Camelia, the Perl 6 bug

IRC log for #darcs, 2013-04-05

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

All times shown according to UTC.

Time Nick Message
00:01 javier_rooster joined #darcs
00:21 mizu_no_oto joined #darcs
01:07 mizu_no_oto joined #darcs
02:11 edwardk joined #darcs
02:57 preflex_ joined #darcs
03:43 javier_rooster joined #darcs
03:57 dolio joined #darcs
05:25 carter joined #darcs
06:09 edwardk joined #darcs
07:37 lelit joined #darcs
08:44 edwardk joined #darcs
08:48 amgarchIn9 joined #darcs
10:38 raichoo joined #darcs
11:09 edwardk joined #darcs
11:47 amgarching joined #darcs
11:58 edwardk joined #darcs
12:22 mizu_no_oto joined #darcs
12:55 whaletechno joined #darcs
13:12 dolio joined #darcs
13:30 mizu_no_oto joined #darcs
14:04 donri joined #darcs
14:34 dolio joined #darcs
14:39 raichoo joined #darcs
14:40 raichoo joined #darcs
15:31 edwardk joined #darcs
15:52 raichoo joined #darcs
17:08 amgarchIn9 joined #darcs
17:13 edwardk joined #darcs
17:16 sm mornfall: do you remember how darcs-fastconvert handles tags ?
17:17 sm a small random-seeming minority of my darcs tags were converted to empty "TAG: xxx" git commits, while the rest became git tags
17:17 sm also I'm wondering if the git tags have the proper date attached, since git tag lists them in alphabetic order
17:28 sm lelit: I gave darcs-fast-export a try. It has nice scripts, but seems super fragile, any unexpected output breaks it
17:29 lelit oh
17:29 lambdabot lelit: You have 1 new message. '/msg lambdabot @messages' to read it.
17:29 Heffalump sm: I suspect it's to do with them not being "clean" in the darcs history
17:29 lelit didn't experience that
17:30 lelit OOC, what kind of "unexpected output"?
17:31 sm more recent darcs warnings and informational output, "you are in a nested repo" is the one I remember
17:31 sm Heffalump: oh, what does that mean ?
17:32 Heffalump sm: if tag A actually depends on patches X, Y, but the order in the repo is X, Y, Z, A
17:32 Heffalump then A isn't clean
17:32 Heffalump whereas if the order is X, Y, A, then it is clean
17:32 Heffalump in other words if there are patches before the tag that could be commuted out, then the tag isn't clean
17:33 sm ah, thanks. Would darcs-fastconvert care about that ?
17:34 Heffalump it has to, because there's no git repository state to tag in general
17:34 Heffalump imagine if it was X, Z, Y, A
17:37 sm I don't grok that yet, but it's more than an easy bugfix in darcs-fastconvert then ? Should I convert my repo to one with all clean tags ?
17:38 Heffalump yes, and if you can - there are some repos where it's impossible
17:38 sm really.. interesting
17:38 Heffalump what are you trying to do, overall? hack up your own bridge?
17:39 sm really, just get the best possible conversion of this repo
17:39 Heffalump actually, if the git repo has merges then it becomes possible, but darcs-fastconvert doesn't do that (and neithr will the bridge, initially)
17:39 donri darcs is a bit like STM: it makes concurrent mutation sane.  git on the other hand is like an immutable data structure.  am I way off here? :)
17:39 sm hledger wants to be on github :-O
17:39 Heffalump donri: kind of :-)
17:39 donri kind of way off?
17:39 Heffalump donri: I can see flaws in the analogy, but it's somewhat valid
17:40 Heffalump sm: how much of a rush are you in?
17:40 sm Heffalump: ah, I see what you mean here: http://hub.darcs.net/simon/darcs-​fastconvert/browse/Export.hs#102
17:40 donri agreed
17:40 sm I'm in a bit of a rush actually, because having one foot in darcs and one git is quite uncomfortable
17:41 sm and kills project momentum a bit
17:41 Heffalump well, you can try to reorder your repo
17:41 Heffalump so you're switching permanently? :-(
17:42 sm I'm afraid it's looking that way, for this repo. I thought I'd keep bridging both for a while but I've grown more nervous about that
17:42 lelit does darcs-fastconvert handle two way sync?
17:42 Heffalump lelit: not really
17:42 Heffalump that's what bridge will do when it gets finished
17:44 Heffalump sm: what have you done to move to git so far?
17:44 sm I've tried all 5 converters and found that darcs-fastconvert works really well except for missing some tags
17:45 sm I've been comparing http://hub.darcs.net/simon/hledger and https://github.com/simonmichael/hledger for issues
17:46 sm incrementally exporting new darcs patches to git also works great
17:46 Heffalump what tags does it miss?
17:47 sm 0.19.3, hledger-web-0.16.2, 0.16.1, and a few more
17:47 sm 0.14.0, 0.6.1, 0.2.
17:47 Heffalump so are you committed to the repo at that github URL yet?
17:48 Heffalump ah, I guess not from the blurb at the top
17:48 sm no, I'll recreate that if needed
17:58 mornfall sm: No idea really. It's been a long while.
17:59 sm it's a good tool, thank you
18:00 * sm went back to GHC 7.0 to get it installing
18:02 * Heffalump is investigating the tags
18:03 mornfall I suspect it only handles optimized tags.
18:03 mornfall Because it runs in sequence.
18:04 Heffalump mornfall: yeah
18:04 Heffalump sm: btw, does your repo have conflicts in it? That's another thing darcs-fc will make a small mess of.
18:05 sm Heffalump: ah, can you remind me how to check ?
18:05 sm I've wondered that for a long time
18:06 Heffalump I can't think of a good way short of grepping in _darcs/patches
