/srv/irclogs.ubuntu.com/2008/09/20/#bzr.txt

=== mw is now known as mw|out
toojaysHi bzr people! Can somebody please tell me how to do a partial commit in bzr?02:38
toojaysIf I were using git, I would do "git add -p".02:39
jelmertoojays, just specify the path02:39
toojaysCan bzr do something like this?02:39
jelmertoojays, bzr commit <path>02:39
toojaysjelmer: The changes I want to commit separately are in the same file.02:39
toojaysOne set of changes I will want to merge to another branch, one set I will not.02:40
toojaysBut they are in the same file.02:40
jelmerah, you'd like bzr to prompt you about each chunk of changes?02:40
toojaysYeah, I guess that's the only UI I know of to do what I want.02:41
fullermdI think somebody wrote a plugin...   bzr record or something.02:41
fullermdYou could also use shelve to set aside the bits you don't want, commit, then bring them back.02:42
toojaysShelve . . . yeah, that sounds like a workaround I could use.02:42
toojaysThis makes me think . . . is there a page for people coming to bzr from other dvcs systems, which compares how to do similar operations?02:43
toojaysI have been using git a lot at work, but am using bzr at home . . . so I'm pretty fluent with git, but have only been doing "easy" stuff with bzr so far.02:43
fullermdWell...  the people wanting that particular thing are usually the darcs people   ;)02:44
fullermdHence 'record'.02:44
fullermdAh, 'interactive'.  Renamed post-loom, I guess.02:45
fullermdhttps://launchpad.net/bzr-interactive02:45
toojaysfullermd: thanks, that sounds like it will do what I want.02:47
markhjelmer: bzr-svn is still reporting as '0.4.13dev0' - is that likely to change for a 1.7 release?04:23
=== sdboyAAAAAR is now known as sdboyer
planetcallshi. I am pretty new to bazar. Mini tuts on website show command line way of doing things. Do we have any UI or shell hooks like SVN for bazar ?05:31
mlhfor windows or linux?05:37
mlhthere's a few in progress05:37
mlhcheck the website05:37
mlhthere's TortoiseBZR in progress, specifically for windows05:38
markhplanetcalls: if you are on windows, try the 1.7rc2 build rather than 1.6 - the GUI stuff is much more polished there.05:38
mlhthere you go -- straight from the source :-)05:39
markh:)05:39
planetcallsmarkh, thanks. From the examples and documents it seems as if bazar is having only the command based binaries.05:40
markhplanetcalls: its very much a current work in progress, so the examples and docs lag a little still05:40
markhplanetcalls: but that also means it is so new that there are issues and things not implemented. It is no where near as mature as TortoiseSVN yet.05:41
markhits functional though05:41
planetcallsyea that is obvious. I am not going to give up Tortoise anytime soon but wanted to give distributed vcs a try05:42
planetcallsjust found TortoiseBzr....sounds it would be easier to adopt bazar in near future05:44
markhplanetcalls: yeah, the most recent TortoiseBzr is in the bzr 1.7rc2 binary on launchpad05:50
markhplanetcalls: but it also includes a plugin which allows you to use a GUI from the command line - eg, "bzr qci" will show a graphical version of "bzr ci" - ie, to do a checkin05:51
planetcallsok thanks......now downloading05:52
_zouafter command "bzr update", I got following messages:05:56
_zouUnable to obtain lock file:///cygdrive/d/work/OpenSrc/gnash_bazaar/trunk/.bzr/branch/lock05:56
_zoulocked 37 hours, 51 minutes ago05:57
_zouWill continue to try until 13:00:3905:57
_zouanyone knows how to solve this?05:58
markh_zou: 'bzr break-lock'06:01
markh_zou: but you should probably also upgrade your bzr to a non-cygwin one - one of the 1.6 or 1.7 binaries would be good.06:02
_zouthanks, trying now.06:02
_zoudoes it work on windows?06:02
_zou'bzr break-lock' works, update again now.06:03
markh_zou: yeah, 1.6 and 1.7 builds for windows are looking good.06:06
markhif that is what you were asking :)06:06
_zouthanks, I'll try it later.06:07
_zou'bzr update'06:08
_zounow it prints '/ 0/0', and no progress.06:09
_zouis there anyway that I can check if the connection is successful?06:10
