IRC log of #schooltool for Wednesday, 2006-12-27

*** jinty has quit IRC00:09
*** jinty has joined #schooltool00:10
*** jinty has quit IRC01:03
*** Bhaskar has joined #schooltool06:43
Bhaskarhi povbot06:54
Bhaskarhello srichter06:56
*** vidasp has joined #schooltool10:56
Bhaskarvidasp,do u know schooltool?10:57
*** Aiste has joined #schooltool11:12
BhaskarAiste, do u know about schooltool11:13
*** didymo has joined #schooltool11:36
*** Bhaskar has quit IRC12:04
*** Bhaskar has joined #schooltool12:06
*** Bhaskar has joined #schooltool12:11
*** ddaa has joined #schooltool12:33
ddaaNew bzr import of the old svn dump12:34
ddaathe corresponding svn2bzr should now reuse file ids correctly when importing svn copies across branches12:34
ddaasvn2bzr code at the usual locations12:34
Bhaskarddaa: schooltool localization how??12:34
Bhaskardo u know schooltool12:35
ddaaI was asked to convert the svn repo to bzr12:35
ddaabut I'm not a schooltool developer12:35
Bhaskarr u student12:39
*** Bhaskar has quit IRC12:42
*** jfroche has joined #schooltool12:50
*** ignas has joined #schooltool13:33
*** mgedmin has joined #schooltool13:36
*** mgedmin has quit IRC13:37
*** alga has joined #SchoolTool13:59
ddaaignas: ping14:18
ignasddaa: pong14:18
ignassaw the changes14:18
ddaasaw the st.bzr-4.tgz ?14:18
ignasno, i bzr pulled and generated a repo myself14:19
ddaaI could not find the lyceum branch you were referring to, so I could not test the merge myself14:19
ddaadid it go better?14:19
ignasi am still getting conflicts, but files are full sized now14:19
ddaaignas: when did you last test it?14:19
ignashmm, christmas eve14:20
ddaaI just fixed the file-id problem yesterday, and uploaded the code today.14:20
ignaswhere is the tgz then ?14:20
ddaacode is at the usual location14:20
ignasok :)14:22
ddaabtw, I'm doing this work on vacation, so it's payable in beer ;)14:22
ignashmm :) going to PyCon or Euro Python ?14:22
ddaamh... europython was fun last year, but I have no plan to go yet14:23
ddaaoh, next year in Vilnius14:24
ignasthis year it's going to be even more fun :D the beer is cheaper in here ;)14:24
ddaaI like the place14:24
ddaawent once for a week on a sprint with Steve14:24
ddaahad a great night at the club just next to the hotel14:25
*** joerg_ has joined #schooltool14:25
ddaaI think I'll try to go to europython14:26
ignasmeet you then :)14:29
ddaaregistrations are not open yet, though14:29
*** joerg has quit IRC14:36
ignashmm, how comes bzr log on a file is more cpu intensive than bzr annotate ?14:44
ddaabecause bzr log on a file is buggy14:45
ignasoh :)14:45
ddaaI saw some commit messages that suggests it will be improved in the next release14:45
ignascan i trust it's output ?14:45
ddaaIME it does not give very meaningful output anyway14:45
ignashmm, seems like it's sane to me ...14:45
ddaamh... ha right it's not a problem in your case because there are no merges14:46
ddaaso yes, it's trustworthy in this case14:46
ignasi see14:46
ddaait's buggy as in "does not use more recent APIs that can make log on file very quick"14:47
ignasone problem with the repo I can see is that the history only spans from the "Switching the trunk: branches/srichter-refactor/schooltool -> trunk/schooltool."14:47
ignaseverything up to then seems lost ...14:47
* ddaa checks14:47
ignas in svn14:49
ddaaI see14:50
ddaaokay, that's a difficult problem14:52
ddaar4984 | mg | 2005-09-15 17:08:33 +0200 (Thu, 15 Sep 2005) | 3 lines14:52
ddaaChanged paths:14:52
ddaa   D /branches/srichter-refactor/schooltool14:52
ddaa   A /trunk/schooltool (from /branches/srichter-refactor/schooltool:4983)14:52
ddaaSwitching the trunk: branches/srichter-refactor/schooltool -> trunk/schooltool.14:52
ignashmm, why ? i mean can't svn2bzr just perform a branch rename ?14:52
ddaaso, at this point, the branch root is /branches/srichter-refactor14:53
ddaait can perform a branch rename14:53
ddaabut here it's actually splitting a branch... making a branch of part of an existing branch14:53
ddaaso... it's possible to handle this case by branching14:53
ddaabut then it needs to introduce a change on the split-away branch where anything that was in srichter-refactor but not in srichter-refactor/schooltool appears deleted14:55
ddaaanother option is to make branches/srichter-refactor not be treated a branch, so all its dir contents are treated as branches14:55
ddaabut then its direct file content (not in a subdir) will be lost14:56
ddaathere can be way, but now we are starting to go into project-specific heuristics14:56
ignasnow i have looked at this one too:
ignasiirc DVCS'es do not handle such things easily ...14:58
*** jinty has joined #schooltool14:59
ddaaweird, I though I had an assert in the code to catch this sort of case15:02
ignasit would make a lot of sense to implement this switch in a schooltool specific way to keep at least a  part of the premerge history15:02
ddaaignas: if srichter-refactor is a branch container, it's easy15:03
ignas and click on next changeset to see what was done ...15:03
ddaayeah, I saw that15:04
ddaalooks like it's okay to make it a branch container15:07
ddaasvn log -v -r 1:6293 file://$PWD/st.repo/branches/srichter-refactor | grep '/branches/srichter-refactor' | grep -v '^   . /branches/srichter-refactor/\(schooltool\|schoolbell\)'15:08
ddaa   A /branches/srichter-refactor15:08
ddaa   A /trunk/schooltool (from /branches/srichter-refactor/schooltool:4983)15:08
ddaaso... easy one15:08
ignashope this will work out15:09
ddaait's a correct thing to say15:10
