Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2014-11-25

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

All times shown according to UTC.

Time Nick Message
01:28 gh__ Heffalump, I'm just saying you can change the on-disk format
01:29 gh__ instead of having to mentionning backwards compatibility in a comment of the "maintain the list" patch
01:39 gh__ apart from that, yeah, memory consumption on PI creation is fixed :) I'll see if on use it has changed or not
01:47 gh__ ok, it hasn't changed
01:47 Rastus_Vernon joined #darcs
03:03 Thaalos joined #darcs
03:31 gh__ Heffalump, I'm screening your bundle
04:12 lambdabot joined #darcs
04:42 lambdabot joined #darcs
05:16 vikraman joined #darcs
06:14 Heffalump gh__: oh, I see. I thought that might cause very confusing errors given that the reading would succeed but produce the wrong data.
09:17 Thaalos joined #darcs
10:54 amgarching joined #darcs
12:51 gh__ Heffalump, I don't know about other people, but I systematically disable (delete) and enable PI to test new PI-related patches
12:57 gh__ Heffalump, OTOH it's true an error could happen to users of HEAD
12:59 gh__ we could put a warning message in that repo's MOTD
15:05 amgarching joined #darcs
15:24 amgarching joined #darcs
15:30 alexei joined #darcs
16:25 byorgey joined #darcs
17:26 lelit joined #darcs
17:47 dolio joined #darcs
18:47 Heffalump gh__: so far I seem not to be disciplined enough to do that
19:30 lelit amazing, darcs is the future, we know, but how did it send back to us "Darcs News #109" from Nov. 2015?? :-)
19:35 amgarchIn9 joined #darcs
20:25 xauth joined #darcs
20:31 lelit Heffalump: trying out the new export command, it seems to me that it formats the darcs patch log as one long line, because it does not separate the patchname from the log with a double eoln...
20:31 lelit in other words:
20:32 lelit $ darcs wha
20:32 lelit hunk ./src/Darcs/UI/Commands/Convert.hs 628
20:32 lelit -                                 plog -> BLU.fromString ('\n' : plog)
20:32 lelit +                                 plog -> BLU.fromString ("\n\n" ++ plog)
20:32 lelit
20:32 lelit seems needed
20:33 * lelit retries to mirror a repo with that...
20:35 aristid joined #darcs
20:37 lelit yeah, way better :)
20:38 lelit Heffalump: lemme know if you prefer I do a darcs send
20:42 xauth Hi all. I've converted a git repo with multiple branches to a bunch of darcs repos. I'd like to find and eliminate duplicate patches among the darcs repos. Anyone have ideas how to go about that?
20:43 xauth LIke, a semi-automated way to find duplicates between darcs repos?
20:43 aristid joined #darcs
20:44 Heffalump xauth: does that actually produce duplicate patches in the sense that darcs thinks they are the same?
20:44 xauth Yes, it does.
20:45 Heffalump xauth: then you don't really need to do anything, do you?
20:45 xauth It happens when git branches cherry pick each other's commits.
20:45 Heffalump lelit: I'm not really familiar with convert/export, I guess you should send in a patch or talk to gh__
20:46 Heffalump xauth: so the metadata ends up identical too, i.e. darcs doesn't offer you the duplicated patch to pull when you try to pull from one repo to another?
20:46 xauth Well, yes and no. I'd like to clean up the duplicates to make the reuse of the patches across different branches explict.
20:46 Heffalump hmm, that sounds non-trivial
20:47 xauth No, the metadata is different. That's basically the problem I'm trying to solve.
20:47 lelit Heffalump: ok, thank you. I assume sending a patch will be noticed by him, right?
20:47 Heffalump there is a sort of bug in darcs-2 patch handling that in theory you could exploit: if you pull in the duplicate patch then unpull the other one then in theory you'll be left with a repo that just dpeends on the duplicate
20:48 Heffalump but that is likely to be rather dicey in practice
20:48 Heffalump lelit: I assume so, yes.
20:48 xauth The problem is it's hard to find the "other" one.
