IRC_Meetings: ghc-meeting-2008-08-06.log

File ghc-meeting-2008-08-06.log, 24.9 KB (added by nominolo, 6 years ago)

4th meeting log

Line 
117:00 < JaffaCake> welcome to the GHC IRC meeting!
217:00 < allbery_b> yep
317:00 <   dejones> :)
417:01 <  nominolo> topic?
517:01              nominolo didn't get the memo
617:01 < JaffaCake> have we anything to discuss about the git switchover?
717:01 < JaffaCake> so we discussed branches earlier
817:01 <  nominolo> the Git repo is now automatically kept in sync
917:02 < JaffaCake> great
1017:02 <      BSP_> are we going to have Git mirrors of the libraries?
1117:02 <  gbeshers> Is the darcs repository going to be kept going?
1217:02 < JaffaCake> BSP_: I only glanced over your patches so far
1317:02 <     Igloo> git mirros of the libraries aren't very useful, as you can't push to them
1417:02 < JaffaCake> BSP_: we could, but we'd have to make it very clear that you can't commit to them
1517:03 <      BSP_> right
1617:03 <      BSP_> but you do get the ability to checkout consistent snapshots easily
1717:03 < JaffaCake> gbeshers: nobody has asked for that yet...
1817:03 <      BSP_> which as nominolo pointed out is necessary for bisect
1917:03 <  nominolo> Igloo: well, they are useful to mark things as buildable
2017:03 <     Igloo> BSP_'s patches look like they rely on GHC's repo being a git repo, so we can't apply them as they are before the switchover
2117:03 <   dejones> JaffaCake: I guess someone needs to update the GHC wiki to include info for using Git?  http://hackage.haskell.org/trac/ghc/wiki/Building/GettingTheSources
2217:03 < lambdabot> Title: Building/GettingTheSources - GHC - Trac, http://tinyurl.com/y3nvjj
2317:03 <  dcoutts_> BSP_: and they'd have to be totally automatic. I can't sort out problems in a git mirror of the repos I work with for example.
2417:03 < JaffaCake> dejones: yes, that will happen in due course
2517:03 <      BSP_> Igloo: you have to keep darcs-all in there and use that while it's still darcs for the git repo
2617:04 <      BSP_> s/for the git repo/for the main repo/
2717:04 <     therp> are we going to switch to git?
2817:04 <  nominolo> BSP_: your patches are in the git repo already?
2917:04 <  nominolo> s/in/against/
3017:04 < JaffaCake> therp: yes
3117:04 <   dejones> JaffaCake: ok :)
3217:04 <      BSP_> dcoutts_: yeah, the git mirror can be maintained by a posthook like we have for ghc currently
3317:04 <     therp> JaffaCake: leaving darcs behind totally?
3417:04 <     Igloo> BSP_: The aclocal patch will break, too
3517:04 <      BSP_> nominolo: not yet
3617:04 <     Igloo> Why does git make it easier to have consistent repos?
3717:05 < JaffaCake> therp: we keep darcs for the libraries, for now
3817:05 <      BSP_> Igloo: i added a compat version of the aclocal patch that will work if the GHC repo is darcs
3917:05 <  gbeshers> JaffaCake: I have some interest in keeping it going as a) a good Haskell example and b) integration with cabal some day.
4017:05 <      BSP_> Igloo: because we can use submodules to maintain the library checkouts
4117:05 <     Igloo> Aha, *reloads*
4217:05              >>> gwright_!n=user@pool-71-127-251-140.nwrknj.fios.verizon.net
4317:06 < JaffaCake> BSP_: so what does sync-all do?
4417:06 <      BSP_> it's just darcs-all and push-all conglomerated into one script
4517:06 <      BSP_> it does stuff like translate the commands you issue it into git/darcs language as appropriate
4617:06 <      BSP_> (though that's only very basic at the moment)
4717:07              >>> waern!i=53915df2@gateway/web/ajax/mibbit.com/x-8770447c8d7386de
4817:07 < JaffaCake> ok, so you haven't touched darcs-all
4917:07 <      BSP_> right
5017:07 <      BSP_> the buildbots can change first, to use git and sync-all
5117:08 <     Igloo> But you've changed the packages format, haven't you?