_zouconnection to bzr server.06:10
_zouno timeout message, no error message, 'bzr update' just stalled.06:12
markh_zou: it can sometimes appear to do that.  Check .bzr.log and you might see activity06:19
markhin your "my documents" folder06:19
markhbut let it go - the problem is in the progress reporting rather than anything more serious06:20
markhI actually opened a bug on that - it mentions the 'branch' command but I think its the same - bug 26461506:21
ubottuLaunchpad bug 264615 in bzr "progress bar for pull does nothing until near completion" [Medium,Triaged] https://launchpad.net/bugs/26461506:21
markh(oops - mentions 'pull' :)06:22
_zoumarkh: yes, I found the file under "my documents", but no message for today.06:23
_zouThere are some old messages.06:23
markhhrmph06:23
markh'up -v' might help - but I'd let it go in the hope it actually working06:24
markhobviously not forever, but many minutes isn't unusual if there are many revisions to pull.06:24
_zouI wait a whole morning last time, still no progress or unfinished. Then I used "ctrl + C" to break the command. That's why I need the 'bzr break-lock' command today.06:33
_zouyes, there are many revisions.06:33
_zouI haven't update for a long time.06:33
markhhrm - 'up -v' might give some insight, but it might not :(  if its connecting via ssh it might be strangeness related to that somehow06:37
=== beaumonta is now known as abeaumont
_zoumarkh: "bzr up -v" also printed "- 0/0" and stalled here.  Thanks anyway, might be a network problem here.08:33
Leonidashi08:49
Spazhello08:50
Leonidaschich change is needed so that I get a file called 'TREE_ROOT' in iter_changes?08:50
Leonidass/chich/which/08:50
LeonidasI see that the empty path has the id TREE_ROOT, but how to make a change to the empty path that is tracked using bzr?08:56
mwhudsonyou don't, i think08:58
Leonidasmwhudson: so why bzr.dev has such changes?09:00
mwhudsonmaybe i think wrong then :)09:00
mwhudsonbut bzr.dev has a long and rather more ... eccentric ... history than most branches09:00
mwhudson(like having a revision with revid "A")09:00
mwhudsonLeonidas: i guess you can look at the revisions in bzr.dev that appear to modify TREE_ROOT, right?09:01
Leonidasmwhudson: I currently don't know how to do it, but hmm, that is probably the way to go.09:02
mwhudsonthere is a way to find "all revisions touching a file"09:03
mwhudsoni forget what it is though, i guess read the log code09:03
Leonidasmwhudson: the point is, if I can process bzr.dev properly, I shouldn't have any problem with bzr repos in general :)09:03
mwhudsonLeonidas: ah :)09:03
* Leonidas checks the cmd_log09:04
mwhudsongrunk, i always forget how horrible the log code is09:08
Leonidasshow_log(bzrdev, lf, specific_fileid='TREE_ROOT') doesn't show anything. *sigh*09:14
mwhudsonyeah, there seem to be inconsistencies around09:15
Leonidasthat wouldn't be a big problem, if my stack would be correct, but somehow I only get one frame in pdb.09:17
mwhudsonyay generators!09:18
LarstiQLeonidas: there is tree.set_root_id()10:41
LarstiQLeonidas: I'm not entirely sure what you're asking about10:42
LarstiQLeonidas: and things like bzrlib.transform.adjust_root_path10:43
LarstiQLeonidas: take care with the transform code though10:43
=== Daviey_ is now known as Daviey
=== beaumonta is now known as abeaumont
clementeThe preferred way to convert from git to bzr is still fastimport, isn't it?11:36
Odd_Blokeclemente: I believe so, yes.11:49
Odd_Blokebzr-git might work, but I don't know.11:49
Pieteror git-bzr, but that just uses fastimport11:52
clementeok. Apparently development for bzr-fastimport is stuck since 2 months :-(11:57
Odd_Blokeclemente: Why's that a problem?  It works.12:03
clementeNot for me; I have bug 23836512:04
ubottuLaunchpad bug 238365 in bzr-fastimport "Symbolic links to files with names in UTF-8 (Unicode)" [Undecided,New] https://launchpad.net/bugs/23836512:04
clementeI'd like to fix it but I don't know where to start12:04
Pieter    ie.symlink_target = data.encode('utf8')12:06
