Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2014-07-08

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

All times shown according to UTC.

Time Nick Message
00:27 Rastus_Vernon joined #darcs
00:39 mizu_no_oto joined #darcs
00:53 alegadea joined #darcs
01:48 enthropy joined #darcs
01:49 mizu_no_oto joined #darcs
01:52 enthropy hello. I have two repositories that have diverged it doesn't seem that darcs is able to pull one patch
01:52 haasn joined #darcs
01:52 enthropy darcs get http://code.haskell.org/~aavogt/xmonad/hlist/
01:53 enthropy then in that repo  darcs pull http://code.haskell.org/xmonad
01:53 enthropy the first patch there "Fri May  2 14:01:46 EDT 2014  Adam Vogt update testsuite" doesn't seem to merge in finite time
01:55 enthropy can something be done to fix this? I'm don't think I upgraded either repo to darcs-2 format... but if I do that this will cause more problems?
02:37 mizu_no_oto joined #darcs
03:21 dolio joined #darcs
05:26 lelit joined #darcs
09:35 donri joined #darcs
10:59 mizu_no_oto joined #darcs
11:19 mizu_no_oto joined #darcs
12:00 mizu_no_oto joined #darcs
13:16 mizu_no_oto joined #darcs
15:52 jeltsch joined #darcs
15:52 jeltsch Hi, I just compiled darcs 2.8.4, but cannot use it to record patches.
15:53 jeltsch I get the following error message:
15:53 jeltsch darcs failed:  Couldn't fetch `0000000621-4ad28c4e1df911c845235e21f755b1bf3108e265aec94b52544e51220779f8d1'
15:53 jeltsch in subdir inventories from sources:
15:53 jeltsch thisrepo:/home/wolfgang/Informatik/Entwicklung/Haskell/incremental-computation
15:53 jeltsch cache:/home/wolfgang/.darcs/cache
15:53 jeltsch Failed to record patch 'much development'
15:53 jeltsch HINT: I could not reach the following repositories:
15:53 jeltsch /home/wolfgang/.darcs/cache
15:53 jeltsch /home/wolfgang/Informatik/Entwicklung/Haskell/incremental-computation
15:53 jeltsch If you're not using them, you should probably delete
15:53 jeltsch the corresponding entries from _darcs/prefs/sources.
15:53 jeltsch The interesting thing is that there actually is no file _darcs/prefs/sources.
15:54 jeltsch Any ideas what to do?
16:00 jeltsch I should perhaps mention that I have deleted the darcs cache directory (I have actually done a fresh OS install). Could this be the cause? This would be a
16:00 jeltsch bit scary for me, since it would mean that even non-lazy repos are not self-contained.
16:00 jeltsch Is this the case? So far I have relied on repos being self-contained.
16:01 pseudolio You built it with GHC 7.8.
16:02 jeltsch pseudolio: Yes, that’s right. Is this a problem?
16:02 pseudolio Yes.
16:02 jeltsch pseudolio: This is bad.
16:03 jeltsch So should I now install a complete GHC 7.6 suite to get darcs?
16:03 pseudolio There's a bug that was fixed in head where there's a {- NOINLINE ... -} instead of {-# NOINLINE ... #-}
16:03 jeltsch pseudolio: How can inlining change behavior?
16:03 pseudolio It's an unsafePerformIOed IORef.
16:04 jeltsch pseudolio: :-O
16:04 pseudolio So it duplicates a 'top level' mutable variable.
16:04 jeltsch pseudolio: Okay, do you remember which module this is?
16:05 pseudolio No. But it should be easy to grep for {- NOINLINE, I think.
16:05 jeltsch pseudolio: I can patch it myself. I have patched darcs a bit already anyhow.
16:05 pseudolio Or just NOINLINE and look for the malformed pragma. There probably aren't many.
16:06 jeltsch pseudolio: There are two of them. Should both be made proper pragmas, or is one being a comment intentional?
16:06 pseudolio Are they both mutable references?
16:07 jeltsch pseudolio: Yes.
