Camelia, the Perl 6 bug

IRC log for #darcs, 2013-02-17

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

All times shown according to UTC.

Time Nick Message
00:02 dixie ssss
00:02 dixie oops
00:13 mizu_no_oto joined #darcs
01:41 mizu_no_oto joined #darcs
01:47 markstos joined #darcs
02:27 M30W joined #darcs
04:49 preflex_ joined #darcs
05:35 xymox joined #darcs
05:43 xymox joined #darcs
06:42 kowey joined #darcs
06:42 stepkut sm: have you given any thought on how to use the darcs send http POST option with hub.darcs.net?
07:47 Heffalump does it actually have such an option? I thought we needed to create one.
07:47 Heffalump (which we definietly want to do)
07:48 owst joined #darcs
07:50 donri i think that's what stepkut is asking about
07:52 Heffalump ok, I thought "use the darcs send ..." implied it existed
07:53 Heffalump but I sometimes fail to be aware of features darcs has had for years, so my impression that it didn't could be wrong
07:58 kowey the remote end needs to have a post file, or something
07:58 kowey but it's there
07:59 donri wait what, this is implemented?
07:59 kowey yup
07:59 kowey by David, ages ago
07:59 kowey but it sort of got forgotten
07:59 Heffalump what does the remote end need to do?
07:59 donri wat
07:59 kowey _darcs/prefs/post
08:00 kowey needs to have a url, and there needs to be a remote thingy that can accept it
08:00 kowey I think somebody submitted a Perl script to illustrate
08:00 kowey contrib/upload.cgi
08:08 owst joined #darcs
08:50 lelit joined #darcs
08:58 owst joined #darcs
09:18 nomeata joined #darcs
09:21 dixie ll
09:21 dixie ops
09:42 gal_bolle joined #darcs
09:43 gal_bolle morning every one
09:44 owst Don't you mean: *bang* morning *bang* *bang* every *bang* one *bang*?
09:45 dixie :)
10:33 gh_ joined #darcs
10:34 gh_ hi
10:34 dixie morning
10:47 gh_ interesting read if you have nothing to do when darcs compiles or when darcs-test runs http://www.infoworld.com/d/open-sou​rce-software/what-you-can-learn-the​-monster-libreoffice-project-212908
11:02 gh_ more opinions about: http://bugs.darcs.net/patch1012 ?
11:04 gh_ I could reintroduce --hashed and make darcs fail when the flag is used. but it's a little ugly, and adding code lines just for that doesn't seem worth it to me.
11:05 Heffalump I would rather not silently take away options immediately, so I'd prefer that, but I don't feel strongly
11:07 gh_ ok, reintroducing it
11:27 gh_ done, if someone wants to review this bundle (which is screened) http://bugs.darcs.net/patch1012
11:29 Heffalump done
11:29 gh_ cool
11:31 iago joined #darcs
11:35 dixie gh_: the tests/bin/convert-writer.sh still uses the --hashed parameter. maybe we should replace it with --darcs-1
11:35 dixie but not sure
11:36 gh_ oh, I look into it⋅
11:38 dixie gh_: I found it only with the grep, no tested yet
11:39 dixie convert.sh (Darcs2): [Skipped]
11:39 gh_ ok that's why it did not appear to me as failing
11:39 gh_ I wonder what made it be "Skipped"?
11:39 dixie grep old-fashioned $HOME/.darcs/defaults || grep hashed $HOME/.darcs/defaults || exit 200
11:40 gh_ wow there's a lot of them
11:42 gh_ ok I had updated the harness as writing darcs-1 to the default files, but all scripts that ought to read the default file still have hashed instead of darcs-1
11:44 kmels joined #darcs
12:03 markstos joined #darcs
12:04 Heffalump morning
12:05 markstos Thanks. Good afternoon.
12:05 Heffalump btw, how are you sending your patches to the tracker? For some reasno they're not coming up as attachments.
12:06 markstos sendmail isn't set up on my laptop, so sent them using Thunderbird manually, as attachments.
12:06 markstos Did they get inlined?
12:06 Heffalump yeah :-/
12:07 Heffalump biab, lunch
12:07 markstos Bummer. I can find an alternate method to send them.
12:13 kowey joined #darcs
12:36 gh_ (last regrets discussion in the sprint room)
12:43 amgarchIn9 joined #darcs
12:46 Heffalump conclusion is we keep it at is, but with some documentation, and add Return as an alias for Space to pick the default option at any prompt
12:47 gal_bolle some documentation being "did you know you can avoid the last question by pressing 'a' or 'd' when you're done?"