Pieterjust change that to ie.symlink_target = data12:06
Pieteror so12:06
clementeThen all data will be stored in binary, even if it contains ¬utf-8 or even invalid utf-8 sequences12:09
clementeI don't know if that is allowed in bzr; but I think that git allows git12:09
clemente*it :-)12:09
Pietera symlink is a symlink, it doesn't have any encoding specified.. I don't know why you'd try to convert it to utf-8 anyway12:10
clementeWhy does data.encode('utf8') fail if data is aleardy utf8? Is it trying to a double conversion?12:14
LarstiQwithout looking at the code, I'd say that is because it's data supplied from the user environment12:14
* LarstiQ looks at the code12:14
clementeMmm... if fastimport leaves the data as it is, then bazaar complains later:12:19
clemente  File "/w/bzr/bzr.dev-oficial/bzrlib/dirstate.py", line 1862, in _inv_entry_to_details12:19
clementeUnicodeEncodeError: 'ascii' codec can't encode character u'\xe9' in position 1: ordinal not in range(128)12:19
=== beaumonta is now known as abeaumont
clementeAnd the problem is really in Bazaar: bzrlib/dirstate.py, 1860:12:28
clemente        elif kind == 'symlink':12:28
clemente            # We don't support non-ascii targets for symlinks yet.12:28
clemente            fingerprint = str(inv_entry.symlink_target or '')12:28
clementeI will report a new bug on that and link the old similar ones to it12:33
clementebug 27244412:44
ubottuLaunchpad bug 272444 in bzr "Support symlinks to non-ascii file names" [Undecided,New] https://launchpad.net/bugs/27244412:44
clementePieter: bzr refuses to store just any string in the original character encoding, and tries to convert everything to Unicode instead12:47
clementeFrom bug 59968: „ This root cause of this problem is that we store this data in XML, and XML1.0 doesn't permit most ASCII control codes. “12:48
ubottuLaunchpad bug 59968 in bzr "add should refuse bad filenames" [Medium,Confirmed] https://launchpad.net/bugs/5996812:48
clementeSo ie.symlink_target = data would be wrong and   ie.symlink_target = data.encode('utf8') better12:48
Pietermessing up filenames just because your backend doesn't support the real data sounds wrong12:48
clementeThere are lots of Unicode bugs in Bazaar...13:01
clementeLarstiQ: you said that it did data.encode('utf8') because it comes from the user. Would it be better to „convert it to utf8 only if it is not already utf8“?13:07
LarstiQclemente: eh no, you can only encode unicode, encoding an already encoded string doesn't make sense13:08
clementeOr „try to convert it to utf8, but if it fails with an exception, leave the data intact“13:08
Pieterclemente: that last part already happens somewhere else in fast-import13:08
LarstiQclemente: so if it's utf8, and then you try to encode utf8, something is wrong13:09
LarstiQPieter: it's not messing up filenames, it's being able to reproduce them13:09
LarstiQwhich, without extra provisions, clashes with being able to store arbitrary binary streams13:10
Pietercan't you just base64 them?13:10
clementeLarstiQ: so a try+catch around that conversion would be needed to prevent the wrong conversion13:10
clementetry+expect actually13:11
clementeexcept !13:12
LarstiQPieter: that's not the point, Linux treats its filesystem paths as binary data, others don't13:12
LarstiQPieter: so you need to know what encoding a filename is in, or you are not able to check it out everywhere13:12
LarstiQclemente: no13:12
LarstiQclemente: in general, the entire pipeline should be: data from environment -> decode into unicode, live in unicode in the entire system -> encode on the way out13:13
LarstiQclemente: there should be no mixing of encoded and unencoded data within13:13
LarstiQPieter: now, some people don't care that you can't check it out, they just want to store a binary stream and tough luck if it fais somewhere else13:14
LarstiQPieter: that should be possible13:14
Pieteryeah.. that seems the right way to go13:14
LarstiQwell13:15
LarstiQit gives rise to working trees that only work in certain corner cases, which is a bad thing13:15
