Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2014-05-12

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

All times shown according to UTC.

Time Nick Message
00:13 pgiarrusso left #darcs
00:14 whaletechno joined #darcs
00:36 kowey joined #darcs
00:50 Igloo_ joined #darcs
00:50 dino-_ joined #darcs
01:28 kowey joined #darcs
01:51 sm_ joined #darcs
01:51 mizu_no_oto joined #darcs
04:17 vikraman joined #darcs
04:17 carter_ joined #darcs
04:33 carter joined #darcs
04:56 carter joined #darcs
05:09 Heffalump owst: from what I recall, that whole phase of discussions was to do with representing conflicts as dags, not the entire patch graph. It turned into conflictors. One of the objections was Igloo's nasty (ABC)(XYZ) example which (from memory) has A B C all commuting, X Y Z all conflicting, and X conflicting with A and depending on BC, Y conflicting with B and depending on AC, Z conflicting with C and depending on AB.
05:10 Heffalump That example shows you can't represent X Y and Z directly in their correct contexts without duplicating A B or C.
05:11 Heffalump the idea of graphictors that I've been throwing around more recently is fairly similar but it says that you always represent A B C in minimal context form (relative to the base of the conflict) and so you have to merge them to calculate the correct context for X Y and Z.
05:45 edwardk joined #darcs
06:13 lelit joined #darcs
06:44 favonia joined #darcs
06:55 slyfox joined #darcs
06:55 slyfox joined #darcs
07:37 hvr I'm trying to 'darcs get http://code.haskell.org/~ross/mtl'
07:37 hvr but somehow the resulting checkout seems broken
07:37 hvr does anybody know what's up here?
07:39 hvr (I'm using '2.9.8 (+ 36 patches)')
08:06 edwardk joined #darcs
08:34 owst joined #darcs
08:35 owst Heffalump: interesting. The log/emails I found seems that David was pro representing the whole patch graph as a DAG, but I haven't got to the emails/log where he dismisses that idea yet :-)
08:35 owst The link I pasted was meant to be a link to something like Haskell DAG using GADTs, I wondered if you got anywhere with that?
08:40 owst I'm not sure I see why that example doesn't allow you to represent a b c uniquely - something like this: http://i.imgur.com/pnsFNr5.jpg where certain node sets can be marked as conflicting (distinguished using different shapes, here)
08:43 owst To be honest, I haven't thought about how resolutions work yet, but presumably you would want to do something like: solve z,c, call it R(z,c) and then solve a,x - R(a,x), then y, b - R(y, b) and finally something like R(a,x), R(y, b) and R(z, c) together, since those resolutions might somehow "inherit" the conflict set x,y,z (empty circle in my diag). Again, I've not really thought about it :-)
08:43 owst And I have no doubt I am and will be re-hashing things that were discussed a long time ago.
08:44 owst Anyway, this thesis won't write itself. I'll be back around this evening to discuss again.
08:45 owst And finally, I'm not sure what you mean by needing to merge ABC to get a context for XYZ - aren't X Y Z all individual patches? Surely they each have an indepentent context that can be satisfied, something like my diagram (which admitedly isn't using any well-defined notation ;-)) Or does XYZ exist only as a whole?
08:45 * owst gone
08:50 gal_bolle joined #darcs
09:25 LnL joined #darcs
10:36 edwardk_ joined #darcs
11:59 edwardk joined #darcs
12:02 haasn joined #darcs
12:16 Heffalump hvr: try rolling back to 2.9.8 or 2.9.7 and using GHC 7.6, there are some issues with HEAD caused by attempts to fix it for GHC 7.8
12:17 hvr Heffalump: I see... I managed to get a working clone with an old darcs-2.8.4
12:23 Heffalump it is the current release version :-)
12:33 mizu_no_oto joined #darcs
12:59 mizu_no_oto joined #darcs
13:11 c74d joined #darcs
14:25 slyfox^w_ joined #darcs
14:27 mizu_no_oto joined #darcs
14:38 lelit joined #darcs
16:36 mizu_no_oto joined #darcs
17:05 Heffalump owst: Do you mean the IRC logs link you posted where I said I should try to write something like that? I had a go at doing it type safely but it was pretty messy and I'm not sure if it worked. If I did it again e.g. for graphictosr I'd eschew so much type safety in exchange for a clean/small core implementation
17:05 Heffalump You're right, the general idea was to have a DAG of patches, but DAGginess would only have arisen in the case of a conflicts and most branches would have been capped off as "dead ends", from what I recall
17:06 Heffalump the problem with reprsetning a, b, c uniquely comes about if you want x, y, z to be in their correct contexts with a chain of patches leading up to them
17:07 Heffalump if you have a dag like yours it's fine
17:24 alexei joined #darcs
18:21 gh_ joined #darcs
18:32 gh_ Heffalump, about http://irclog.perlgeek.de/darcs/2014-05-12#i_8711792  , shall we rollback the culprit patch?
18:43 Heffalump I'm not really sure what the best approach is. I haven't had time to investigate any new failures caused by slyfox's patch. Someone mentioned he was looking at fixing them but I haven't heard any more since then.
18:47 slyfox i'll try to investigate 'darcs record' oddities today/tomorrow and report back
18:47 gh_ okay
18:47 slyfox i need some time to understand how files from inventory/ get to patches/
18:48 slyfox (for some reason the same code use used as for cache filling even files are guaranteed not to present outside)
18:49 slyfox (and a side note: 'darcs record' does not respect --no-cache)
19:16 alexei joined #darcs
20:47 lelix joined #darcs
21:49 owst joined #darcs

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