Camelia, the Perl 6 bug

IRC log for #darcs, 2013-08-15

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

All times shown according to UTC.

Time Nick Message
00:13 kofno joined #darcs
00:27 markstos joined #darcs
00:30 markstos Anyone here who could help me get darcs to compile again on Ubuntu Linux? http://bugs.darcs.net/issue2337
00:32 markstos I'm going to ask in #haskell, too.
00:36 * Heffalump appears
00:36 Heffalump -frts is what I thought should work
00:37 Heffalump have you done a clean build?
00:37 Heffalump oh, wait
00:37 epta joined #darcs
00:37 Heffalump your incantation suggests you're asking *cabal* to run with -K100M
00:37 Heffalump you should do GHCRTS=-K100M when you *run* darcs, not when you build it
00:38 markstos Heffalump: I think that's what I usually do. I keep GHCRTC defined in my .bashrc. I don't usually specify it to cabal,
00:38 markstos but I thought maybe I should in my write-up in case it was affecting things.
00:38 Heffalump I think the problem is that cabal is complaining about it because wasn't itself build with -rtsopts
00:39 markstos OH. AH HAI. So maybe this was my first build since I put that in my .bashrc? I'll try unsetting it.
00:39 Heffalump the fact that the message is prefixed with "setup" implies it's coming from a program called "setup" (which cabal might have built) and not darcs.
00:42 markstos OK. trying again.
00:42 markstos might be working now. Waiting for the build.
00:43 Heffalump if darcs didn't even start building the first time, that's further confirmation of the diagnosis
00:48 markstos Heffalump: that fixed it! I'll write it up in the bug report for darcs, and possibly open a new bug report with cabal suggesting that they consider they provide a more helpful error message in this case.
00:48 Heffalump they don't control the message
00:48 Heffalump it's come from the GHC RTS
00:48 Heffalump on the other hand, arguably cabal and its setup programs should be linked with -rtsopts by default
00:49 Heffalump it's only really a security risk when the Haskell program is running with higher privileges than the person who can control its environment or flags, because some options can be used to write arbitrary files.
00:50 markstos OK, all I know is that would have been VERY helpful is the diagnostic had let me know that the failure was related to a particular environment variable and not the software itself. I did not specify the env variable, or realize I still had it in my .bashrc, and had no reason to suspect it was related.
00:50 Heffalump I doubt that applies to cabal etc (and it only applies to darcs in occasional situations, but enough that we plumped on the side of default to off)
00:50 Heffalump ok, so the message should be clearer too - that's a GHC bug
00:50 markstos Or was, with GHC 7.4.2.
00:51 Heffalump got to go, back in a few hours
00:52 markstos If cabal doesn't every need the flag, it would be a feature to 'cabal' to explicitly unset it, and avoid the situation.
00:52 markstos (ever)
00:52 kofno joined #darcs
01:03 intripoon joined #darcs
01:05 mizu_no_oto joined #darcs
01:35 markstos joined #darcs
02:38 mizu_no_oto joined #darcs
03:16 mizu_no_oto joined #darcs
03:29 preflex_ joined #darcs
03:39 alexei___ joined #darcs
04:04 kofno joined #darcs
06:53 raichoo joined #darcs
07:25 raichoo joined #darcs
07:35 lelit joined #darcs
08:49 adnap Is it possible to change the e-mail in commits in a darcs repo without leaving traces of the old e-mail?
09:01 donri joined #darcs
09:41 owst joined #darcs
09:44 owst adnap: no, not AFAIK. It's certainly something that could be possible but I don't think we've got it implemented (it would of course change all the patches identities and make them incompatible with the existing repo...)
09:45 whaletechno joined #darcs
09:59 adnap owst: I have a repo and the commits have my personal e-mail, but I would like to change it to a software-specific e-mail before I make it public. Any advice?
10:00 adnap owst: The only idea I have is to use the existing repo to manually generate patches, create a new repo, and then apply them all again
10:05 owst I wonder if you could do something with scripting darcs rebase
10:05 owst I think rebase allows you to "steal" other people's patches, so you could steal each patch, using the correct user
10:06 owst let me just try something
10:07 owst Yeah, it's easy.
10:08 owst Take a copy of the repo, then try this:
10:08 owst `darcs rebase suspend -a && darcs rebase unsuspend -a -A 'Author Name <email@example.com>'
10:08 owst err, ignore the backtick at the start
10:10 owst (if you want to keep the original date, pass --keep-date to the unsuspend)
10:17 adnap owst: I'm getting "No such command 'rebase'"
10:19 owst Which version of darcs do you have?
10:19 owst I'm not sure if rebase is in a released version of darcs...
10:19 adnap owst: 2.8.4
10:20 owst Right, so if you want to use rebase you'll need to use the darcs from the darcs repo: www.darcs.net
10:21 owst (`darcs get www.darcs.net && cd www.darcs.net && cabal install`)
10:21 owst That assumes you have cabal/ghc on your machine?
10:25 adnap I do
12:00 kofno joined #darcs
12:19 bfrank joined #darcs
12:23 hualet_deepin joined #darcs
12:24 hualet_deepin left #darcs
13:29 owst adnap: any luck with the rebase?
13:30 adnap owst: No. I haven't bothered to update darcs yet
13:31 owst Ah, fair enough.
13:52 tcs joined #darcs
13:53 tcs Hi
13:54 tcs Is anybody out there?
13:54 tcs How do I get darcs-1 repository into a darcs-2 one?
14:13 owst tcs: using a darcs-2.X binary, `darcs convert existing_path output_path`
14:36 mizu_no_oto joined #darcs
16:46 mizu_no_oto joined #darcs
16:48 pointfree-x60 joined #darcs
16:53 raichoo joined #darcs
16:55 bsrk joined #darcs
16:58 * Heffalump appears
16:58 Heffalump bsrk: how's it going?
17:02 bsrk Hi
17:03 bsrk I have put the development server back online, and have started implementing changes on files
17:04 bsrk I am debugging darcs send, too
17:04 bsrk that's it
17:04 Heffalump is the darcs send problem on the darcsden side?
17:05 Heffalump btw  you didn't add the stuff we discussed about darcs send and anonymous patchse to the todo list (as far as I can tell)
17:07 bsrk well until I get the POST request to work, I can't do that can I?
17:07 bsrk the problem is atleast partially on darcs side, as post requests from the browser do work.
17:10 bsrk I am also having trouble with changes on files
17:10 bsrk currently only the first patch is showing up when running changes on files
17:12 bsrk filterPatchesByNames gives all patches that modified the file(s) given in the arguments, does it not?
17:27 Heffalump you can't do what?
17:32 bsrk When I run darcs send --mail
17:32 bsrk it makes a patch, and makes a post request
17:32 bsrk looking a the logs, I can confirm that it is at the right location
17:33 bsrk however, it is not getting directed to the right handler
17:33 bsrk (I can confirm that post requests from browser is getting redirected to the proper handler)
17:34 Heffalump have you tried tcpdump?
17:35 bsrk no, what does it do?
17:36 Heffalump it or any other packet sniffing tool can be used to capture the data going between client and server
17:37 Heffalump so you can compare the working case (browser) with the failing case (darcs send)
17:37 bsrk It is able to make a request, and get a response
17:38 bsrk but the response it gets is from a completely different handler that what it should be.
17:38 bsrk so, it is not a case of the request itself failing
17:38 Heffalump but something must be different
17:38 bsrk but the request getting directed to the wrong place
17:39 Heffalump and if you compare what the client is sending in each case, you might be able to isolate that
17:39 Heffalump either that, or debug the server logic with trace statements or whatever
17:40 bsrk that was what I was trying. I will look into tcpdump
17:41 Heffalump actually I seem to recall that sniffit is better at capturing packet data
17:43 Heffalump another item for the todo list is to make darcs send not need --mail when it can use POST
17:43 bsrk okay
17:45 Heffalump re darcs changes, did you read what the changes command does?
17:45 bsrk it shows the list of patches, list of patches filtered by file if a file option is given?
17:46 Heffalump I mean in code terms
17:46 bsrk yes, that is what I am trying
17:47 Heffalump if darcs changes does what you'd expect, presumably you can get the correct list by copying what it does
17:49 bsrk yes, but it probably better to rely on the smallest piece of code that does the job
17:49 Heffalump yes, but you can cut down from there if necessary
17:50 Heffalump anyway, I think that's it for today - next meeting Monday?
17:51 bsrk okay
17:51 bsrk left #darcs
18:16 gh_ joined #darcs
18:28 gh_ joined #darcs
18:43 Heffalump vikraman: hi!
18:43 Heffalump I think camp is pretty much inactive at the moment.
18:44 vikraman Heffalump: hi