LarstiQPieter: the root cause here being that the common denominator doesn't allow that13:15
clementeBut both modes are incompatible13:16
LarstiQclemente: bug 59968 shouldn't be relevant anymore13:17
ubottuLaunchpad bug 59968 in bzr "add should refuse bad filenames" [Medium,Confirmed] https://launchpad.net/bugs/5996813:17
LarstiQclemente: at least the xml part13:17
=== jelmer is now known as Guest25912
LarstiQhi Guest2591213:18
LarstiQclemente: so just fix 27244413:18
clementeLarstiQ: How? I think two modes are needed, configurable by the user. Some people would like „don't change my encoding, leave the data as it is even if it later fails somewhere else“ and others „please make that my data is available everywhere without encoding problems“13:20
LarstiQclemente: your use case of tres should work perfectly fine with the latter, we _know_ the encoding13:22
LarstiQclemente: if we know the encoding, there is no problem13:22
LarstiQclemente: the problem arises when we do not know the encoding of binary data13:22
clementeSo the problem is not knowing the encoding used to store the target of a symlink13:24
LarstiQclemente: that's a problem, not your problem. If you look at the git-fastexport output you'll see it declares the data to be utf8 encoded13:25
PieterLarstiQ: only the commit messages afaik13:26
LarstiQlink to a file with a name in utf-813:26
LarstiQPieter: ^^ that's in 'expo'13:27
Pieter              It is recommended that <path> always be encoded using UTF-8.13:27
Pieterthat's in the manpage13:27
LarstiQPieter: right, I just looked at the output of clemente's series of git commands13:27
LarstiQclemente: if you try without the symlink, it works fine, right?13:28
clementeLarstiQ: I think so13:29
* LarstiQ commented out the git add prova, and that works13:30
LarstiQclemente: so the concrete problem bzr problem we're dealing with is an (now) arbitrary limitation on symlink targets13:31
LarstiQhmm, maybe not13:31
LarstiQclemente: if I add the symlink and commit it in bzr, it works. Branching that branch however does not.13:32
PieterLarstiQ: that path is only utf-8 in the git output because it was stored as utf-813:32
LarstiQclemente: so some parts allow it, others don't.13:32
LarstiQPieter: right, and I'd like to solve clemente's specific problem without tackling the larger issue13:33
LarstiQthat is a wee bit more work13:33
PieterLarstiQ: by assuming it's utf-8?13:33
LarstiQPieter: by being told it is utf-813:33
LarstiQPieter: which is true in clemente's case13:34
clementeSo the problem is in branch, not in add,rm,log,... A simpler testcase is: mkdir br1; cd br1; bzr init .; touch més; ln -s més prova; bzr add prova; bzr commit -m "link to utf-8 file name"; cd ..; bzr branch br1 br213:41
LarstiQclemente: right13:43
LarstiQclemente: can you write a testcase for that?13:43
LarstiQa bzrlib/tests/ testcase that is13:43
clementeLarstiQ: I'll try that in a while, yes (first I'll need to make them run)13:46
LarstiQclemente: ./bzr selftest <pattern>13:46
clementeok, that was easy :-)  (Although there was no documentation on bzrlib/tests/ explaining how to run them)13:47
clementeor how to run all13:48
clemente./bzr selftest13:49
clementewell, that was also easy... only it wasn't documented there13:49
LarstiQclemente: bzr help selftest?13:49
LarstiQthat, or the developer documentation13:49
clementeYes, but I saw lots of tests in the bzrlib/tests/ and didn't know how to run them :-)13:50
clementeMaybe a README or INFO helps there13:50
clementebzrlib/tests/INFO, I mean13:50
clementeI'll work more on this later, and write or extend some tests for utf813:56
LarstiQluks: qannotate is nice!14:07
LarstiQluks: and qbrowse!14:13
LarstiQOdd_Bloke: hmmm, something funny is going on when I bisect run, it seems to keep repeating going back to the tip14:45
LarstiQho hum, why was ie.get_tar_item deprecated16:05
Guest55688LarstiQ, it sucks16:08
=== Guest55688 is now known as jelmer
LarstiQjelmer: the removal sucks, or the function sucks?16:09
LarstiQjelmer: or your nick change sucks?16:14
jelmerLarstiQ, get_tar_item can't provide a last modification time16:15