12:47 owst With future fixes as we find them, including: "Did you mean to record nothing?" if you've said n to each "do you want to record this?"
12:55 iago is rebase planned to be in the next Darcs release?
12:55 Heffalump iago: yes
12:55 Heffalump it's in head, it basically has to be :-)
12:56 amgarchIn9 joined #darcs
12:58 gal_bolle you could always rebase it out of there
13:01 Heffalump :-p
13:14 amgarchIn9 joined #darcs
13:15 mizu_no_oto joined #darcs
13:18 amgarchIn9 joined #darcs
13:19 amgarchIn9 joined #darcs
13:43 marmoute I'm a bit late on schedule but I should come to see you in the next hour
13:43 marmoute I hope you won't be all leaving to the airport by then.
13:44 gh_ marmoute, ok, if you can make it through the chinese new year parade :)
13:45 dixie they makes here the supportive environment for owst work on darcs-bridge
13:45 gh_ and the fireworks
13:45 gh_ dixie, :)
13:48 marmoute That why I've no estimate more precise than 1 hour ^^
13:49 Heffalump owst and I will be here for about 2 - 2.5 hours more
13:49 Heffalump dunno about the others
13:50 gal_bolle actionsView aThing someThings Nothing =
13:50 kowey sounds like aThing/someThings is something that needs to go into/use English
13:50 kowey (the module)
13:51 gal_bolle i think  it does
13:51 Heffalump someone should have added a French module
13:51 gh_ I'll leave at 18hs, maximum
13:51 gal_bolle likewise
13:51 kowey :-)
13:56 dixie I'll leave to the airport about ~17:00, too.
14:00 marmoute Ok, leaving Gobelins now for snakes and dragons.
14:28 gh_ (marmoute is with us right now)
14:30 sm g'day all
14:31 Heffalump hi
14:31 Heffalump did you see the conversation above about HTTP POST support already existing in darcs send?
14:32 sm Heffalump: yes just now
14:32 sm that would be nice
14:33 sm I'm spending an hour or two investigating stale lock files
14:55 nomeata joined #darcs
14:56 gh_ marmoute gave us a short version of his fosdem talk https://fosdem.org/2013/schedule/even​t/changeset_evolution_with_mercurial_​the_next_generation_of_dvcs_features/
14:56 gh_ no we're explaining the darcs model to him
14:57 sm neat
14:58 sm I think I have confirmed interrupting a local darcs push does not leave a lock file, but interrupting one in progress over darcshub-ssh does most of the time
14:59 sm also fwiw if the linux OOM killer kills a large darcs add, the lock file is left
15:00 sm I had the impression darcs was really careful to clean these up, to the point of not letting you kill it until it's done
15:02 gh_ marmoute explains mercurial's cherry picking
15:02 alexei joined #darcs
15:07 owst sm: the OOMKiller is not polite!
15:07 owst (i.e. darcs won't get a chance to clean up)
15:07 owst AFAIK (will check)
15:08 sm fair enough
15:09 sm we expect darcshub ssh to be more polite though. I'm looking at how lock file gets cleaned up normally
15:09 gh_ gal_bolle does a demo of darcs rebase
15:17 gh_ eveything goes 100% as planned
15:17 kowey wow!
15:17 kowey demo gods pleased
15:17 gh_ that was sarcastic :)
15:17 gh_ I'm sorry
15:17 gh_ :)
15:17 kowey too good to be true
15:17 gh_ we wanted to rebase something beyond a tag
15:21 owst sm: do_send_sig_info(SIGKILL, SEND_SIG_FORCED, victim, true);
15:21 owst victim being the memory-hog
15:21 owst linux takes no prisoners!
15:25 exlevan joined #darcs
15:26 exlevan is there a way to pull a patch without its dependencies?
15:28 gh_ pull is an operation that inherently respect dependencies
15:28 nomeata joined #darcs
15:29 gh_ maybe you want something weaker than "pull", as in just grab the patch bundle?
15:30 exlevan I'm trying to get a patch from one branch to another, but it happens to depend on something I don't want to pull
15:30 exlevan maybe rebase can help
15:31 owst If it doesn't textually depend on the unwanted patch, then rebase will be very easy
15:34 exlevan is there an example on how to use it? `darcs help rebase` doesn't show much info
15:36 gh_ rebase demo is over, discussing git-hg bridging
15:40 sm during a push, darcshub-ssh executes darcs transfer-mode & darcs apply via ssh's spawnProcess. This does forkIO and then waits for darcs's output and termination. If the ssh session is terminated, it calls killThread on the darcs process.
