Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2015-12-05

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

All times shown according to UTC.

Time Nick Message
00:44 Big_G joined #darcs
01:37 Riastradh joined #darcs
04:42 mizu_no_oto joined #darcs
05:05 mizu_no_oto joined #darcs
11:30 seumiremorque joined #darcs
17:24 Big_G joined #darcs
18:14 Riastradh joined #darcs
22:13 maerwald sm, Heffalump: in snap they said "given that without -threaded you still get cooperative concurrency I don't see how that is any safer" wrt darcs relying on $cwd, so it seems we are already relying on unsafe behavior
22:15 dixie joined #darcs
22:22 Heffalump Do we actually fork off multiple threads anywhere?
22:23 Heffalump Don't get me wrong, I think the situation is a mess and dangerous. I just don't have any simple solutions.
22:24 sm maerwald: sorry, can't chat now, but building without -threaded definitely fixed a class of problems we were seeing
22:24 Heffalump I spent quite a while a couple of years back trying to implement the "overload over relative or absolute paths" idea by making a custom monad for Darcs IO operations to run in. But it foundered on integration with other things, and in places where Darcs itself forks off other threads.
22:26 maerwald Heffalump: that's probably done by snap, not by us
22:27 Heffalump as sm says, it seems that things work out ok in practice without -threaded, and not with it
22:27 Heffalump anyway, I mostly feel that switching to absolute paths is probably fine, but we need to see what the actual performance hit is
22:27 maerwald I have had directory errors on hub.darcs.net too
22:28 maerwald maybe without threaded we are not exposing the problem that much, but it's not gone
22:29 dolio It's not even cooperative, it's still preemptive.
22:29 * sm feels any current errors can be traced to other causes
22:29 dolio There are just never two threads running simultaneously.
22:30 maerwald sm: not sure I want to rely on that feeling ;)
22:31 sm test on your instance and get your own feelings :)
22:32 maerwald I don't rely on feelings
22:32 maerwald it's not clear what the code does currently, that's bad
22:33 maerwald that's why we use haskell, no? To be clear what the code does ;)
22:40 sm sure
22:41 sm got to go, later all o/

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