4 Dec 2023 |
geekosaur | Redacted or Malformed Event | 02:53:46 |
geekosaur | (method definitions for other types are optional, but are implemented as calling toList on the value and then using the list implementation, so the list implementations are mandatory) | 02:56:09 |
geekosaur | * before Foldable , GHC.OldList was Data.List | 03:34:00 |
geekosaur | * so it was renamed to Data.OldList and the instance Foldable [a] where… added to it | 03:34:11 |
geekosaur | * so it was renamed to GHC.OldList and the instance Foldable [a] where… added to it | 03:34:22 |
geekosaur | * so it was renamed to GHC.OldList and the instance Foldable [] where… added to it | 03:49:58 |
| lainon joined the room. | 06:40:22 |
| tfaz.xyz changed their profile picture. | 10:29:07 |
| tfaz.xyz changed their profile picture. | 10:29:45 |
Profpatsch | can we please, please, please monomorphize Data.List functions again | 10:53:21 |
Profpatsch | We can still export the overloaded ones from Prelude for all I care | 10:53:32 |
Profpatsch | this situation is untenable | 10:54:21 |
teo (they/he) | @Profpatsch : last I heard this was the current blocker https://gitlab.haskell.org/ghc/ghc/-/issues/21791. There's been a lot of ghc-proposal Stability talk recently though so maybe stuff has changed | 10:54:52 |
Profpatsch | I’d really love to see this | 11:01:35 |
Profpatsch | Container modules should always export functions that are as monomorphic as possible | 11:02:02 |
Profpatsch | this really helps with production code & avoiding bugs due to unwanted overloading | 11:02:21 |
ic.rbow | are there any conditions when forkFinally may not fire? | 14:22:10 |
ic.rbow | * are there any conditions when forkFinally callback may not fire? | 14:22:18 |
| sebrut left the room. | 14:43:14 |
@21it:matrix.org | In reply to @ic.rbow:matrix.org are there any conditions when forkFinally callback may not fire? Maybe some unmasked/async exceptions? I bet there should be a lot of scenarious where it's not working as expected | 18:41:54 |
geekosaur | a key one is if the main thread terminates then child threads are simply killed by the OS | 18:44:25 |
| reddorf joined the room. | 19:12:55 |
ic.rbow | I try to use forkFinally to shoot off threads for accept d sockets. After accepting I increment a counter, and decrement it back in forkFinally handler (and close that connection). Most of the time everything is okay, but after a while, a gaggle of sockets remain lingering in "established" state and the counter matches the number of such sockets. I would just go fix my network setup if only the OS were to show the leak. But with the counter to back it, it surely looks like forkFinally handler didn't fire (both the counter didn't decrement and a connection didn't go down). | 19:26:22 |
| * geekosaur wonders if
The newly created thread has an exception handler that discards the exceptions BlockedIndefinitelyOnMVar , BlockedIndefinitelyOnSTM , and ThreadKilled , and passes all other exceptions to the uncaught exception handler.
means the finalizer won't be called on any of those. (although that would be odd given ThreadKilled )
| 19:33:06 |
| * geekosaur * wonders if
The newly created thread has an exception handler that discards the exceptions BlockedIndefinitelyOnMVar , BlockedIndefinitelyOnSTM , and ThreadKilled , and passes all other exceptions to the uncaught exception handler.
means the finalizer won't be called on any of those. (although that would be odd given ThreadKilled ) | 19:36:23 |
geekosaur | anyway I'd ask over in #GHC:matrix.org | 19:45:20 |
ic.rbow | In reply to * @geekosaur:matrix.org
wonders if
The newly created thread has an exception handler that discards the exceptions BlockedIndefinitelyOnMVar , BlockedIndefinitelyOnSTM , and ThreadKilled , and passes all other exceptions to the uncaught exception handler.
means the finalizer won't be called on any of those. (although that would be odd given ThreadKilled ) I think those would be caught in the try inside forkFinally | 19:47:36 |
sm | oh wow.. I didn't know pandoc supported completely custom readers as lua plugins (for 2 years) | 22:10:01 |
sm | * oh wow.. I didn't know pandoc supported completely custom readers as lua plugins (for 2 years)
https://pandoc.org/custom-readers.html | 22:10:23 |
5 Dec 2023 |
| Federico Damián Schonborn changed their profile picture. | 00:38:08 |