15:41 gh_ gal_bolle asks about hg with dumb servers
15:48 sm that raises the ThreadKilled async exception in darcs, which I don't see that darcs catches anywhere
15:55 sm well, except that its running most actions inside withLock, which I'd expect would call releaseLock in this case
15:57 Heffalump we're packing up to go
15:58 * kowey waves to the sprinters
15:58 kowey hope you guys had fun
15:58 kowey bon voyage
15:58 sm good job sprinters
15:58 kowey well, for those of you who are voyaging anyway
16:03 exlevan maybe it could help if there was a way to replace a dependency with a (resolved) conflict
16:08 lelit joined #darcs
16:19 sm hey, darcs compiled itself only once - nice
16:19 sm I hadn't seen that before
16:29 stepkut Heffalump: yeah, I was looking at how hard it would be to implement the post option, and found in the code that it already exists :)
16:30 stepkut Heffalump: so, I thought you must have added it, but then I checked the changelog and found it has been there for ages
16:33 stepkut sm: i think it would be pretty killer if you could do: darcs get ; darcs record ; darcs send, and send a patch back upstream with out ever having to fork, touch the hub.darcs.net web interface, or have a properly configured sendmail -- plus the patch would actually be on the site and not lost in someones inbox. Not quite clear yet on how to make it easiest for the maintainer to merge that patch bundle
16:38 sm stepkut: yes it would
16:39 sm why does http post belong in send, couldn't it just be darcs get http://..., darcs push ?
16:40 kowey guess it depends if we expect it to be applied or not
16:40 kowey putting in send leaves it ambiguous
16:40 kowey you wouldn't want to push for something that generates pull reqs would you?
16:40 sm send would mean don't apply, just deliver it to maintainer for approval
16:40 kowey yup
16:41 kowey I think a darcs push over http might be a good thing at some point, but imagine there's lots and lots of details to sort out
16:41 kowey and that a darcs send sort of thing would still be wanted
16:41 kowey in any case, we do need to remember in the back of our minds that for lots of people setting up something like sendmail for send is a bit of a roadblock [tangent]
16:42 kowey the super-gigantic-neverending-todo-list
16:42 stepkut if you do a 'darcs get http://' and then 'darcs push', what happens now?
16:42 sm so hub would maintain some kind of temporary incoming queue for maintainers, that anyone can send mail to
16:42 kowey probably tells you it doesn't know how to push to http:// urls
16:42 sm but what about spam
16:42 kowey code written with sort of notion of pluggability, ie. if ever darcs learned about more uri schemes…
16:43 stepkut kowey: so, you just get some sort of error?
16:44 stepkut I'm thinking.. if you try to 'darcs push' to an http:// repo *and* _darcs/pref/post exists, it could recommend trying 'darcs send' in the error message
16:44 kowey that'd be nice
16:45 kowey I like a sort of gentle helpfulness in darcs
16:45 kowey not pushy
16:45 stepkut yes
16:45 kowey just oh by the way, you might not have made this (mental) connection
16:46 stepkut right
16:47 mekeor joined #darcs
16:49 stepkut sm: I was thinking you would use the darcs send HTTP POST mechanism instead of email..
16:50 stepkut sm: since 'darcs send' via email requires a working sendmail ... which isn't found nearly so often these days
16:51 stepkut sm: the nice thing about darcs send via HTTP is that it doesn't require the user to configure anything.. you just create _darcs/prefs/post on the serverside (and possibly remove _darcs/prefs/email) and 'darcs send' automatically does a POST
16:52 stepkut to the url listed in the _darcs/prefs/post file
16:53 stepkut and, conveniently, this has already been in darcs for a long time, so it doesn't require people to have a special new version of darcs either :)
16:58 sm yeah, sorry I misspoke, but this would allow anyone to post stuff to (somewhere on) hub really easily
16:58 sm but that's probably ok
16:59 sm personally my priority is that hub do the basic things well
16:59 sm but experiments are welcome too
16:59 sm afk a minute, breakfast
17:03 stepkut i think that being able to send a patch upstream is a pretty 'basic thing' for a source control site :)
17:11 stepkut also, given the fact that most people do not have hub.darcs.net accounts (and don't really want them)... a patch sending method that doesn't require them to create one seems good
17:42 sm stepkut: hub currently supports sending a patch upstream just like github: fork repo, commit patch, let upstream know about it
17:44 stepkut yes.. i think that is good for github junkies.. but, in general, seems like a crazy workflow for sending a one-time patch to someone
17:44 stepkut I've never understood why github users think it is so great.. but then again .. they do use git :)
17:46 sm I hear you
17:47 stepkut let's say I am trying out some new project.. so I do a darcs get http:// and get it. Then i find a minor bug. I do darcs record ; darcs send. done! Or.. do I instead, go to a website, create a new account, find the project url again, create a fork, check out the project a second time, record a patch, push it upstream, and then go back to the website and submit a merge request :-/
17:48 stepkut and ... now there is this fork hanging around that will likely never see another patch
17:48 stepkut but will start cluttering up google search results, etc
17:49 stepkut and make it more confusing as to where the real source is
17:58 teratorn joined #darcs
18:22 shlevy So I ran cabal init and then darcs record broke the generated cabal file into two hunks, was it darcs that did that or cabal init?