5217:08 <      BSP_> yep
5317:08 <      BSP_> but it turns out it still works with the old darcs-all
5417:08 <      BSP_> the existing regex was permissive enough
5517:08 <      BSP_> (well, i had to remove one line of darcs-all, since i added the root repo explicitly to the packages list: you can remove that refactoring if you like)
5617:09 <     Igloo> Ah, I forgot the $. Fair enough
5717:10 <     Igloo> Wait, I'm sure this won't work properly
5817:10 <      BSP_> oh?
5917:10 <     Igloo> It'll treat   "dir repo VCS"   as   "dir tag repo"
6017:10 <     Igloo> Where "tag" is something like "extralibs"
6117:11 < JaffaCake> yep
6217:11 <      BSP_> i wonder why it worked when i tested it?
6317:11 <     Igloo> Only some commands will break
6417:11 <      BSP_> ok, that's pretty easy to fix
6517:11 <     Igloo> Yup
6617:11 <      BSP_> maybe we should discuss this outside the meeting, though?
6717:12 <     Igloo> Sure
6817:12 < JaffaCake> so BSP_, will you push these patches when they're ready?
6917:13 <      BSP_> JaffaCake: i can if you like (assuming i have the permissions to do that)
7017:13 <      BSP_> but i can't change the buildbots
7117:13 < JaffaCake> do you have an account? I don't remember
7217:13 <     Igloo> I'll do that once the patches are in
7317:13 < JaffaCake> ok, great
7417:13 <     Igloo> I can push the patches too, if you can't
7517:13 <  nominolo> we have to be careful about one thing:  at one point we need to add a .gitignore file similar to the .darcs-boring file
7617:13 < JaffaCake> Igloo: thanks
7717:13 <     Igloo> (But changing the buildbots shouldn't be neccessary for the patches, of course)
7817:13 <  nominolo> this may interfere with the sync-script
7917:14 <  nominolo> (i did it locally and had some issues)
8017:14 <     Igloo> nominolo: Why?
8117:14 <      BSP_> could we push the .gitignore to just the git repo?
8217:14 <  nominolo> it could stop some files from being imported
8317:14 < JaffaCake> will that mess up the sync?
8417:14 <     Igloo> I don't understand the problem
8517:15 <  nominolo> Igloo: darcs-to-git works by pulling a patch, then reflecting the changes with the appropriate darcs commands
8617:15 <     Igloo> Oh, you mean if files that are in the repo are listed in gitignore then it won't add the changes for them?
8717:15 <  nominolo> if the .gitignore file masks out some changes this may mean that some files are not exported
8817:15 <     Igloo> Do you mean "appropriate git commands"?
8917:15 <     Igloo> OK, but that's a bug in teh .gitignore file, isn't it?
9017:16 <  nominolo> Igloo: yes, git
9117:16 <  nominolo> not darcs
9217:16 <  nominolo> Igloo: no
9317:16 <  nominolo> sometimes users explicitly add an ignored file
9417:17 <     Igloo> But if this goes wrong in the sync, then it'll also go wrong in our repos once we start using git
9517:17 <     Igloo> git commit -a    will miss things
9617:17 < JaffaCake> nominolo: shouldn't the sync override .gitignore?
9717:17 <  nominolo> yes
9817:17 <      BSP_> Igloo: not if they are already in the git repo
9917:17 <  nominolo> Igloo: but you could override it with an explicit git add
10017:17 <  nominolo> Igloo: like in darcs
10117:17 <  nominolo> JaffaCake: yes
10217:17 <  nominolo> JaffaCake: that's what i'm going to do
10317:18 <     Igloo> "yes" to what?
10417:18 <  nominolo> Igloo: to git commit -a behaviour
10517:18 <     Igloo> So hang on. Suppose foo is already in the repo, and listed in gitignore
10617:18 <     Igloo> If I make a change to foo, will commit -a see it?
