Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2014-07-30

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
01:00 stepkut joined #darcs
01:15 amgarchIn9 joined #darcs
01:17 mizu_no_oto joined #darcs
02:15 edwardk joined #darcs
03:11 mornfall joined #darcs
06:33 alegadea joined #darcs
07:47 lelf joined #darcs
09:46 favonia joined #darcs
09:56 c74d joined #darcs
10:30 c74d joined #darcs
11:24 mdiaz joined #darcs
12:01 gh_ joined #darcs
12:34 mdiaz joined #darcs
12:51 mizu_no_oto joined #darcs
15:56 amgarchIn9 joined #darcs
16:11 stepkut I have a sad question. What is the best tool these days for a one time darcs->git migration?
16:13 sm hi stepkut
16:13 sm coming from you, that *is* sad :)
16:13 sm darcs (HEAD) convert --export | git fast-import or similar
16:15 stepkut I have a bunch of subdirectories in a single darcs repo that I want to split into separate git repos. It is best to first convert to git, and then do the splitting?
16:20 sm are your darcs commits mostly segregated by directory ? If not, then I'd guess it'll be easier on the git side
16:22 stepkut mostly, but not entirely
16:25 sm so then darcs get ./REPO --match 'touch DIR' ./REPO-DIR for each directory will mostly work of course, but http://imwill.com/split-a-large-git-repository/#.U9kb_lpdVCc looks like it will entirely work
16:26 stepkut spiffy
16:26 amgarchIn9 joined #darcs
16:28 sm I like the point made there that having independently changing areas in one git repo is a suboptimal (because of all the merging), something that darcs handles with ease
16:29 stepkut yeah.. if only people picked technology based on its merits instead of its popularity
16:29 sm well.. git has its merits too, you'll see
16:30 stepkut :p
16:30 sm I know :)
16:30 stepkut unfortunately, the only merit I need is  github
16:30 stepkut never had any other issues with darcs
17:00 favonia joined #darcs
17:17 mornfall stepkut: actually, git filter-branch
17:18 mornfall stepkut: that'll also work if you have patches that cross subdir boundaries
17:18 stepkut cool
17:18 mornfall (might be the same thing sm posted, didn't check enough backlog)
17:18 stepkut still installing the build dependencies for darcs :)
17:28 stepkut src/Darcs/Patch/Match.hs:360:41: Not in scope: ‘toLower’
17:31 amgarchIn9 joined #darcs
17:32 stepkut adding the obvious import seems to fix that
17:33 stepkut sadly, export died with, 'quit (core dumped)'
17:33 stepkut or perhaps fast-import died.. checking now
17:34 stepkut ohh..
17:34 stepkut I should not do this in eshell.. that sometimes causes problems I think
17:35 stepkut much better
17:45 gh_ joined #darcs
17:48 gh_ I'm experiencing stack overflows at the end of each command with darcs HEAD
17:59 gh_ stack overflow seems to occur before working copy is updated, since obliterate behaves like unrecord now
18:03 Heffalump gh_: on trivial repos?
18:03 gh_ on darcs.net
18:07 gh_ hmm seems to be provoked by my patch "show patch hash in UI"
18:07 gh_ I'm on another manchine than the one I used to write that patch
18:09 gh_ hmm could it be because of patch index.. since patch hash changed
18:12 gh_ ok I nuked the patch index (that was build with the previous patch hash format), but I'm still experiencing stack overflows each time darcs wants to build it from scratch
18:14 gh_ by the way I'm building darcs with profiling enabled on this machine
19:40 gh_ alright.. nothing to see with the aforementioned patch
19:44 gh_ it's just that compiling without optimizations makes patch index creation leak space
19:44 gh_ which is noticeable on big repositories (eg darcs.net)
19:47 Heffalump oh, I now vaguely recall that this was sort of known
19:48 gh_ can it be fixed?
19:49 Heffalump if it was known about, I don't know if it was analysed in any great detail
19:49 Heffalump if it's only without optimisations it's probably not worth too much effort
19:49 Heffalump though it could be unstable under a new GHC version I guess
19:55 gh_ https://dl.dropboxusercontent.com/u/6239815/patchindex_creation_optimized.pdf
19:55 gh_ https://dl.dropboxusercontent.com/u/6239815/patchindex_creation_not_optimized.pdf
19:56 gh_ I think even in the case where darcs si built normally (ie with optimizations), the memory consumption is big
19:57 Heffalump mdiaz: how's things going?
20:00 mdiaz hi Heffalump. All fine. I'm testing the behavior of the  commands under undo. For now, it seems that: record, unrecord and amend works well. I'll see what happens with pull and apply.
20:01 Heffalump are you actually writing tests, or testing manually?
20:02 mdiaz For now, just testing manually.
20:04 Heffalump you're building up quite a big backlog of tests for multiple features :-)
20:04 mdiaz The commands that I have to see if are YesUndoable or NoUndoable are: apply, record, pull, amend, unrevert, tag, rollback, convert and rebase.
20:05 Heffalump hmm, lots of things change working
20:06 mdiaz What are the NoUndoable according to you? :)
20:07 Heffalump apply, pull, rollback, unrevert all change working
20:07 Heffalump rebase has lots of subcommands, you'll need to look at them individually
20:07 Heffalump you've misssed out obliterate
20:07 Heffalump which also changes working
20:07 Heffalump convert might in theory be undoable but it seems like quite a complicated command
20:09 mdiaz Ok.
20:12 gh_ well "convert darcs2" and "convert export" are read-only on the origin repository
20:12 gh_ I don't think there is anything to undo
20:13 Heffalump yeah, that's part of what I was thinking about it being complicated
20:13 Heffalump but I think finalizeRepositoryChanges should be exactly the right way to identify the things that need a decision
20:16 mdiaz Ok, then the more complicated command is rebase. Do you have any suggestions for improving the patch?
20:17 * gh_ runs
20:23 Heffalump I haven't really read the code that stores the undo state yet
20:23 Heffalump it might be useful to write a higher-level description of the format of the undo state
20:26 lelf left #darcs
20:30 mdiaz Mmm... higher-level description... like what?
20:35 mdiaz I would also like to start with 'darcs undelete'. But, I don't know where to start.
20:44 lelf joined #darcs
20:51 Heffalump higher-level description like "the undo state is stored in folder XXX, and contains files YYY and ZZZ..., which should be interepreted as follows: ..."
20:51 Heffalump I think it'd be best to drop darcs undelete. It's not that useful and polishing the existing features is more important.
21:03 mdiaz Ah, ok :). And where I write the description?
21:10 Heffalump hmm, not sure :-)
21:10 Heffalump a comment in the code, probably
21:32 alegadea joined #darcs
21:33 mdiaz Ok, or in the wiki.
21:52 gh_ joined #darcs
22:22 lelf joined #darcs

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary