github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
02:04 klapperl_ joined 02:05 klapperl left 03:01 leedo left 03:03 leedo joined
nwc10 good *, #moarvm 06:19
06:25 klapperl_ left 06:57 domidumont joined 07:53 Kaiepi left 07:57 Kaiepi joined
MasterDuke am i correct in assuming you can do a hash bind of an entry, and then fill in the entries members? or do they have to be populated first? 08:08
nwc10 either order works with UTHash
MasterDuke good, thanks 08:09
08:17 Altai-man_ joined 08:27 sena_kun joined 08:29 Altai-man_ left 08:42 klapperl joined 10:16 Kaeipi joined, Kaiepi left 10:26 Altai-man_ joined 10:29 sena_kun left 11:19 konvertex joined 11:52 Geth left, Geth joined 12:27 sena_kun joined 12:29 Altai-man_ left
jnthn So...that pesky segv from Friday... 13:00
nwc10: Interesting that paste.scsys.co.uk/589021 blows up after test 6 with ASAN, but valgrind...uh, hmm, exists with segfault instead of reporting a memory error...wat 13:03
And not until after test 14.
Also On My Machine ASAN also only blows up after 14 13:06
Though the trace is useful
goodness, something is very busted... 13:18
MasterDuke if anyone is interested, gist.github.com/MasterDuke17/7844a...c6f6bc4e4e is the segv i get on my branch that adds lots of frees before throws. the repl errors on master with `Unhandled lexical type 'num32' in lexprimspec` (the test is todo'ed), but doesn't segv 13:38
13:43 Merfont joined
Geth MoarVM/new-disp: 5676b70c85 | (Jonathan Worthington)++ | 3 files
Fix bytecode/C function/bytecode dispatch sandwich

Need to make sure we properly track when we thunked, so that we don't scribble over the invoked frame with a place we thought we were returning to.
13:44
MasterDuke well, that turns out to have been easy to diagnose and fix. don't know why i missed it earlier (got sidetracked thinking it was that change in container.c), but i was freeing something that hadn't been alloced 13:45
13:46 Kaeipi left
jnthn OK, now I need to go for the guards, I guess... 13:50
MasterDuke and everything tests clean. wonder if i remember where in the list of stuff the plugin found i was 13:51
nwc10 jnthn: All tests successful. 13:57
MasterDuke oh, maybe i was at the end except for the MVM_HASH_(BIND|GET) stuff (which is now taken care of) and the MVM_ASSIGN_REF stuff (which was decided isn't needed) 14:06
jnthn nwc10: yay 14:08
Updates around gist.github.com/jnthn/e81634dec57a...9#guarding and in capture transformation to reflect latest design tweaks
Now to implement the guarding bits
nwc10 got a SEGV in "Stage optimize" for the setting 14:23
14:26 Altai-man_ joined 14:29 sena_kun left 14:54 Merfont left 14:57 Kaiepi joined
Geth MoarVM/new-disp: 1e45dfb999 | (Jonathan Worthington)++ | 11 files
Bring in tracked values in dispatchers

These will cover values from incoming argument captures, but also later from read attributes, so we get a safe and uniform interface for doing guards. For now, we can just insert such an arg into another capture.
15:08
MoarVM/new-disp: 85d0608404 | (Jonathan Worthington)++ | src/disp/syscall.c
Tweak name of insert literal obj arg, add str too
15:18
15:36 domidumont left
Geth MoarVM/new-disp: ab7ed52a42 | (Jonathan Worthington)++ | 3 files
Stub in all dispatcher guard syscalls
15:40
15:57 domidumont joined 16:19 domidumont left 16:27 sena_kun joined 16:29 Altai-man_ left
Geth MoarVM/new-disp: 8a623816ce | (Jonathan Worthington)++ | 2 files
Sketch out data structures for dispatch recording
16:44
nwc10 jnthn: still a SEGV in core setting - "==16923==Hint: address points to the zero page. 16:45
"
#0 0x7f7f5bb5ef74 in MVM_frame_unwind_to src/core/frame.c:1310
...
jnthn nwc10: Will care later. 16:46
nwc10 OK :-)
that's why I didn't go fine the nopaste URL
jnthn I want to revist a few things in that area anyway, but would rather get some progress in on the dispatch stuff rather than shave another yak :) 16:47
I mean, they have to be shaved eventually, but... :) 16:48
Geth MoarVM/new-disp: 379b01942d | (Jonathan Worthington)++ | src/disp/program.h
Model multi-level capture derivations too
16:51
jnthn Glad to have figured out that data structure design; back to the next steps being a SMOP again.
dinner & 16:58
timotimo /* The index involved in the inert or drop. */ - the dispatch programs are inert? :) 16:59
nine timotimo: well for now ;) 17:19
timotimo keep poking it with a stick until it moves 17:21
nine I'd think poking with a keyboard would be more promising 17:22
18:27 Altai-man_ joined 18:29 sena_kun left 19:54 zakharyas joined 20:28 sena_kun joined 20:29 Altai-man_ left 21:04 zakharyas left
MasterDuke how easy would it be to create a dummy STable to attach a REPR_data to, just for the purpose of calling gc_free_repr_data() with it? 21:18
jnthn Why would that be needed? 21:59
When does repr data preceed the type existing? 22:00
MasterDuke when it's being composed
e.g., github.com/MoarVM/MoarVM/blob/mast...que.c#L876 22:02
i just added a free_repr_data() that's a virtual copy of gc_free_repr_data() but just takes an MVMP6opaqueREPRData 22:03
github.com/MoarVM/MoarVM/pull/1291 will be updated soon if you're curious 22:04
22:05 patrickb joined 22:27 Altai-man_ joined 22:29 sena_kun left 22:47 patrickb left
MasterDuke updated, think it's ready to be looked at 23:08
23:23 MasterDuke left