10717:18 <      BSP_> i'm pretty sure if you change foo it will be picked up by -a
10817:18 <      BSP_> as long as it's in the repo
10917:18 < JaffaCake> nominolo: ok, so we need to wait until your fix is in place before we commit .gitignore
11017:18              <   isaacd!n=me@71-10-234-12.dhcp.oxfr.ma.charter.com [Connection timed out]
11117:19 <  nominolo> Igloo: we should try it
11217:19 <     Igloo> OK, so the only problem is if we add files that are being ignored
11317:19 <     Igloo> If we can ignore gitignore then that sounds best, anyway
11417:19 <  nominolo> Igloo: yes, i already had this problem locally
11517:19 <      BSP_> nominolo: by the way, i've got an updated .gitignore on my darcsconversion wiki page that includes some stuff that should have been in darcs-boring but wasn't
11617:20 <  nominolo> Igloo: actually, a workaround would be to p
11717:20              >>> isaacd!n=me@71-10-234-12.dhcp.oxfr.ma.charter.com
11817:20 <  nominolo> p
11917:20 <  nominolo> arg
12017:20 < JaffaCake> ok, in terms of priorities, the post-receive-email is essential, but GitPlugin and gitweb are not
12117:21 < JaffaCake> since we have github for browsing
12217:21 <     Igloo> GitPlugin for trax?
12317:21 < JaffaCake> yep
12417:21 < JaffaCake> would be very nice, but not essential
12517:22 < JaffaCake> I'd like to be able to link to commits from tickets, instead of cut/pasting
12617:22              JaffaCake has been playing with gitsum today
12717:22 < JaffaCake> works quite nicely
12817:23 < JaffaCake> it's an emacs mode that lets you interactively edit a patch before committing/reverting
12917:23 <  nominolo> Igloo: a workaround to all this would be:  keep the gitsync repo as it is, clone this repo, add the .gitignore there and automatically push all imported changes into that repo
13017:24 < JaffaCake> nominolo: wouldn't that need to do push+rebase?
13117:24 <  nominolo> JaffaCake: hm
13217:25 <  nominolo> JaffaCake: or merge
13317:25 <  nominolo> ok, it's probably cleaner to explicitly ignore the .gitignore
13417:25 < JaffaCake> .gitignore should probably be on the "gitignore" feature branch :)
13517:26              >>> not_here!n=claus_re@host213-122-18-69.in-addr.btopenworld.com
13617:27 <     therp> huch, I get 10kb/s to darcs.haskell.org
13717:27 < JaffaCake> oh yes, simonpj just sent an email to Galois about that
13817:27 <  dcoutts_> I suspect too many people are getting it with git
13917:27 <  malcolmw> Let's say that (hypothetically) darcs-2 improves to the point where ghc might want to start using it again.  Would it be easy to convert back to darcs from git?  Would it be possible to maintain an ongoing darcs-2 mirror of ghc's git repo?
14017:27 <  dcoutts_> and the http server is still throttled for some reason
14117:27 <     Igloo> It's orders of magnitude less throttled than it was
14217:28 < JaffaCake> reddit isn't helping
14317:28 <  dcoutts_> aye, we need redirect filtering :-)
14417:28 <     Igloo> b/w use for the last half day or so on http://darcs.haskell.org/mrtg/external-bandwidth.html looks like it's significantly higher
14517:28 < lambdabot> Title: Traffic Analysis for external bandwidth -- monk
14617:28 <  dcoutts_> erm referrer
14717:28 <     therp> btw what's "our" affiliation with galois? why are they in charge of darcs.haskell.org?
14817:28 <     Igloo> They bought the machine and bandwidth
14917:29 < JaffaCake> therp: they donate the machine and bandwidth for free
15017:29 <     Igloo> And also donate sysadmin time
15117:29 <  dcoutts_> therp: they're nice people :-)
15217:29 <   dejones> :)
15317:29 <     Igloo> And back it up for us
15417:29 < JaffaCake> right, but they only donate a fixed amount of bandwidth, and it isn't always enough :)
15517:29              <   dolio!n=dolio@216.68.186.59 [Read error: 104 (Connection reset by peer)]
15617:29 <  dcoutts_> JaffaCake: perhaps hackage should not be on the same box
15717:30              dcoutts_ would like to use the whole hackage.h.o domain for hackagedb in any case
15817:30 <  dcoutts_> if we could move the trac instances somewhere
15917:30 <     therp> that sounds nice @ galois
16017:30 < JaffaCake> yeah, but we have those  hackage.haskell.org/trac URLs all over the place
16117:30              >>> chrik!n=usereins@host-091-096-133-212.ewe-ip-backbone.de
16217:31 <  dcoutts_> JaffaCake: true, though redirects are quite possible
16317:31 < JaffaCake> yep
16417:31              <   chrik!n=usereins@host-091-096-133-212.ewe-ip-backbone.de []
16517:31 < JaffaCake> darcs.haskell.org isn't quite suitable any more either :)
16617:33 < JaffaCake> does anyone have any thoughts on the "Unfriendly HEAD unfriendly" thread on cvs-ghc?
