Camelia, the Perl 6 bug

IRC log for #darcs, 2013-02-25

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

All times shown according to UTC.

Time Nick Message
01:41 favonia joined #darcs
02:32 intripoon joined #darcs
03:00 mizu_no_oto joined #darcs
03:36 carter joined #darcs
03:52 whaletechno joined #darcs
04:00 mizu_no_oto joined #darcs
04:19 mizu_no_oto joined #darcs
04:40 preflex_ joined #darcs
05:25 mulander joined #darcs
06:35 lelit joined #darcs
08:09 raichoo joined #darcs
08:12 delamonpansie joined #darcs
09:03 amgarchIn9 joined #darcs
09:31 gh_ joined #darcs
09:41 nomeata joined #darcs
10:23 owst joined #darcs
10:33 schlaftier joined #darcs
12:53 nomeata_ joined #darcs
13:13 iago joined #darcs
13:16 donri joined #darcs
13:26 mizu_no_oto joined #darcs
14:07 gbeshers joined #darcs
14:11 favonia joined #darcs
14:37 mizu_no_oto joined #darcs
14:47 lpierron joined #darcs
14:51 lpierron Hi, I want to patch master branch (600 patches) of a project with my branch (25 patches). I've started 'darcs pull' friday at noon, it's always running 100% of cpu, how long  does it take ?
14:53 lpierron What is the good practice ton merge big assymetric patches ? Maybe start from master branch and add my new functionalities without trying to patch ?
14:54 donri lpierron: what version of darcs are you using and what version of the repo format is the repos? (darcs show repo)
14:56 lpierron darcs --version
14:56 lpierron 2.7.99.1 (release candidate 1)
14:57 donri darcs show repo | grep Format ?
14:57 lpierron hashed
14:58 donri lpierron: no darcs-2?
14:59 donri lpierron: you might want to consider darcs convert'ing to the new darcs-2 format. the old one had problems with exponential merges
14:59 lpierron how can I convert in darcs-2 format ?
15:00 donri lpierron: i think it's just, darcs convert .
15:00 owst yes
15:01 owst be sure to read `darcs convert --help`
15:01 owst There's likely a bunch of stuff on the wiki, too
15:01 owst wiki.darcs.net
15:02 lpierron OK I try it
15:03 donri this should help you in the future but not sure it'll help you now, since you're not supposed to merge patches between separately converted repos ...
15:04 donri so i'm not sure you can convert master and your branch separately and then pull your patches into master
15:04 donri you could sidestep darcs with darcs diff | patch...
15:06 donri if your branch has all patches from master though you could convert your branch and then replace master with it manually
15:09 donri lpierron: ^
15:09 lpierron I converted both repo
15:10 donri note that convert creates new repos
15:11 lpierron yes I see, now I'm a little bit hesitating between : darcs pull from the old master branch or from the new one
15:12 lpierron I read on Wiki : "it’s important to know that you can keep Darcs 1 semantics but still get a lot of the same benefits of the Darcs 2 format by invoking darcs optimize --upgrade (from Darcs 2.4 and up)."
15:12 lpierron is it a really good idea to use "optimize — upgrade" rather than "convert" ?
15:13 donri i don't know
15:16 lpierron I tried to pull and I have a strange bug : darcs: bug at src/Darcs/Patch/Depends.hs:249 compiled Apr  3 2012 09:13:23
15:16 lpierron I pulled master converted in working branch converted
15:18 donri lpierron: as i said you shouldn't merge between separately converted repos
15:20 lpierron Ah yes, I understand now your remark. So I will try some darcs diff
15:22 lpierron So I try, it's not possible to pull from an old format repository, grrr !
15:24 donri hm?
15:28 mizu_no_oto joined #darcs
15:37 gh_ lpierron, `optimize --upgrade` will not do anything since your repositorie sare already hashed
15:38 gh_ lpierron, have you tried pulling with --skip-conflicts ?
15:38 gh_ you may manage to pull a few patches like this
15:43 lpierron oh yes, I've got 200 patches from 600, good !
15:44 gh_ I thought you wanted to pull 25 patches
15:51 lpierron I can try in the other way
15:55 lpierron OK, 3/25, but my repo is messy, I have to revert to a repo before tryning merging
17:25 owst interesting email from Henning Thielem on darcs-users, I wonder who (if anyone) speaks German, and could speak about darcs?
17:26 owst *Henning Thielemann (stupid mutt chopping names!)
17:27 raichoo joined #darcs
17:29 mizu_no_oto joined #darcs
17:59 amgarchIn9 joined #darcs
18:14 schlaftier joined #darcs
18:57 favonia joined #darcs
19:18 markstos joined #darcs
19:18 markstos I compiled darcs this morning and just tried to push with it, and I'm getting this error:
19:19 markstos darcs: Bad default option: command 'apply' has no option '+RTS'.
19:19 markstos I believe I compiled darcs with the RTS flag:
19:19 markstos cabal  -frts -ftest --with-ghc=/home/mark/ghc/bin/ghc install
19:19 markstos In the target repo, I have this:
19:20 markstos apply +RTS -K100M -RTS
19:20 Heffalump in the prefs?