18:44 vikraman yeah, the latest commits were last year
18:44 Heffalump the darcs patch code is quite well separated from the rest of darcs, but has some historical complexities
18:44 Heffalump but in general implementing a new patch type in darcs shouldn't be too much effort if you ignore UI
18:47 vikraman http://darcs.net/Theory#future-darcs-3--
18:47 Heffalump I think your approach to defining an issue tracker sounds just right. Defining the commute rules will be the tricky bit.
18:47 vikraman you mention writing a pluggable patch theory implementation
18:47 Heffalump nothing's happened about that
18:48 Heffalump but the point about it being pluggable is to make it easy to distribute it separately from darcs itself
18:48 * vikraman curses his flaky internet
18:48 Heffalump which is mostly orthogonal to just defining the patch type itself
18:48 Heffalump I'm having trouble with that page too.
18:49 Heffalump I'm not sure if the wiki is having a problem
18:49 Heffalump hmm, now it works
18:49 Heffalump the nice thing about issue tracking in patches is that it's completely independent of "normal" patch types
18:50 Heffalump also, if you're right that conflicts can be avoided/automatically resolved, then you're not dependent on the conflict resolution code in darcs which is probably the worst aspect of darcs-2 patches
18:51 Heffalump that said, the camp code is almost certainly simpler and cleaner. But less likely to end up in a real software release.
18:51 Heffalump so to an extent the right place to work depends on to what extent you'd like to work out the theoretical issues in a clean sandbox versus working towards something directly usable
18:52 Heffalump I would love to make the darcsden issue tracker under version control
18:56 vikraman Heffalump: did you receive that last message?
18:56 mizu_no_oto joined #darcs
18:57 Heffalump nothing from you for 8 minutes. The channel is logged (see the topic) if you want to see what got through.
18:57 vikraman ok, sorry about that
18:57 Heffalump no problem
18:58 vikraman yeah, so I do want to put the core patch theory into a separate module
19:00 gh_ joined #darcs
19:01 Heffalump in darcs you make a patch type by implementing various classes, so it can live in separate modules in code
19:02 vikraman it's almost similar in camp also
19:02 vikraman I actually started writing some part of it, I'll clean it up and put it online soon
19:03 vikraman also, what about graphictors?
19:04 Heffalump they're vapourware :-)
19:04 vikraman is there some place I can read about it in more detail though?
19:04 Heffalump we need to improve the conflict handling algorithm. darcs-2 is better than darcs-1, and camp has a cleaner less buggy implementation of the same algorithm that darcs-2 uses.
19:05 Heffalump nothing beyond what's on the wiki
19:05 Heffalump though I'm happy to talk about them
19:05 vikraman it's probably bad to have too many conflicts in an issue tracker though
19:05 Heffalump ah. There's basically nothing on the wiki.
19:05 Heffalump yes, I think having an underyling patch type that's "conflict free" is better if possible
19:05 vikraman the simplest way would be to commute patches based on timestamp!
19:06 Heffalump perhaps, though I'm not sure you can satisfy the laws that way (perhaps you can). The other approach is described in http://darcs.net/Ideas/AddAddConflicts
19:06 Heffalump the concept of "semi-conflict"
19:08 vikraman ah, that is interesting
19:08 vikraman there's one more issue to think, ie, how to store the patches/issues
19:09 vikraman we could have a generic backend, that allows saving to files, databases, (memory?) etc.
19:09 Heffalump the darcs storage model is roughly to have a sequence of patches on disk for the history along with the efficient representation of the "current state"
19:10 Heffalump but something different might well be appropriate for issue tracking as you want good access to the history too
19:12 vikraman ok great, so I think this idea is good, and I will take it up
19:13 vikraman I will regularly check back here with whatever progress I make
19:13 Heffalump great!
19:13 Heffalump what's the timescale for your project?
19:14 vikraman 2 semesters
19:14 vikraman 1 year that is
19:36 pointfree-x60 joined #darcs
19:44 alexei___ joined #darcs
19:44 alexei___ joined #darcs
21:09 mizu_no_oto joined #darcs
23:28 burp_ joined #darcs
23:45 kofno joined #darcs

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