ignasi know code from srichter-refactor/schoolbell was moved to srichter-refactor/schooltool ...15:10
ddaahow do you want this situation to be represented in bzr?15:11
ddaaa problem we have now is that moving or copying a file across branches loses the ancestry of the file15:12
ddaawe could fix this by making bzr see those cases are merges15:12
ddaabut then other files in the merged may have their annotation clobbered15:13
ddaabecause bzr cannot represent (yet) things like "merge only this file from this branch"15:13
ddaaso a partial merge is recorded as "merge whole trees, and revert the other changes"15:14
ddaaso... I see two choices15:14
*** mgedmin has joined #schooltool15:15
ddaa1. either make srichter refactor a branch container (which looks correct), or represent the move of trunk out of it as a branch+renames+"remove schoolbell from this branch"15:16
ddaa2. either lose annotation for files copied or moved across branches, or add merges in the bzr history.15:17
ignashmm, the first one will at least keep more of the schooltool branch history15:17
ignasi don't know how the second one will work out15:17
ddaawell... in 1. we can also represent "copy of a branch into an existing branch" as a "merge + renames"15:18
ignashmm, i am still trying to find the revision that does the copy ...15:19
ddaawhich one?15:19
ddaaI have a few notes here that can help15:19
ignasinitially refactor had schooltool and schoolbell in it, and then is some rev - there was only schooltool left ...15:20
ddaathat remove schoolbell15:21
ddaa   D /branches/srichter-refactor/schoolbell15:21
ignashmm, and files out from it got moved some time earlier ?15:22
ddaasvn ls -R -r 4947 file://$PWD/st.repo/branches/srichter-refactor/schoolbell@4947 | less15:23
ddaashows there was still quite a lot of content in this branch at this time15:23
ddaahard to say how much was moved away15:24
ignashmm, i just thought if it was just 1 large move (nor many merges+removes) or a lot of single file copies ...15:25
ddaasvn log -v -r 1:6293 file://$PWD/st.repo/branches/srichter-refactor | grep '/branches/srichter-refactor' | grep 'schooltool' | grep 'schoolbell' | less15:25
ddaathat will give some idea, it's one screenful15:25
ignas8 revisions15:27
ddaalooks like the bulk of the copying happened in r481015:27
ddaabut it was mostly copying15:27
ddaanot renaming15:28
ddaaso if we use a single branch, we'll lose the history of the files anyway because bzr cannot represent copies.15:29
ddaaUnless we go really out of our way to say "in this svn revision, a copy renames the source over the destination and recreates the source with a new id"15:29
ignashmm, the source get's deleted anyway15:29
ddaabut not in this very revision15:30
ignasi mean - we don't care about what happens to it ...15:31
ddaaI guess we can get good results by making srichter-refactor a non-branch, and causing copy-across-branches be recorded as merge15:32
ignasmight just work15:32
ddaajust depends on how much you guys were copying single files across branches instead of doing "svn merge"15:33
ddaaDo you see what would be the problem?15:33
ignashmm, looking at the svn2bzr.log i can see a lot of copies15:34
ignasi think a lot of merges are treated as copies by svn (not sure about that though ...)15:34
ddaaat least, for all files that are new in the branche merged from15:35
ddaa(I actually tested it)15:35
ddaawhich is what caused the conflict with the .po files you had the last time15:36
ignasi see15:36
ddaatricky dickey15:37
ignasif it was easy i would have migrated to bzr 6 months ago :)15:39
ddaamy job if solving tricky bzr migration problems15:39
ddaaCall me Migration Item15:40
ddaaI'm starting to be annoyed by svn2bzr... its rename detection is way too simplistic, and it got no tests, so I can confidently do the major surgery it needs...15:42
ddaagot an idea15:44
ddaathe problem with the merge detection15:44
ddaais that in r4810 we see files copied from two places15:45
ddaaactually several... from r4808 and r480915:45
ddaafrom self branch, and from the schoolbell branch15:45
ddaaall the stuff that comes from schoolbell comes from 480915:46
ddaathe problem is how do we find the right bzr revision to use as merge parent?15:46
ddaawith svn habit of not updating its inventories on commit, it's tricky to map svn-path+svn-revision to bzr-branch+bzr-revision15:48
mgedminmany of these copies are actually moves15:49
mgedminsvn treats moves as svn cp + svn rm15:49
mgedminthe other ones are all branching15:49
ddaabut it looks like we can do r4810 just fine if we just filter the copyfrom paths that come from self branch. Then all the other copies come from the schoolbell branch 4809, and we can easily find the bzr revision for this.15:49
ddaamgedmin: ack, it's actually mostly renames involving 'R' actions15:50
ddaaand deep (directory) deletes15:50
ddaaand svn2bzr rename handling sucks too much to deal correctly with those15:50
ddaaand it's also change-based instead of changeset-based, so it does not have the big picture15:50
* ddaa goes to hack something15:53
ddaaI think I'll also try to hack the bzr-cvsps MinimalTree in there, I'm tired of waiting hours for the import to run.15:55
*** jinty has quit IRC18:04
*** alga has quit IRC18:10
*** Aiste has quit IRC19:34
*** Aiste has joined #schooltool19:37
*** ignas has quit IRC19:42
*** ignas has joined #schooltool21:00
*** mgedmin has quit IRC21:20
*** jinty has joined #schooltool21:50
*** jfroche has quit IRC22:18

Generated by 2.15.1 by Marius Gedminas - find it at!