20:49 Heffalump xauth: in theory you might pull them all into one big repo then unpull all the extra ones. But I can think of all sorts of things that might go wrong in practice.
20:49 xauth Like, there will be lots of conflicts if you throw everything onto one great heap. :(
20:50 xauth But it might be worth a try.
20:50 Heffalump yes, and I have no idea whether the duplicate hack will actually work in practice if you stress it like that
20:50 Heffalump how did you do the conversion?
20:51 xauth Semi-automated. There's around 500 commits/patches.
20:52 xauth Usually just up-arrow and enter for the simple cases.
20:52 Heffalump you mean you manually recorded the darcs patches?
20:52 xauth Yes.
20:52 Heffalump hmm
20:53 Heffalump and do even the heads of the repos where the commits are the same end up as separate darcs patches, or is it only for the cherry picks?
20:53 Heffalump I don't mean heads, I mean the earlier commits
20:53 Heffalump i.e. if you have a fork, do the commits up to the fork end up as the same darcs patches in the different repos
20:54 xauth Lemme think about that question. I'm no git guru.
20:54 Heffalump it more depends on what darcs did. If you took each git branch and started again from the origin, they are probably different darcs patches each time.
20:54 xauth Yes. It's basically, darcs get the parent repo(s), git checkout, darcs record.
20:55 Heffalump so you have a sense of what darcs repo corresponds to each git commit?
20:55 xauth Yes. I actually have 500 darcs repos. :-]
20:56 Heffalump and you're explicitly mapping commit to repo, and then pulling the right one for the parent before you record a new commit?
20:56 xauth Exactly.
20:56 Heffalump ok, so then it's only the cherry picks that would be different patches, I guess (by construction)
20:56 Heffalump what are you doing about metadata (commit message, author, date/time)?
20:56 xauth Hmm, yes.
20:57 * Heffalump is very interested in this problem in general btw because I want to make a tool that automates it
20:57 xauth git commit id goes into the darcs patch name. git log message goes into the darcs comment. After that, a darcs tag does on top of it with the same name and comment.
20:58 xauth s/does/goes/
20:58 xauth Wife's calling from the bedroom. Bed time soon... :-/
20:58 Heffalump np
20:59 Heffalump come back tomorrow :-)
20:59 xauth I just might. :)
20:59 Heffalump is there any metadata at all that would identify cherry picks?
21:00 xauth Unreliable. Sometimes is written in the log, but that's surely not enough even for semi-automation.
21:00 xauth The info just isn't in git, formally.
21:02 xauth Sometimes cherry picks are modified before commit, but usually they aren't. I can live with not detecting modified cherry picks.
21:03 xauth Anyway. Off. Thanks so far.
21:03 Heffalump my gut feeling is that the recording phase needs to do some kind of heuristic search, offer that to the user, and then record any corrections on top
21:04 xauth For the modified cherry picks, yes, that would be best I suppose
21:04 * xauth waves.
21:04 xauth left #darcs
21:46 lelit Heffalump, gh__: I sent three patches, hopefully they got delivered at last
21:50 Heffalump no sign yet
21:51 lelit uhm
21:55 lelit sigh, my machine has a poorly configured MTA, so the message "darcs send" sent bounced back... I wrote the bundle in a file, and attached it to a mail I sent to patches@darcs.net on my own
22:06 lelit Heffalump: nothing yet? if so, I will resend the bundle with an external webmail...
22:09 Heffalump something arrived, I'm just checking the format
22:12 lelit thankyou
22:14 Heffalump ok, I had to rename the attachment to a .dpatch extension for it to be properly recognised, but other than that it's fine
22:14 Heffalump and "properly recognised" may only apply to my custom "watch the patch tracker" tool
22:15 lelit I'll try to figure out how to configure my local MTA for the next time :)
22:18 Heffalump in time we'll switch to http submission
22:18 Heffalump local MTAs are now the exception rather than the rule :-(
22:42 lelit g'night

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