16717:33 <   dejones> JaffaCake: Maybe let universities mirror the GHC source code for load balancing?
16817:33              >>> dolio!n=dolio@216.68.186.59
16917:33 <   dejones> I would be willing to ask my CS department if they would.  I know they mirror some Linux distros.
17017:34 < JaffaCake> mirrors are certainly useful
17117:34 < thoughtpo> JaffaCake: i started that, and have gotten the HEAD to build since
17217:34 <  dcoutts_> I doubt that the darcs/git repos are the ones that suck bandwidth
17317:34 <  dcoutts_> dejones: we could easily mirror release tarballs of course
17417:34 < JaffaCake> thoughtpolice: so we plan to try to communicate our activities better in the future
17517:34 <   dejones> dcoutts_: Oh?  Any clue what does suck the bandwidth and can it be mirrored?
17617:35 <  dcoutts_> dejones: I'm not sure we know precisely the breakdown by url
17717:35 <     Igloo> http://www.haskell.org/haskellwiki/Haskell.org_domain has links to analog reports
17817:35 < lambdabot> Title: Haskell.org domain - HaskellWiki
17917:35 < thoughtpo> JaffaCake: excellent! I hope it didn't seem like badgering, I just felt some people were left out of the loop (ChilliX initially prompted me to write to cvs-ghc after me and a few others mentioned it wouldn't build)
18017:36 < JaffaCake> not at all, constructive criticism is always a good thing
18117:36 < thoughtpo> and as I would like to perhaps contribute some stuff (reading a bunch of the commentary now) it would be nice to know when to expect changes that might cause lots of wackiness, although if we now have an OS X buildbot and GHC HQ is getting one too hopefully errs can be caught much quicker
18217:36 < JaffaCake> and it was the last round of badgering that prompted us to start using validate
18317:36 <  dcoutts_> which was a major improvement
18417:37 <  dcoutts_> and if there's access to an OSX box in future that ought to do it
18517:37 < JaffaCake> it's not free, of course - it'll cost Igloo time to validate on OS X too
18617:37 < thoughtpo> dcoutts_: yes malcolmw said earlier there is now an x86 os x bot, and somewhere on cvs-ghc it was mentioned that Igloo should be getting one as well
18717:38 < thoughtpo> which is a tremendous improvement - it has only become apparent (to me at least) in the past few days how many GHC users are also OS X users
18817:38 < thoughtpo> so having a hole like that seemed like a big one
18917:38 <  malcolmw> quick Q on buildbots: the buildslave page seems to have some odd ideas about the date.  Today is apparently 2008-01-06, and I see other dates like 2008-41-05 and 2008-59-06
19017:38 <      BSP_> is it possible to set up a service where you can request the bots to validate your git branch?
19117:38 < JaffaCake> Igloo: there are a lot of buildbots, and several OS X ones, are any of them dormant now?
19217:38 <     Igloo> malcolmw: URL?
19317:38 <  dcoutts_> thoughtpolice: yes, just not many of the developers, which was the problem. OSX hardware is expensive.
19417:39 <  malcolmw> Igloo: http://darcs.haskell.org/buildbot/all/buildslaves
19517:39 <  dcoutts_> BSP_: we've talked about this idea quite a bit. People think it's a good thing, but setting it up seems non-trivial.
