bzr-launchpad package in Debian. [14:28] debian folks dislike launchpad [14:30] debian folks dislike * [14:32] jelmer: If people really object, they can remove the python files themselves. Have you encountered people who object so strongly that it warrants the (archive) overhead of an extra package? [14:33] jelmer: if I understand you correctly, it is never the case of wanting to install the plugin extra, but only to remove it. [14:34] thats a bit ewww, tho [14:35] it would be trivial to have the bzr source package build a bzr and bzr-launchpad, not really a hardship at all.. [14:36] dato, http://lists.alioth.debian.org/pipermail/pkg-bazaar-maint/2008-May/000843.html [14:36] dato: perhaps I'm underestimating the overhead on the archive [14:36] It just seems cleaner to give plugins their own package [14:40] Jc2k: we (Debian) need less packages, not more. [14:43] Jc2k, the bug you reported that affected a bunch of gnome svn branches should be fixed now, btw [14:43] jelmer: i saw, rerunning my script now - all perfect so far [14:43] jelmer: if i want to keep a bunch of branches up to date, is there anything cheaper than keep running svn-import ? [14:44] Jc2k: bzr multi-pull [14:44] that will update the existing branches [14:44] i was looking at multi-pull, but that won't take account of new branches [14:44] but it will not fetch new branches [14:44] right [14:45] other than that, svn-import is the cheapest available - it should be pretty quick when branches are up to date [14:45] jelmer: its quite slow against evolution [14:45] about an hour.. [14:45] wow.. [14:45] 23,000 revisions [14:45] quite a lot of branches too [14:45] it imports 23000 revisions in an hour? [14:46] Jc2k, it's probably somthing funny in the repository layout [14:46] Pieter: no, it already has 23000 revisions [14:46] oh.. then I'm less surprised :) [14:47] Pieter: about 1,000 revisions per 5 minutes iirc [14:47] jelmer: any suggestions? [14:47] jelmer: you can grab my evolution repo clone if you want [14:48] actually, i dont know how you clone a repo in bazaar.. [14:48] there needs to be a bzr import-bzr :D... [14:48] Jc2k: Any chance you can run with -Dtransport and send me the output in ~/.bzr.log ? [14:48] Jc2k, there's no easy way but Odd_Bloke worked on a plugin to allow cloning of repositories [14:49] cool [14:49] dato: Anyway, if everybody else objects to this I'd be happy to pull it out again [14:49] dato: but nobody yelled on the list, that's why I went ahead and put it in [14:49] jelmer: well, I follow -commits more closely than -maint :P [14:50] jelmer: anyway, splits in Debian are normally done because of (a) size, to allow users to unbloat their systems, (b) dependencies, same "bloat" argument [14:51] jelmer: but I'm slowly getting uninvolved, so I guess it's your call... [14:52] dato: Ok, I'll see if I cna revert it [14:52] dato: Getting uninvolved with Debian or just Debian+Bazaar? [14:53] Debian+Bazaar [14:53] using that I am, *cough*, using mostly git nowadays... [14:54] well, same here but that's not by choice :-) [14:54] * dato goes for lunch [14:54] yay spain :P [14:55] lol, I know some folks here in .nl that eat dinner at 5... === kiko-zzz is now known as kiko [15:34] if I want to find whether two branches have diverged is this correct?: [15:34] graph = branch1.repository.get_graph() [15:35] diverged = len(graph.heads([branch1.last_revision(), branch2.last_revision()])) > 1 [15:36] also, do I need to do a fetch beforehand if they don't share all ancestory? [15:37] in keeping with the whole english vs english thing, bzr vizualize should work. :-) [15:43] james_w: If branch1's repository may not have the relevant revisions for branch2, you should do branch1.repository.get_graph(branch2.repository). [15:43] The branch1 repository will be preferred. [15:43] ah great, thanks. [15:44] heads > 1 is a valid test for divergence. [15:47] how can I check out a previous revision? [15:49] if you want a separate working area then "branch -r" will get you that, if you just want to look at one then "revert -r", if you want to move to an old one and discard the revisions after then "pull --overwrite -r ." or "bzr uncommit -r && bzr revert" will do it. [15:53] right, revert -r it is [15:53] thanks [16:32] Is there any way to convert a Perforce repository to bzr, preserving history? [16:34] http://bazaar-vcs.org/BzrFastImport/FrontEnds [16:34] or if you want to be particularly elaborate, there's git importer and git has p4 importer ;) [16:34] are there any git importers other than through the fastimport frontend? [16:35] (oh, actually, the p4 importer on the frontend page is git's one; it's nice idea to reimplement the fast-import interface in bzr) [16:35] bazaar has a fast-import [16:35] but it's kinda crashy and eats tons of memory [16:36] So the only option is to go from p4 to git and then to bzr? [16:37] no, you can use the p4 fast-export and bzr fast-import [16:37] it might work [16:40] doesn't look like the git-p4 has a standalone exporter === mw|out is now known as mw [16:49] heya [17:08] I could just go for a slush puppie === cprov is now known as cprov-out [18:09] Hi [18:09] I got a branch with several revisions [18:09] I now discovered that, by magic, one important file got lost [18:09] How would I best go about restoring that single file from a previous revision? [18:16] dennda: bzr revert -r XXX filename [18:16]