Camelia, the Perl 6 bug

IRC log for #darcs, 2013-04-22

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

All times shown according to UTC.

Time Nick Message
00:09 edwardk joined #darcs
05:46 kmels__ joined #darcs
05:54 gh_ joined #darcs
07:18 lelit joined #darcs
07:33 DarkFox joined #darcs
07:33 DarkFox Thanks sm :)
08:07 javier_rooster joined #darcs
08:17 donri joined #darcs
08:34 owst joined #darcs
08:54 raichoo joined #darcs
09:00 raichoo joined #darcs
09:06 raichoo joined #darcs
09:08 kmels_ joined #darcs
10:32 amgarchIn9 joined #darcs
11:13 schlaftier joined #darcs
12:19 mizu_no_oto joined #darcs
12:25 whaletechno joined #darcs
12:37 povman joined #darcs
12:39 edwardk joined #darcs
12:43 mizu_no_oto joined #darcs
13:28 javier_rooster joined #darcs
13:41 lelit joined #darcs
13:53 kmels_ joined #darcs
13:59 uniquenick joined #darcs
14:00 mizu_no_oto joined #darcs
14:06 javier_rooster joined #darcs
14:21 gh_ joined #darcs
14:48 barbapapa joined #darcs
14:55 kmels_ joined #darcs
15:03 favonia joined #darcs
15:24 favonia joined #darcs
15:25 dolio joined #darcs
15:28 javier_rooster joined #darcs
15:36 favonia joined #darcs
16:34 raichoo joined #darcs
16:37 mndrix joined #darcs
17:00 * Heffalump looks for a list of things darcs doesn't have (for a talk): so far I have "multi-head repos", "lots of tools", "good conflict handling", "github", "short revision identifiers". Any other thoughts
17:00 Heffalump ?
17:03 amgarchIn9 joined #darcs
17:05 Igloo An incomprehensible UI
17:05 Igloo (we're looking for differences between git and darcs, right?  :-)  )
17:06 lelit :)
17:06 uniquenick to go along with the current UIs "if you didn't specify, we'll ask you" theme, it would be nice if I could say "darcs push" and have it ask which repo to push to
17:07 javier_rooster joined #darcs
17:07 dolio You should say that it doesn't have Launchpad.
17:07 dolio Not github.
17:07 Igloo uniquenick: If you configure push to use --no-set-default by default then you'll get that, I think
17:08 mizu_no_oto joined #darcs
17:23 edwardk Heffalump: users =)
17:24 kmels_ joined #darcs
17:25 Heffalump edwardk: yeah, I considered that :-)
17:25 Heffalump dolio: why?
17:27 uniquenick if I got a "Failed to commute common patches" error, have no idea how I managed to get it, and am using 2.8.1, do you guys still want a bug report?
17:31 dolio joined #darcs
17:32 Heffalump probably not worth it if you can't reproduce it, sorry :-(
17:32 Heffalump dolio: why launchpad rather than github?
17:33 Heffalump uniquenick: though I have no reason to think it wouldn't have happened in 2.8.4
17:33 uniquenick is that a "your repo might work if you upgrade" or a "your repo is boned, too bad you hadn't upgraded before"?
17:34 dolio Heffalump: Launchpad has a better model for some things.
17:35 dolio Especially if you don't have multi-head repos, like darcs doesn't (as you mentioned).
17:35 Heffalump uniquenick: if you keep getting it in the same repo, then I'm definitely interested in helping you fix it
17:35 Heffalump dolio: perhaps, but it's not launchpad I see in the URL for every software project
17:36 Heffalump uniquenick: or if your repo is broken afterwards in other ways, likewise
17:36 dolio Why is that always the argument?
17:37 uniquenick I have two developers (one being me), we each have our repo on our laptop, and we randomly pull from each other whenever we feel like it
17:37 Heffalump dolio: well from the specific perspective of giving a talk, it's something the audience is likely to relate to
17:37 dolio Oh, well if that's the point, yeah.
17:37 uniquenick everything worked fine for about a year, then friday I tried to pull and got that error with a list of patches going back to sept 2012
17:38 uniquenick pulling the other direction does the same thing
17:38 Heffalump uniquenick: can you make the reops public? Also, ideally I'd help you on Thursday rather than now owing to needing to finish said talk.
17:39 Heffalump but I'll at least have a quick look if I can get the repos
17:40 uniquenick that's the big problem, I doubt I can get permission to give anyone a copy of the repo
17:41 Heffalump ok. Can you paste the full error, if it includes any other info? If the patch names are secret then truncate it before them.
17:41 Heffalump (hpaste or whatever)
17:43 uniquenick the patch names aren't secret, just really long.  do you want them?  if not the error is just "darcs: bug at src/Darcs/Patch/Depends.hs:249 compiled Dec 10 2012 04:49:40"
17:46 favonia joined #darcs
17:47 Heffalump it would be useful to get a sense of how long the list is and to be able to ask you questions about specific patches
17:47 Heffalump does your repo have a lot of conflicts and/or a very recent one?
17:49 uniquenick I think we've only had a few conflicts occasionally.  am I supposed to do something special when I get them?  I just fix it manually and record it, I don't run any special "the conflict is cool now" command or anything
17:51 Heffalump no, you shouldn't need to do anything special, it's just that conflict handling is a good place to look out for darcs bugs
