Camelia, the Perl 6 bug

IRC log for #darcs, 2013-09-22

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

All times shown according to UTC.

Time Nick Message
00:12 mizu_no_oto joined #darcs
01:40 intripoon_ joined #darcs
02:41 mizu_no_oto joined #darcs
03:14 arpunk_ joined #darcs
03:30 mizu_no_oto joined #darcs
03:59 preflex_ joined #darcs
04:29 byorgey joined #darcs
04:35 adnap joined #darcs
05:11 dolio joined #darcs
07:31 donri joined #darcs
07:36 lelit joined #darcs
08:26 alexei___ joined #darcs
10:29 alexei___ joined #darcs
10:58 arpunk__ joined #darcs
12:38 mizu_no_oto joined #darcs
13:08 gh_ joined #darcs
13:14 arpunk_ joined #darcs
13:39 schlaftier joined #darcs
14:17 MasseR joined #darcs
16:40 rdesfo joined #darcs
16:41 rdesfo I get 'checking whether to use -liconv... not needed.' when I run 'cabal configure -ftest'.  How do I get pass this?
16:44 int-e that doesn't look like an error to me.
16:53 rdesfo I don't see -liconv in cabal configure --help
16:53 int-e it's done in Setup.lhs
16:53 int-e which checks a flag calld  libiconv
16:57 int-e that said I too have trouble with this check ... it says "checking whether to use -liconv... setup: Unable to link against the iconv library." but the test program compiles just fine with gcc.
17:00 int-e Oh. The test fails here because libcurl is missing. PAH.
17:03 sm what is this sinister libiconv-libcurl axis
17:03 int-e (configure -ftest -f-curl  worked)
17:07 int-e sm: It depends on the C library whether -liconv is needed to link programs that use iconv functionality. So Setup.lhs tests this by trying to compile and link a simple program that uses iconv_open. This trial compilation tries to link all extra-libraries with the program. If compilation fails, that's interpreted as -liconv being needed to link to iconv. In my case this was wrong, because the failure was a missing libcurl.
17:09 sm int-e: I see. That's a really helpful explanation, I wonder where you could put it so people affected will see it
17:09 rdesfo $ cabal configure -ftest -f-curl
17:09 rdesfo Resolving dependencies...
17:09 rdesfo Configuring darcs-2.9.7...
17:09 rdesfo checking whether to use -liconv... not needed.
17:09 int-e rdesfo: that's fine, you should  cabal build  next.
17:10 int-e or cabal test?
17:10 rdesfo 'cabal build' works, but I get a message when I 'cabal test'
17:10 int-e (and from what I saw you probably didn't need -f-curl)
17:11 int-e (Sorry for distracting you; I ran into another problem which was related, but different.)
17:15 rdesfo np, cabal test worked :)
17:36 int-e oh why is the text field for the description on the bug tracker so tiny ...
17:45 favonia joined #darcs
17:56 int-e sm: so it's issue 2345 now (nice number!)
18:04 Heffalump int-e: ahah. That explains the other problems people have seen. Thanks!
18:04 Heffalump we've had a few reports of this but I haven't had time to get to the ottom of them
18:05 int-e Hmm, are they on the bugtracker? I didn't find anything but I only did a quick search for iconv.
18:12 sm int-e: good one
18:36 * sm tries uploading a 26M zip file to darcsden as a patch bundle.. hope I don't regret it
18:38 sm hi Heffalump
18:40 sm I was going to ask about making bundles but now I remember. darcs send -O, right
18:41 sm and if I apply bsrk's second patch, I should be able to just darcs send and see it appear in the the repo's issue tracker (or owner's mailbox), right ?
18:42 sm I'm unclear on how that works, do I need to set up a special email address ?
18:44 Heffalump sm: you as the patch sender?
18:45 Heffalump that's part of how "darcs send" works already (it uses the author info in ~/.darcs/author etc)
18:45 Heffalump because normally it wuold send an email
18:48 sm Heffalump: right, so.. it a does a http post instead ? to where, and how does it know to do that ?
18:52 Heffalump it does it to a URL on the darcsden site - it knows how to do that because that's what one of the files on the repo in the darcsden site tells it to do
18:52 Heffalump darcs has supported this from the client side for ages, but there hasn't been any easy way to configure it server-side
18:54 sm Heffalump: how do I set it up ?
18:54 Heffalump in what sense?
18:55 Heffalump on the server side? If you just enable the issue tracker with aditya's patches applied, it ought to just work.
18:55 Heffalump (if not, that's a bug)