19:20 markstos I have that there, because as been discussed, there is some issue related patch index that causes darcs to fail otherwise
19:20 markstos Yes, _darcs/refs/defaults
19:20 Heffalump (train just getting in, back later)
19:21 Heffalump I didn't realise you could do that
19:21 Heffalump has it worked before
19:21 Heffalump ?
19:21 markstos Yes.
19:21 Heffalump I'm surprised!
19:21 Heffalump <gone>
19:21 markstos So, either I didn't really compile darcs with RTS (I'm not sure how to easily verify that), or something changed so this is no longer allowed in the prefs file.
19:23 donri markstos: you could try darcs +RTS -N<your cores + 1>
19:23 donri should error if not compiled with -rtsopts
19:24 donri markstos: also you can set rts options with $GHCRTS as well
19:24 markstos donri: that command succeeds (prints the help)
19:25 markstos donri: Would that get passed through to the environment when "apply " is called?
19:25 markstos (in a  target repo (on the same machine)?
19:26 donri i think so?
19:26 markstos I'm trying to just take out that prefs like for now..... that worked.
19:26 donri children inherit their parent's environment by default methinks
19:26 markstos I'll leave it at that for now.
19:27 markstos I'll keep that in mind for the next time this blows up.  So: GHCRTS="-K100M" ?
19:28 sm interesting
19:29 Heffalump markstos: what I'm surprised about is that +RTS ... -RTS in prefs/defaults gets processed in a way that the GHC RTS sees it
19:29 Heffalump because I would expect that (a) those arguments only get processed when you invoke a darcs executable, before darcs itself sees them and (b) that when darcs is handling its defaults from its own configuration files, it does so internally
19:29 markstos I'm not sure I can claim that it was truly working before (although I thought it was),only that it didn't parse-fail before.
19:30 Heffalump I see - though I'm still surprised, because the parse failure is exactly the error I'd expect given the way I'd have assumed things work :-)
19:30 markstos Maybe it was a bug fix that darcs started processing the prefs more strictly?
19:31 Heffalump again, not that I can think of :-)
19:31 Heffalump anyway, the main question is how you can make it work the way you do wnat, I guess..
19:32 markstos Removing the RTS prefs is working for the moment. I'll try the GHCRTS environment variable next time I get an RTS blow-up.
19:33 Heffalump ok
19:33 Heffalump perhaps in the meantime we'll fix the performance problem :-)
19:34 markstos That would be awesome.
19:47 stepkut joined #darcs
20:00 favonia joined #darcs
20:01 carter joined #darcs
20:04 dixie hi
20:04 dixie I'm blocked with issue 2163 correction :(
20:05 dixie the basic idea is "darcs amend-record -A should default to my author id"
20:06 dixie The -A expects argument at this time. There two forms of flags for author, either -A <authorId> or --author=<authorID>.
20:09 dixie if the -A has the <authorId> option optional, it is not clear string after -A <somestring> is the authorId or next parameter (e.g. directory fo amend-record)
20:10 dixie not sure if it would be consistent if the -A will always require parameter (authorId) and --author will allow having some defaut value (e.g. author defined for repository).
20:16 Heffalump I'm not entirely sure I understand the bug
20:23 dixie Heffalump: Do you mean my noisy explanation of the problem I have with issue 2163 or issue 2163 itself?
20:25 dixie The amend-record once it has -A passed on command line expects the string value (author id). The proposal in the issue is to have possibility passing -A without string value, so the darcs amend-record will use author id either from _darcs/ or ~/.darcs/ if available.
20:26 dixie my problem is that -A can not be provided without string value since the command line arguments parsing in darcs will not recognize if this value is part of -A or next parameter.
20:27 Heffalump I don't understand the issue itself
20:27 Heffalump oh, I see, he wants to steal the patch
20:27 Heffalump i.e. change the author to himself
20:28 markstos left #darcs
20:28 Heffalump I think we shouldn't make whether or not a flag takes an argument ambiguous
20:28 Heffalump so it would be better to have a new flag
20:28 Heffalump or, something in the interactive prompt to offer it maybe
20:29 dixie ok, I was not sure whether there is other command line parameter which has the similar behavior like it is expected in this issue from -A/--author.
20:30 dixie seems only -O has similar behavior, but it takes the argument without space (-O[<directory>]). But probably it is not good idea to change how the -A is currently handled. I'll check the other suggestions you said. Thanks.
20:33 favonia joined #darcs
20:39 markstos joined #darcs