16:07 pseudolio Then I don't think they're intentionally comments.
16:08 jeltsch pseudolio: Okay, darcs is being rebuilt. Let’s see.
16:09 jeltsch pseudolio: Works now! :-)
16:09 jeltsch pseudolio: Thanks a lot.
16:09 pseudolio No problem.
16:13 byorgey wow, that's a really insidious bug.
16:13 pseudolio Took them a while to figure out.
16:13 byorgey I'll bet.
16:13 pseudolio I think it was when someone noticed that -O0 fixed the bug that it really got solved.
16:15 pseudolio I had no idea that GHC up until 7.6 accepted '{- NOINLINE ... -}' as a pragma, though.
16:16 xstill joined #darcs
16:17 byorgey what, it did?  haha, even more insidious than I thought.
16:17 pseudolio It must have. Because this bug only triggered in 7.8.
16:17 pseudolio Either that, or 7.6 wasn't inlining those things.
16:18 pseudolio Even without the pragma.
16:18 pseudolio But that's hard to believe.
16:31 jeltsch left #darcs
16:41 amgarchIn9 joined #darcs
17:03 favonia joined #darcs
17:16 edwardk joined #darcs
17:17 amgarchIn9 joined #darcs
18:46 amgarchIn9 joined #darcs
19:23 favonia joined #darcs
19:25 edwardk joined #darcs
19:45 edwardk joined #darcs
19:59 mdiaz joined #darcs
19:59 mdiaz_ joined #darcs
19:59 Heffalump hi mdiaz
19:59 Heffalump how's it going?
20:00 mdiaz Hi Heffalump. Everything is fine, I'm still working in the patch for the global cache, but it's ok.
20:00 Heffalump which patch do you mean?
20:01 Heffalump the garbage collection or the bucketing?
20:03 mdiaz_ joined #darcs
20:03 mdiaz_ the patch for the garbage collection
20:04 Heffalump ok, great, how far has it got?
20:04 mdiaz_ It's almost ready.
20:04 Heffalump how does it work?
20:05 mdiaz_ But I've a question about lazy repos.
20:05 mdiaz_ How I get the list of patches?
20:06 Heffalump I'm not entirelyy sure, to be honest. What have you tried so far?
20:07 mdiaz_ Nothing.
20:08 mdiaz_ I just did the part regardless non lazy repositories.
20:08 Heffalump Have you looked at what's stored on disk?
20:09 mdiaz_ Yes. I just read the patches in _darcs.
20:10 mdiaz_ And I don't like the idea of downloading the invetories list...
20:10 Heffalump have you looked at what inventories are available already in a lazy repo?
20:11 Heffalump is it just the inventories to cover the patches we already have, or is it all the patches?
20:12 mdiaz_ When you get the repo for the first time, the inventories dirrectory is empty.
20:12 mdiaz_ But I don't know later...
20:13 Heffalump in a completely lazy repo, what does hashed_inventory contain?
20:13 Heffalump does it contain a pointer to the (not present) previous inventory?
20:15 mdiaz_ I think so.
20:16 Heffalump ok, so then operationally one algorithm would be to visit the inventories we do have, started from hashed_inventory, and mark any patches listed in them as "to be kept"
20:16 Heffalump (also mark the inventories we visit as to be kept)
20:16 Heffalump I'm not sure what existing code will do that without also silently downloading more, though.
20:17 mdiaz_ Oh, ok.
20:17 Heffalump you should look in the existing methods in Darcs.Repository.* to figure that out
20:18 mdiaz_ Ok, thanks.
20:18 Heffalump so, can you describe a bit what your nearly-ready patch for garbage collection does/
20:19 mdiaz_ I think that, now, is useless :).
20:19 Heffalump oh?
20:19 mdiaz_ I was just looking at _darcs/patches.
20:20 mdiaz_ For selecting the patches that I need to keep.
20:20 Heffalump ok, so could that work in conjunction with your existing code to gc individual repos?
20:20 Heffalump i.e. first gc the repos, then use _darcs/patches to gc the global cache?