18:29 kowey did you mean the addfile and then contents of the file?
18:29 shlevy I'm getting permission denied when pushing to a hub.darcs.net repo for the first time
18:29 shlevy kowey: No, two hunks after the addfile
18:29 kowey hmm, that's odd
18:30 shlevy darcs: /tmp/darcs.log: openFile: permission denied (Permission denied)
18:31 stepkut shlevy: I am getting similar errors
18:31 stepkut /home/darcsden/users/stepcut/web​-plugins//_darcs/prefs/sources: canonicalizePath: does not exist (No such file or directory)
18:31 stepkut darcs: /tmp/darcs.log: openFile: permission denied (Permission denied)
18:31 stepkut so, I don't think it is just you :)
18:37 alexei_ joined #darcs
18:51 amgarchIn9 joined #darcs
18:51 shlevy stepkut: OK :)
18:51 shlevy Who maintains hub.darcs.net?
18:53 * shlevy emails simon@joyful.com
18:53 shlevy Too many simons in Haskell
18:55 gpiero1 joined #darcs
19:04 kmels joined #darcs
19:07 sm shlevy, stepkut: oh! so sorry, that was me debugging
19:08 sm should be fixed now
19:19 stepkut works. thanks!
19:24 mizu_no_oto joined #darcs
19:28 stepcut joined #darcs
19:35 Heffalump shlevy: the two huns are an artefact of a slightly dubious design choice of darcs patches: a newly added file has a single newline in it already from the darcs point of view.
19:41 Heffalump (so as a result any new file with a blank line in ends up being two hunks - one above the blank line and one below it)
19:48 sm darcs does remove the lock when killed - the problem is darcshub-ssh is not actually killing it, the process is left sitting there forever
19:49 Heffalump but darcs itself might end up being OOMed, right?
19:50 sm it could, but more often it seems to be just sitting there waiting on something but not actually growing or doing much
19:50 sm it's not retrying to take the lock.. I don't know what it's doing
19:51 Heffalump I see
19:51 Heffalump well, easy enough fix for darcshub-ssh to see where that goes then
19:51 Heffalump I mean to see how it improves things
19:51 sm what's the fix ?
19:51 Heffalump maybe I misunderstood
19:52 sm well I'm still debugging why it doesn't kill it.. it calls killThread on it
19:52 Heffalump oh, I see
19:53 Heffalump Darcs.Util.SignalHandler is supposed to turn signals into eceptions
19:53 Heffalump oh, killThread is already an exception, duh
19:53 Heffalump withLock in Darcs.Repository.Lock definitely uses finally
19:53 Heffalump (tunnel)
20:01 sm Heffalump: yeah I think darcs is behaving properly here
20:01 sm when I manually kill the hung darcs apply process I see it clear the lock
20:04 amgarchIn9 joined #darcs
20:31 Heffalump so the question is what's happening to it in this case?
20:40 mizu_no_oto joined #darcs
20:55 amgarchIn9 joined #darcs
21:16 gh_ joined #darcs
21:17 gh_ hi again
21:17 carter joined #darcs
21:23 Heffalump hi
21:24 Heffalump sm: can it be reproduced consistently? Or is it a question of waiting around to see.
21:31 marmoute Thanks for your explanation on darcs guys
21:32 sm Heffalump: sorry was away. Yes it's consistent
21:32 sm I'll drop in more debug logging
21:38 Heffalump marmoute: nice to meet you!
21:39 alexei joined #darcs
21:43 * marmoute wonder which one Heffalump is.
21:56 Heffalump Ganesh
22:33 shlevy sm: Thanks! Working now
22:34 shlevy Heffalump: Ah, that explains it.
23:35 alexei joined #darcs
23:36 kmels joined #darcs
23:47 donri joined #darcs

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