17:52 schlaftier joined #darcs
17:52 Heffalump so given that you've had a bug, it's a useful question to ask
17:53 uniquenick is there a way to tell when I had a conflict?
17:55 xymox joined #darcs
17:56 Heffalump I think darcs changes --verbose will display the conflicts (look for the text "conflictor" assuming you're using darcs 2 patches - check _darcs/format to be sure of that)
17:57 Heffalump anyway, your description of a "a few conflcits occasionally" makes it less likely as a cause
18:15 uniquenick is it normal for "darcs changes --verbose" to show newlines as a $ character if they come after whitespace?
18:16 Heffalump yes, it's intentional
18:17 Heffalump I think the idea is to encourage not having trailing whitespace by making it look ugly, though I'm not 100% certain of that precise nuance.
18:17 Heffalump Certainly it's useful to see that you have it.
18:19 uniquenick so to get things working again, can I just delete one repo and re-clone the other?  or do I have to delete them both and init a fresh one and import everything?
18:19 Heffalump I would expect deleting one and re-cloning the other would work
18:19 Heffalump but I'm not 100% certain without a better understanding of what's wrong with them
18:19 Heffalump does darcs check on each repo individual report any problems?
18:20 uniquenick nope, it thinks they are both fine
18:20 Heffalump I think the most likely cause of your problem is that you have somehow got two patches, one on each side, that darcs thinks are the same (because they have the same metadata including timestamps) but have different contents
18:21 Heffalump so throwing away one side is very likely to solve the problem
18:21 Heffalump (back in a while)
19:45 Heffalump (back)
19:47 uniquenick what's the difference between _darcs/prefs/repos and _darcs/prefs/sources?
19:50 Heffalump sources is where it looks for cached data (patches and pristine)
19:50 Heffalump so shoudl be invisible to you apart from the speed impact
19:50 Heffalump repos is for pull/push
20:07 carter joined #darcs
20:18 mizu_no_oto joined #darcs
21:02 uniquenick would identical patches at identical times with different names break things?
21:13 Heffalump uniquenick: no
21:14 Heffalump uniquenick: also, darcs introduces some random junk into patch names when you record them (and then hides that junk in the UI), so you can't really do it even with identical names and identical times
21:14 uniquenick well, I noticed the oldest patch it listed has different names
21:15 Heffalump if you circumvent the junk - by writing patches directly, or by using an old version of darcs that doesn't include the junk, then that could lead to breakage. But it'd have to be quite an old version.
21:15 Heffalump in what sense?
21:15 uniquenick in the sense that my copy said "Removed padding on call_to_action list elements"
21:16 uniquenick and the other copy got hit with a search+replace and was call-to-action
21:16 Heffalump oh, ick.
21:16 Heffalump that would break things.
21:16 uniquenick so, I would assume the search+replace is the obvious culprit, just not sure what all would break or how badly
21:16 Heffalump any idea how? The data is stored in compressed files to make that unlikely, except _darcs/inventory
21:16 Heffalump or _darcs/hashed_inventory I think actually
21:17 uniquenick hashed_inventory appears to have been hit yeah
21:17 Heffalump probably nothing else would be
21:17 Heffalump so you could try changing it back and see if things start working
21:18 Heffalump I'm surprised darcs check passed, actually
21:19 uniquenick yeah, changing hashed_inventory fixes it
21:19 uniquenick so, since everything else is compressed, I'm probably safe and don't need to re-import everything into a fresh repo right?
21:20 Heffalump that's an unfortunate failure :-( darcs used to have a format that was very prone to this kind of problem, which was part of the reason for moving to compressing and hashing things
21:20 Heffalump but the inventory file is an unfortunate omission
21:20 Heffalump uniquenick: yeah, I think so
21:24 Heffalump uniquenick: well done on debugging it! I've reported http://bugs.darcs.net/issue2318
21:25 uniquenick thank you for the help, I wasn't sure where to even start
21:26 uniquenick would there be any objections to making "darcs pull" with no default present a list of whats in _darcs/prefs/repos and letting the user pick with 1, 2, 3, etc?  instead of just yelling at me to specify a repo?
21:27 Heffalump hmm, interesting idea. Not sure if it's come up before. http://bugs.darcs.net/patch948 is kind of related.
21:27 Heffalump but not quite the same
21:27 Heffalump it seems like a sensible idea to me
21:28 Heffalump though we like to be careful about adding new features for fear of ending up with an inconsistent mess: http://darcs.net/Development/NewFeature
21:31 uniquenick that would work too
21:55 dolio joined #darcs
22:22 gh_ joined #darcs
23:10 gh_ joined #darcs
23:16 mizu_no_oto joined #darcs
23:35 gh_ joined #darcs

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