jelmerLarstiQ, which caused issues exporting the same .tar.gz file for the same revision16:16
jelmerLarstiQ, not sure if that's the reason it was deprecated though16:16
LarstiQjelmer: I don't think that's fixed now though16:16
LarstiQitem.mtime = now16:16
LarstiQjelmer: http://thread.gmane.org/gmane.comp.version-control.bazaar-ng.general/44475/focus=4447616:17
LarstiQjelmer: but now it seems the exporter code suffers from duplication16:17
jelmerLarstiQ, that was there before too, but it means you have to pass in "now"16:17
LarstiQjelmer: mkay16:19
LarstiQat least I only have to fix _export_iter_entries() once16:21
=== ferrouswheel__ is now known as ferrouswheel
LarstiQright, back to the bzrlib/merge.py conflict16:50
=== beaumonta is now known as abeaumont
fdvdoes anybody know whether (and possibly how) "shallow" checkouts / branches are supported in 1.6? I believe I heard they would be18:30
uwsfdv: I think the name is "stacked"18:33
LarstiQyup18:34
LarstiQfdv: see --stacked on branch and push for instance18:35
fdvright18:38
fdvbut it depends on the availability of it's parent branch for all operations, according to the docs18:38
fdvdo you know if bzr-svn supports it?18:40
fdvheh. seems to :)18:41
LarstiQfdv: should only be the case for operations which it doesn't have the revisions itself for18:41
fdvah, ok18:41
LarstiQfdv: iirc for bzr-svn the layering wasn't entirely right18:41
fdvoh18:41
LarstiQbut it's been a while since I asked jelmer about it18:41
fdvthe bzr-svn faq recommends it18:41
LarstiQreally? Then I guess he fixed it :)18:42
fdv(http://samba.org/~jelmer/bzr-svn/FAQ.html#cloning-a-large-subversion-branch-is-very-slow)18:42
fdvnow I've just got to hassle the svn guys to upgrade to 1.5, then the world will be a better place :)18:42
LarstiQfdv: any idea how large of a difference upgrading to 1.5 makes?18:43
fdvLarstiQ: nope, just read what that faq said :)18:44
fdvit took a week to clone the 1.4 repo to bzr, can't be worse than that ;)18:44
fdvLarstiQ: if and when I figure it out I can try to remember to let you know18:45
LarstiQfdv: yes please, I could go badger some of my svn admins then ;)18:45
fdv:)18:46
fdvLarstiQ: svn 1.4 is so broken that there should be ample reason in itself, afaik18:46
fdvmerging support is paramount, imo18:46
fdv(*working* merge support, that is)18:47
LarstiQI haven't had cause to use it on svn, I heard there is this one python script you need to accomplish it with 1.5?18:47
fdvwith 1.4 you can use svnmerge18:48
fdvit's supposedly fixed in 1.518:48
fdvLarstiQ: have you tried cloning the repo from 1.4?18:48
LarstiQok18:48
LarstiQfdv: with bzr? Yes, although not the entire repository, but projects within it.18:49
fdvright18:49
fdvhm. it might seem to be issues with stacked branches from svn20:11
fdvno matter what type of repo I init, (I tried --1.6.1-rich-root, --rich-root-pack, and --subversion), the result is the same20:14
fdvbzr reports that it uses RepositoryFormatKnitPack5RichRoot for stacking20:15
fdv(after a branch command)20:15
fdvthen it says that SvnRepository('<svnurl>') is not compatible with KnitPackRepository('local-repo-path'), different serializers20:16
fdvdoes this sound familiar to anybody?20:17
dleeWhen I push to a bzr+ssh branch or commit when bound to it, is there yet a way to have it email people about my push or commit?  The bzr+ssh branch is on a Unix (FreeBSD) system I control.20:20
fdvuhm. I had a slightly old version of bzr-svn, sorry about the hassle20:33
=== beaumonta is now known as abeaumont
=== beaumonta is now known as abeaumont
=== rick_h_ is now known as rick_h
=== rick_h is now known as rick_h_
=== rick_h_ is now known as rick_h_mobile
=== rick_h_mobile is now known as rick_h
=== rick_h is now known as rick_h_
alefteristrying to push over sftp and i get "bzr: ERROR: Permission denied: "/blog": [Errno 13] Permission denied"23:20
alefteriswhat can i do to see what is wrong?23:20
alefterisdoes the directory need any special permissions?23:25
alefterisnevermind, solved it23:31

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