20:40 markstos Some feedback on the "darcs changes / darcs log" split: I generally like the idea. I like that  "darcs changes" is interactive by default. A lot of darcs commands are interactive by default, and something I enjoy about the darcs design. I'm on the fence about "darcs log" remaining a hidden command. I realize it's there to help folks coming from other systems.
20:41 markstos My final bit of feedback is the the interactive-by-default broke a shell script I'd developed. that  works like this:
20:41 markstos darcs changes -p $@ --summary | grep '\./'  | cut -d ' ' -f 6 | sort -u
20:41 markstos I wonder how many other tools out there expect darcs changes to not be interactive in a case like this?
20:43 Heffalump yeah, breaking scripts is a general problem we have
20:43 Heffalump I'm not really sure if there's a good approach
20:43 markstos It's also not clear how to turn off --interactive. Since "darcs log" is hidden, I don't think you expect people to guess that. I think the answer is "--all", but that's not really intuitive in the case where I want to go "--all -p foo".
20:44 markstos Can you dect that I'm piping the output to a pipe, and disable "interactive by default" in that case?
20:44 markstos I think the "ack-grep" tool works like that.
20:44 markstos Heck, I think darcs may work like that: Color to stdout, no coloring codes to a pipe.
20:45 markstos The other thing is: If --interactive is the default, having an --interactive flag is not nearly as useful as the flag used to disable interactivity.
20:46 markstos Also, if interactive is going to remain the default, and the --interactive flag remains, the description should be updated to include: "[DEFAULT]".
20:47 markstos EOF. end-of-feedback.
20:48 markstos Don't some of the darcs web-based  repo browsers shell out to calling "darcs changes" ?
20:49 dixie markstos: the flags and command line parameters is open theme on darcs, see http://darcs.net/Ideas/Flags
20:50 markstos I see. I would add to the mix: Consider making --disable an "advanced" flag. I have used lots of flags in my years with darcs, and have never used that one.
20:51 markstos It's really not for normal use, but more of a lock-down command as I understand, for darcs repos on servers, being accessed somewhat "read-only"
21:06 dixie hmm, it is good idea to check also the flags Options/Advanced options membership.
21:30 Heffalump markstos: I guess --no-interactive would be better?
21:30 Heffalump what you say about the pager makes sense, though I wonder about consistency with e.g. record - which you can't sanely pipe into a pager
21:30 markstos Heffalump: I think that makes sense, if --interactive is the default.
21:32 markstos It's OK to be inconsistent with record, since the case can't sensibly apply to record.
21:32 Heffalump the two options are next to each other in the help
21:32 Heffalump (--all and --itneractive)
21:32 Heffalump but there's no indication of grouping
21:33 markstos But what about my case where I wanted to use "darcs changes -p" ? It looks like a mistake to use "darcs changes --all -p".It appears that I'm both requesting to all patches back, and to also get back just the patches matching -p.
21:33 Heffalump agreed
21:33 markstos With "--no-interactive -p", that tension doesn't exist.
21:34 Heffalump I'm not mulling it over
21:34 Heffalump s/not/just/
21:34 Heffalump trying to think about getting consistency with other commands and the right thing here - --all is very standard
21:41 Heffalump so adding --no-interactive as an alias for --all is trivial, but how to document it in help
21:44 markstos Include the same help for both: "non-interactive, select all patches, subject to other filters"
21:45 markstos People will look for it for different reasons, so I think it's good to have both names, but also good to document the same, since they really are the same.
21:47 markstos Actually, perhaps "--all" should be hidden. --interactive considers all patches, too.
21:47 markstos The opposite of --all isn't interactive, it is "some"
21:47 Heffalump -a         --all                   answer yes to all patches
21:47 Heffalump -i         --interactive           prompt user interactively
21:47 Heffalump the issue I see here is again consistency with other commands
21:48 Heffalump in part to simplify the implementation, but mainly to give users the message that the options are fundamentally the same
21:48 markstos From the wiki page referenced above, the future looks more like --interactive=[yes/no] ?
21:49 markstos --interactive is kind of like "--lazy". :)
21:49 Heffalump :-)
21:49 Heffalump markstos: good point
21:50 Heffalump though I'm focusing on a short-term locally good fix
21:50 Heffalump not sure if anyone is working on a reworking in time for 2.10
21:50 markstos Yeah, all the yes/no makeover stuff should happen later, all at once.
22:29 favonia joined #darcs
23:26 iago joined #darcs
23:49 dleverton_ joined #darcs

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