19617:39 < thoughtpo> dcoutts_: indeed, to tell you the truth I didn't buy this macbook - it was a present
19717:40 <      pejo> For mirroring releases/etc: how many mega/gigabytes are we talking about?
19817:40 < thoughtpo> so i can see that as a limiting factor for sure
19917:40 < JaffaCake> dcoutts_: fwiw, the cost of the hardware has not been the limiting factor (at least for me)
20017:40 <     Igloo> malcolmw: That looks like someone's used %m for month when it means minutes
20117:41 <  dcoutts_> JaffaCake: heh, it is for me :-)
20217:41 <  dcoutts_> (actually, I'd only get one if required to)
20317:41 < JaffaCake> I don't have room in my brain for OS X too
20417:41 <     Igloo> JaffaCake: Doesn't look like useful stats are immediately available, but I'm sure I can dig some out if it would be useful
20517:42              Igloo will have to work out how to make room on his desk for OS X  :-)
20617:42              >>> simonpj!n=simonpj@nat/microsoft/x-35c26f73bc5a0bbf
20717:42 < JaffaCake> hehe
20817:42 <      BSP_> the buildbots seem to be pervasively broken, even outside os x
20917:42 <     Igloo> I need something like a CD changer, but for laptops
21017:42 < thoughtpo> hehe, OS X doesn't take up much brainpower generally but for you guys it might be different :)
21117:42 < JaffaCake> I hadn't considered that square footage might be the limiting factor
21217:43 < JaffaCake> BSP_: it's mostly the MS firewall breakage
21317:43 <      BSP_> ok
21417:43 < JaffaCake> it severely limits the usefulness, and I think it's got worse lately
21517:44 <      BSP_> would it be possible to host the master on a galois server then?
21617:44 < JaffaCake> we used to get 50% or so, now we don't seem to get any
21717:44 < JaffaCake> the master is on darcs.haskell.org
21817:44 <  malcolmw> Igloo: get a Mac mini - they are only the size of a stack of 5 CD cases.
21917:44 < JaffaCake> many of the clients are here behind our broken firewall, though
22017:44 <     Igloo> full builds are also failing because of orphan warnings at the moment, but we'll fix that in due course
22117:45 <      BSP_> JaffaCake: ah, i see
22217:45 <     Igloo> malcolmw: I'm getting whatever's sent to me  :-)
22317:45 <      BSP_> maybe have a microsoft and a galois master then?
22417:45 <      BSP_> even if it's two things to maintain, at least the bots will /work/..
22517:45 < JaffaCake> Igloo: we've ordered a Mac Pro
22617:45 < JaffaCake> BSP_: that's a thought, yes
22717:45 <  malcolmw> JaffaCake: ooh, nice kit
22817:46 <  dcoutts_> is a Mac Pro a desktop or laptop?
22917:46 <      BSP_> desktop
23017:46 < thoughtpo> desktop
23117:46 <  dcoutts_> ok, good
23217:46 <  malcolmw> Igloo: but it's a floor job, you don't want that on your desktop
23317:46 <     Igloo> "the fastest Mac ever", according to google
23417:46 < JaffaCake> ah no, we ordered a laptop
23517:46 <      BSP_> you mean a macbook pro?
23617:46 < JaffaCake> yep, that's the one :)
23717:46 < JaffaCake> sorry
23817:46 < thoughtpo> hehe, MSR not up to date on macs?!
23917:46 <      BSP_> ;)
24017:47 <  malcolmw> it's the best machine for running Vista you know!
24117:47 < thoughtpo> i'll take your word for it.
24217:47              malcolmw avoids all microsoft products except ghc
24317:47              dcoutts_ avoids all apple products
24417:48              JaffaCake would like to stress that GHC is not an MS product
24517:48 < allbery_b> oh, enough already kids
24617:48 < JaffaCake> :)
24717:48 <  malcolmw> now where were we?
24817:48 <  dcoutts_> done with buildbots?
24917:48 <      BSP_> JaffaCake: have you looked into alternatives to buildbot like tinderbox?
25017:49 < JaffaCake> well, I would like to have an autoamtic patch-testing bot
25117:49 < JaffaCake> BSP_: no
25217:49              nominolo bought his Macbook himself
25317:49 <  dcoutts_> JaffaCake: yes, that would be very useful
25417:49 <   dejones> Maybe someone could ask Apple to donate a machine that could be co-located at Galois for build testing?
