| IRC logs at
Set by AlexDaniel on 12 June 2018.
00:10 Geth left 00:11 Geth joined 00:15 lucasb left 01:12 dogbert17 left 03:00 squashable6 left 03:02 squashable6 joined 03:04 squashable6 left 03:07 squashable6 joined 06:20 squashable6 left 06:21 squashable6 joined 06:44 squashable6 left, squashable6 joined 07:06 domidumont joined 07:10 domidumont left 07:13 domidumont joined
nine brrt: I'd be grateful for any such hints! My understanding of the JITs control flow is limited 07:15
tellable6 nine, I'll pass your message to brrt
09:05 zakharyas joined 09:41 sena_kun joined 10:09 AlexDani` joined 10:13 AlexDaniel left 10:18 brrt joined 10:53 sena_kun left 11:08 sena_kun joined 11:13 zakharyas left 11:15 ZzZombo left, zakharyas joined 11:35 zakharyas left 11:36 domidumont left 11:37 zakharyas joined 11:55 robertle joined 11:56 brrt left 12:17 zakharyas left 12:33 klapperl left, lucasb joined 12:52 sena_kun left
Guest38485 is everyone working on ? 12:54
13:03 domidumont joined 13:08 sena_kun joined 13:46 zakharyas joined 14:54 sena_kun left
nwc10 Guest38485: domm mentions it every so often on over on 14:59
15:08 sena_kun joined
nine Guest38485: I don't need advent calendar puzzles. MoarVM supplies me with a steady flow of the greatest puzzles and mysteries that want solving :) 15:12
15:58 zakharyas1 joined 15:59 zakharyas left 16:01 robertle left 16:54 sena_kun left 16:55 ZzZombo joined 16:58 domidumont left 17:09 ZzZombo left, ZzZombo joined, dogbert17 joined, sena_kun joined
dogbert17 nine: have you solved the JIT puzzle yet? 17:10
17:13 ZzZombo left, domidumont joined
nine dogbert17: no, still on it 17:14
17:19 zakharyas1 left 17:23 zakharyas joined
dogbert17 nine: any wise comments from jnthn or brrt? 18:27
18:46 zakharyas left 18:53 sena_kun left
nine not yet 18:58
19:07 sena_kun joined 19:19 AlexDani` is now known as AlexDaniel, AlexDaniel left, AlexDaniel joined 19:45 Kaiepi left 19:48 Kaiepi joined 20:10 domidumont left
nine No progress today :/ 20:48
I do have some suspicion...that tc->cur_frame ends up pointing to the wrong frame as it never had a spesh_cand but we're in JITed code anyway 20:51
20:54 sena_kun left
nine That tc->cur_frame gets set by NativeCall's callback handler: tc->cur_frame = backup_cur_frame; 20:56
I.e. that's the frame that called the native code that ran the callback 20:57
21:08 sena_kun joined 21:09 patrickb joined 21:35 MasterDuke joined 22:54 sena_kun left
jnthn nine: Hmm...I wonder if it could be something like a deopt all causes an uninlining, creating a new top frame, and then it restores backup_cur_frame to the wrong thing. That doesn't quite match the symptoms though... 23:44
23:44 MasterDuke left