18:56 sm I'm testing at http://hub.darcs.net (:8901) /simon/darcsden-simon/issues , which is running aditya's first bundle patch (not the email one)
18:56 alexei___ joined #darcs
18:56 sm so that should let me upload bundle files to the issue tracker.. anything else ?
18:57 Heffalump that's it
18:57 Heffalump oh, errm, there ought to be a way to actually apply the bundles to a repo. However now I think of it I never actually checked that..
18:58 sm ok, so with this patch a contributor can darcs send -O and then upload that to an issue in the upstream repo, as an alternative to forking the upstream repo
18:58 sm you're right, I saw code for that somehwere
18:59 Heffalump right
19:02 Heffalump just tried to upload a bundle to the :8901 instance, but it failed - probably to do with the db schema?
19:04 sm same here
19:05 sm I should remove "The details have been logged." , it's often a lie
19:06 Heffalump it does work on aditya's test instance, FWIW
19:06 sm oh, no it was logged, just not very helpfully
19:06 sm 404 not found from couchdb
19:06 * sm checks for schema changes
19:07 * sm pines for persistent
19:10 sm he forgot to include it in the cleaned up patch.. but there's http://hub.darcs.net:8901/Aditya/darcsden-gsoc/pat​ch/20130830123150-ae621#src/DarcsDen/State/Schema.hs
19:19 rdesfo left #darcs
19:20 mizu_no_oto joined #darcs
19:22 sm one more mail and then I think I'm done for the moment - anything else you think I should look at Heffalump ?
19:27 Heffalump no, don't think so
19:30 sm ok, thanks Heffalump
19:33 sm I think bundle upload is not terribly useful without the direct darcs send support, right
19:43 Heffalump it's not a great workflow without that, but it's arguably still nicer than an entire fork for one patch
19:45 sm ok.. there's cognitive, ui and maintenance costs, but perhaps
19:46 sm what do you think about the mime-string/iconv requirement, we still need windows compatibility right
19:46 Heffalump if no solution is immediately obvious, I'll live without it for the short-term
19:47 sm ok
19:47 Heffalump I can figure out something to fix it myself, even if it's just hacking the feature into an ifdef
19:47 sm indeed
19:49 sm I see.. the direct darcs send patch adds /_darcs/prefs/post returning "REPOURL/send' and a /send handler
19:49 sm so on the client machine, I just have to make sure my upstream repo is set
19:50 sm I wonder what versions of darcs have this capability
19:51 gh_ joined #darcs
19:55 rdesfo joined #darcs
19:56 alexei___ joined #darcs
20:17 rdesfo left #darcs
20:26 * sm ponders unauthenticated darcs send
20:30 sm as bsrk said, the current implement means anyone can http-post an email message containing a patch bundle to any repo on darcs hub, and that message will be emailed to the repo's owner
20:31 sm and secondly, if the repo has an issue tracker enabled, and the message's from address is that of any darcs hub user, the bundle will be posted in that issue tracker instead, owned by that user
20:33 sm and it's easy to create a user or a repo owner with arbitrary email address
20:35 sm so this can be used to send email anonymously to anyone, as long as the message includes a valid darcs bundle
20:40 sm and also to anonymously create issues with arbitrary content (plus a darcs bundle) in any darcs hub issue tracker. Possibly including private ones
20:47 sm that doesn't sound like a good long-term solution
20:47 sm how could it be tweaked so that darcs send'ing to hub is still easy ?
20:55 schlaftier joined #darcs
21:04 sm we can't be an anonymous remailer. So I think we must not forward the email when there's no issue tracker. Instead, we could reject darcs send's POST attempt, giving the reason and suggesting they try again with --to REPO_OWNER_EMAIL
21:06 sm that way, they can still send junk, but not from our machine. Also the default send target probably needs to be configurable in the darcs hub repo, rather than hard-coding it to the http url.
21:09 sm or, darcs send could be smarter, trying both post and email addresses for you
21:12 gh_ joined #darcs
21:20 sm for the issue-tracker-enabled case, you could: add a repo setting controlling acceptance of anonymous darcs send bundles in your tracker. And save only the bundle and limited metadata - not the full email content. And/or, let repo owners approve incoming bundles before they're generally visible
21:20 sm
21:20 sm o/
21:21 mizu_no_oto joined #darcs

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