20:23 mdiaz_ Mmm, but I can gc global cache while gc the individual repos.
20:23 mdiaz_ right?
20:23 mdiaz_ Ah, no...
20:24 Heffalump sorry, I didn't follow that?
20:25 mdiaz_ I mean, I don't need to read _darcs/patches, if a read the inventories.
20:26 Heffalump I was suggesting an alternative approach where you do read _darcs/patches, after GCing the individual repos
20:26 Heffalump would that be a viable alternative to the algorithm I suggested above?
20:27 edwardk joined #darcs
20:28 mdiaz_ But in that way we're reading the patches twice, right?
20:29 Heffalump it would just be the names, wouldn't it?
20:29 Heffalump i.e. just listing the directory, not actually opening the files
20:30 mdiaz_ yes
20:30 Heffalump I don't think that would hurt too much
20:30 Heffalump in theory it could be cached in RAM, I guess, i.e. accumulate the set of patches to keep at the same time as GCing each repository
20:32 mdiaz_ Ok, I'll see how to do it.
20:33 Heffalump ok, onto other things
20:33 Heffalump you're very overdue for a blogpost - could you do that today/tomorrow?
20:33 mdiaz_ Ok.
20:34 Heffalump you should also get the bucketed cache patch closed out. Did you followup on Guillaume's latest comments/
20:34 * Heffalump whacks his shift key, which always seems to fail when I'm trying to type a '?'
20:37 amgarchIn9 joined #darcs
20:38 mdiaz_ Yes, I think that I just need to add a check to a  file.
20:40 Heffalump ok. I was also wondering if you've thought about tests for that code?
20:42 favonia joined #darcs
20:42 mdiaz_ Yes, but I first wanted to install darcs in windows. I'll need your help for that.
20:43 mdiaz_ I was also waiting for your comments for that patch.
20:43 Heffalump Do you have the prerequisite (Windows? :-)
20:43 mdiaz_ Yes, I've windows :).
20:44 Heffalump since Guillaume has reviewed it, it's fine, but tests would be a good idea, perhaps at some later date.
20:44 Heffalump ok. Have you tried to install darcs yet? Did anything go wrong if you did?
20:44 mdiaz_ When I tried to install,  it complained about some missing dll.
20:44 mdiaz_ But I'll try again.
20:46 Heffalump ok, so you got as far as running cabal install, and that failed?
20:47 mdiaz_ Maybe it would help me if you update the wiki :).
20:47 Heffalump which page?
20:48 mdiaz_ I'll try these days, and I'll give you the exact error.
20:48 Heffalump the most likely problem is that it won't find curl. The easiest workaround is simply to disable it (and have it build against the Haskell HTTP library instead) with -f-curl.
20:48 mdiaz_ BuildingUnderWindows
20:49 Heffalump ok, the page already suggests -f-curl at the top
20:49 Heffalump actually, I think also that some of the latest patches in screened don't work, so if you get an error about zip-archive, try removing the patch that added it
20:52 mdiaz_ Oki.
20:53 Heffalump are you keeping a list of things to do, btw?
20:53 Heffalump it might be useful for keeping track of all the outstanding stuff
20:55 mdiaz_ Yep, I've one.
20:55 mdiaz_ I'll continue to work, I want to finish this soon.
20:56 Heffalump ok. I'll look forward to your blogpost, and talk to you next week.
20:59 mdiaz_ Ok, I think we'll need to talk before (like Friday), but I'll tell you by email.
20:59 Heffalump ok, definitely happy to talk as often as you need
21:00 mdiaz_ :), thank you.
21:00 mdiaz_ Bye.
21:00 Heffalump this time is generally a good one for me, too. Bye.
21:50 dolio joined #darcs
22:27 dolio joined #darcs
22:39 edwardk joined #darcs
23:04 Rastus_Vernon joined #darcs
23:09 mizu_no_oto joined #darcs
23:10 mizu_no_oto joined #darcs
23:22 mizu_no_oto joined #darcs
23:24 mizu_no_oto joined #darcs
23:29 edwardk joined #darcs

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