25517:50 < JaffaCake> you think Apple would donate a machine to support an MS-managed project?
25617:50 <  nominolo> JaffaCake: i would like a testing branch for the new ghc repo
25717:50 <     Igloo> I think we had a quick look at tinterbox and a couple of others before setting up buildbot, but we made our decision based on 0 experience of any of those system
25817:50 < JaffaCake> I'd switch if it was easy
25917:51 <      BSP_> Igloo: right. if we found an alternative that supported stuff like build-a-branch on request it would be nice
26017:51 <  nominolo> JaffaCake: if you commit to this one, buildbots will start and, if successful, merge this to master
26117:51 <   dejones> JaffaCake: I don't think GHC is "managed" exactly my MS.  It is still an open-source project in which anyone can contribute.
26217:51 <      BSP_> i'll do a little research..
26317:51 < JaffaCake> nominolo: but what if multiple people are submitting?
26417:51 <   dejones> s/my/by
26517:51 < JaffaCake> dejones: right, but Apple might not see it that way... you can always ask, of course
26617:52 < allbery_b> there's also that MSR != the business side of Microsoft
26717:52 <   dejones> JaffaCake: do you think Galois would be willing to co-locate the machine?
26817:52 <  dcoutts_> dejones: I got Sun to donate, but that's because of interesting hardware and multi-core work
26917:52 <   dejones> Or that we could find a place for the machine to be connected to the internet so build testing could be remotely performed if needed.
27017:53 <  nominolo> JaffaCake: then first-come-first-serve
27117:53 < JaffaCake> nominolo: there has to be a queue of some kind, I think
27217:53 <  dcoutts_> JaffaCake: there doesn't need to be a queue on non-working, just on working.
27317:54              <   bos!n=bos@hierophant.serpentine.com [Read error: 113 (No route to host)]
27417:54 <  nominolo> JaffaCake: yes.  the tricky part is really how we can refer to the right version of the darcs repos
27517:54 < JaffaCake> ugh, that's hard
27617:55 <  nominolo> JaffaCake: well, we could have git-mirrors of them and integrate them as sub-modules
27717:55 <   dejones> dcoutts_: Yeah, I guess it can't hurt to ask Apple.  :)
27817:55 <   dejones> I'll investigate it more.
27917:56 <  nominolo> dejones: yeah, give steve a call
28017:56 < JaffaCake> heh
28117:58 <   dejones> nominolo: Np, he's on speed dial, along with Bill.  ;)
28217:59 <  nominolo> dejones: now if you were *really* cool you'd have stallman on speed dial, too  (trap, trap! stallman probably doesn't have a phone!)
28317:59 < JaffaCake> Igloo: how difficult would it be to run another buildbot server locally?
28417:59 <     Igloo> Isn't a phone one of the 2 or 3 things Stallman famously /did/ have?
28518:00 <     Igloo> JaffaCake: Locally whhere?
28618:00 < JaffaCake> here
28718:00 < JaffaCake> Igloo: stallman definitely has a phone, he answered it during a talk he gave here at the computer lab
28818:00 <     Igloo> Ah, i see. Should be easy; just pull the repo and make start
28918:01 < JaffaCake> and point the clients at the new server?
29018:01 <     Igloo> Yup
29118:01 < JaffaCake> how about the email summaries?
29218:01 <     Igloo> Oh, hmm
29318:01 < JaffaCake> hmm, but nobody else would be able to access the web pages
29418:01 < JaffaCake> and the IRC bot probbaly wouldn't work
29518:02 < JaffaCake> well, it might
29618:02 <     Igloo> I think the summaries work by e-mail
29718:03 <     Igloo> Yeah, so summaries and IRC bot should be fine, as long as we can send mail and get out on an IRC port
29818:03 <  nominolo> BSP_: is your sync-all finished?  (i.e., working)
29918:03 <     Igloo> So all we'd lose is web pages for the MSR slaves, I think
30018:03 <      BSP_> nominolo: well, it worked for my limited tests. what do you want to do with it?