18:07 * sm tries zgrep -i conflict _darcs/patches/*
18:10 sm what should I really be looking for ?
18:10 sm oh, it's "conflictor" or something eh ? and since I don't see any, no conflicts ?
18:10 mornfall Possibly merger.
18:11 mornfall Depending whether it's darcs 2 or darcs 1 repo.
18:11 sm darcs-2, certainly
18:11 Heffalump ok, it does have some
18:11 sm hashed, darcs-2
18:12 Heffalump at least, zgrep conflict _darcs/patches/* did return some results for me
18:12 Heffalump and that's in a freshly got repo
18:14 sm wow, 12k files in _darcs/patches for a 3k patch repo - does that sound normal ?
18:15 Heffalump it means darcs never GCs
18:15 Heffalump do you know why you didn't find any conflictors when I did?
18:16 sm well, I was speaking of a smaller repo I tried.. I haven't seen zgrep actually finish in my hledger repo yet
18:16 sm clearly time to get a fresh copy :)
18:18 sm even with 12k files, its weirdly non-terminating and silent. Even grepping for "e" shows nothing.
18:18 sm in the new copy, I see the conflictors
18:20 sm 8 patches with conflictors
18:22 sm good to know. Now I want to check all my repos. :) Anyway what problem should I expect in the darcs-fastconvert conversion ?
18:32 sm oh, "conflictor" can be found in darcs changes -v output also
18:32 * sm kills another unending zgrep in darcs repo, I think hard links are confusing it or something
18:36 amgarchIn9 joined #darcs
18:41 Heffalump sm: darcs-fastconvert will just flatten the conflicts, so you'll lose the original patch that conflicted
18:42 Heffalump e.g. in AB[resolve AB] where A and B conflict, the flattened repo will look like A(A^-1)[resolve AB]
18:42 Heffalump so you'll lose the original B
18:49 sm Heffalump: thanks, interesting.. so the result is correct, but the history is missing a bit
18:52 sm I can see it at http://hub.darcs.net/simon/hledger/changes/page/53 , where "Add Ledger.Amount.convertMixedAmountTo function" is a simple conflicting patch that got resolved by the next one
18:52 sm and https://github.com/simonmichael​/hledger/commits/master?page=45 , where the conflicting patch appears at the bottom (outside of github's usual sort order)
18:53 Heffalump my long term plan is to get the bridge to convert these by making merges
18:53 sm we're messing with their heads
18:53 Heffalump but there's a general tension - in general, you could make git repos with lots of merges, where each patch is commuted as far back as possible
18:54 * sm nods
18:54 Heffalump and your unclean tags are another example of the tension, I suspect (haven't properly investigated them yet, will do so tonight though)
18:57 sm actually though, I see both the conflict and resolution patches in git, even if github got a little creative with the display order - https://github.com/simonmichael/hledger/comm​it/cc9882ec26f05275800f626cbcb8990abca26c27 , https://github.com/simonmichael/hledger/comm​it/b379cc7b7259edf62c3706be99bac2b8bad58343 , same as on darcs hub
18:58 sm well time to shift gears, thanks a lot Heffalump
18:58 sm later all o/
19:20 Heffalump sm: what you'll be missing in the above example is the ability to recover the unconflicted B
19:36 amgarchIn9 joined #darcs
19:49 mizu_no_oto joined #darcs
20:02 edwardk joined #darcs
20:15 carter joined #darcs
20:35 lelit` joined #darcs
21:04 amgarchIn9 joined #darcs
21:26 dixie_ joined #darcs
21:28 amgarchIn9 joined #darcs
21:54 mizu_no_oto joined #darcs
22:21 edwardk joined #darcs
22:22 dolio joined #darcs
22:22 Heffalump sm: I've made a version of hledger with the tags all in order (so they should all be clean)- i t'll appear at http://hub.darcs.net/ganesh/hledger-ordered once my push finishes
22:23 Heffalump at least if my push doesn't break!
22:23 sm Heffalump: wow, thanks. I've been working on the same
22:23 Heffalump the procedure was roughly: (1) make an empty repo (2) repeatedly pull from the hledger repo by:
22:23 Heffalump darcs pull ../hledger --tag '.*' --dont-prompt-for-dependencies
22:23 Heffalump answer 'yd' to just grab the first tag
22:24 sm oh, nice
22:24 Heffalump annoying thing is that the pull is a bit slow when there are lots of patches left, I think because it visits every tag before offering the first
22:24 Heffalump I'd still like to try to do something aout your conflicts (i.e. figure out how to convert them in a nicer way) but that'll have to wait till tomorrow if at all
22:26 sm my procedure was pull up to the first tag manually (because pull -t 0.1 muddled the ordering of the initial patches), then pull each tag in turn
22:28 sm is your push taking quite a while ? I did one earlier that seemed much quicker
22:31 alexei joined #darcs
22:37 sm Heffalump: this one does not convert as well as mine.. it only finds one tag
22:38 sm I'm doing: hledger-ordered$ git init hledger.git; darcs-fastconvert export | (cd hledger.git; git fast-import)
22:39 sm and checking the total commits and tags it reports at the end
22:39 sm I have one that's good, this is just FYI
22:42 sm or even: darcs show tags | wc -l; rm -rf hledger.git; git init hledger.git; darcs-fastconvert export | (cd hledger.git; git fast-import |& grep 'tags   :')
22:55 lelit joined #darcs
22:58 saep joined #darcs
23:24 nand` joined #darcs

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