30118:04 < JaffaCake> Igloo: right, hmm
30218:04 <  nominolo> BSP_: build ghc from teh git repo
30318:04 <      BSP_> nominolo: yeah, it'll work fine for that
30418:04 <     Igloo> Oh, and the IRC bot isn't too helpful anyway
30518:04 <     Igloo> It might be for you, if we can get the URLs right
30618:04 <      BSP_> nominolo: make sure you commit the other patches though, since the packages format has changed and libraries/Makefile needed a change since git doesn't track empty directories
30718:04 < JaffaCake> Igloo: I do go and look when there's a failed build usually
30818:04 <  nominolo> Igloo: can't you upload the summaries somewhere via ssh?
30918:05 <     Igloo> Right, but I wouldn't be able to, for example
31018:05 < JaffaCake> true
31118:05 <     Igloo> Yeah, we could probably kludge around it somehow
31218:05 <     Igloo> We can send an e-mail if nothing else
31318:06 <  nominolo> Igloo: aren't those logs *very* big?
31418:06 <     Igloo> I think we could send the last N lines or something
31518:07 <  nominolo> or .gz attachments
31618:07 < JaffaCake> dcoutts, Igloo: btw, I think I fixed the Parser.hs problem
31718:07 <  dcoutts_> JaffaCake: yeah? was it the generated makefiles then?
31818:07 < JaffaCake> yep, it just needed to reorder the suffix rules in the generated makefile
31918:08 <  dcoutts_> great
32018:08 <  dcoutts_> so the makefile and build semantics match
32118:08 < JaffaCake> so I push this to the main Cabal repo, right?
32218:08 <  dcoutts_> JaffaCake: sure, I don't ever touch the makefile generation
32318:09 <  dcoutts_> JaffaCake: oh, I see what you're asking, yes, and then Igloo syncs the Cabal repos occasionally
32418:09 <  dcoutts_> JaffaCake: so you could push to the ghc one too if you like
32518:09 <     Igloo> Right
32618:09 < JaffaCake> I'll do that
32718:10 < JaffaCake> is it /home/darcs/Cabal?
32818:10 <     Igloo> cabal
32918:10 <  dcoutts_>  /srv/darcs/cabal
33018:10 <     Igloo> (lower case 'c')
33118:10 < JaffaCake> ok
33218:10 <  malcolmw> Cabal is a symlink to cabal
33318:10 <     Igloo> Oh, cool
33418:11 <     Igloo> Unfortunately there isn't a way to deprecate the old name
33518:11 <  dcoutts_> malcolmw: oh, I see you made that symlink recently
33618:11 <  malcolmw> dcoutts_: recently, about six months ago, yes
33718:12 < JaffaCake> hmm, Igloo: your patch "Windows fixes" is in the GHC branch but not the main cabal repo, it seems
33818:12 <  dcoutts_> Igloo: there is, remove the symlink and add a http redirect. Or make a repo there with a motd that shouts at people :-)
33918:12 <     Igloo> Ooops
34018:12 <  dcoutts_> Igloo: naughty :-)
34118:13 <     Igloo> Are you pushing it, JaffaCake?
34218:13 < JaffaCake> I can do
34318:13 <     Igloo> Thanks
34418:14 <     Igloo> dcoutts_: If you add an MOTD then it has to be a separate repo, and thus will go out of sync, though
34518:14 <  dcoutts_> Igloo: yep, that's ok
34618:14 <     Igloo> I guess so. Not very pleasant, especially if people start pushing without reading
34718:15 <  dcoutts_> pushing could be blocked
34818:17 < JaffaCake> time for me to go, bye folks
34918:17 <     Igloo> See ya!
35018:18              >>> bos!n=bos@adsl-69-107-87-59.dsl.pltn13.pacbell.net
35118:19 <  dcoutts_> taraa JaffaCake
35218:19              <   tibbe!n=tibell@193.142.125.1 ["Ex-Chat"]
35318:21 <   dejones> JaffaCake: cya
35418:31              >>> tibbe!n=tibbe@193.142.125.1
35518:38 < thoughtpo> i guess this means that this week's #ghc meeting has come to a conclusion?