Welcome to the main channel on the development of MoarVM, a virtual machine for NQP and Rakudo (moarvm.org). This channel is being logged for historical purposes. Set by lizmat on 24 May 2021. |
|||
00:02
reportable6 left
00:04
reportable6 joined
00:24
jgaz left
01:26
MasterDuke joined
03:05
evalable6 left,
linkable6 left
03:06
evalable6 joined
03:07
linkable6 joined
04:34
evalable6 left,
linkable6 left,
linkable6 joined,
evalable6 joined
06:02
reportable6 left
06:03
reportable6 joined
07:03
linkable6 left
07:04
linkable6 joined
|
|||
nine | dogbert17: ah, yes, this cannot really work as such: MVM_args_slurpy_positional(tc, &tc->cur_frame->params, GET_UI16(cur_op, 2)) | 08:02 | |
Since cur_frame could be moved by the GC, that pointer into the structure can get outdated | 08:03 | ||
Geth | MoarVM/new-disp: 539991e410 | (Stefan Seifert)++ | 4 files Fix ctx pointing at fromspace in MVM_args_slurpy_positional tc->cur_frame can be moved by the GC, so a pointer into this structure can get outdated which is what happens when run with GC_DEBUG=3. Since there are only two callers anyway, make the ctx arg optional and fall back to &tc->cur_frame->params if a NULL was passed. This way we always get a fresh pointer. |
08:19 | |
08:45
linkable6 left,
evalable6 left
08:46
evalable6 joined
08:48
linkable6 joined
10:00
squashable6 left
10:03
squashable6 joined
11:03
linkable6 left
11:05
linkable6 joined
|
|||
dogbert17 | nine++, that was fast :) | 11:22 | |
there seems to be at least one more GC related bug lurking somewhere in the new-disp branch | 11:23 | ||
here's what it looks like: gist.github.com/dogbert17/e85f9f02...fe9dfa68c9 | 11:27 | ||
hmm, there's a function called MVM_args_slurpy_named in the stacktrace | 11:36 | ||
12:02
reportable6 left
12:05
reportable6 joined
12:20
lizmat_ joined
12:21
cognominal_ joined
12:22
TempIRCLogger left
12:23
lizmat left
12:25
cognominal left
12:27
lizmat_ left
12:28
lizmat joined
|
|||
nine | Could be a similar thing | 12:41 | |
Didn't have time to check. Visiting parents | |||
dogbert17 | Have a nice visit. | 14:00 | |
Geth | MoarVM: coke++ created pull request #1515: Switch from freenode to libera |
17:14 | |
MoarVM: a0d05b4a81 | Coke++ | 4 files Switch from freenode to libera * remove unhelpful comment * yes, macports is also moved to libera |
17:48 | ||
MoarVM: fb5d26a7e3 | MasterDuke17++ (committed using GitHub Web editor) | 4 files Merge pull request #1515 from coke/libera Switch from freenode to libera |
|||
18:03
reportable6 left,
reportable6 joined
18:17
RakuIRCLogger left
18:18
jgaz joined
18:36
jgaz left
|
|||
[Coke] | MasterDuke++ thanks | 18:50 | |
MasterDuke | [Coke]++ i was surprised when i saw the pr, i thought it had been done already | 18:54 | |
tellable6 | 2021-07-17T23:05:04Z #raku <lizmat> MasterDuke: thanks for the worries, though :-) | ||
[Coke] | ... I was suprised when I went to github and saw it wasn't also! :) | 18:55 | |
afk | |||
nine | dogbert17: this is interesting! Apparently that MVM_args_slurpy_positional issue is not particular to new-disp. In github.com/MoarVM/MoarVM/commit/e0bc6e2f1c I already fixed the same issue in MVM_args_slurpy_named | 19:20 | |
That's 2 years ago | |||
dogbert17 | a mystery | 19:23 | |
nine | In so far as it's surprising that we did not hit that MVM_args_slurpy_positional issue sooner. And that it came up so easily now | ||
dogbert17 | aha | ||
nine | Still there's also the "Unhandled exception: Can only use manipulate a capture known in this dispatch" that comes up now as soon as GC_DEBUG=3 is active | 19:28 | |
dogbert17 | so we have 2-3 bugs at the very least | 19:39 | |
timo | quite possibly the same kind of missing mvmroot in one of the boot dispatchers as the one the other day | 19:43 | |
nine | got it | 19:44 | |
Geth | MoarVM/new-disp: 0c8538cac4 | (Stefan Seifert)++ | src/disp/boot.c Fix possible access to fromspace in boot_code We need to protect the args_capture pointer to the newly allocated capture because MVM_disp_program_record_track_arg will allocate. |
19:45 | |
nine | With this t/01-sanity/01-literals.t now passes with GC_DEBUG=3 | 19:46 | |
dogbert17 | ah, the one timo found yesterday | 19:52 | |
nine | Oh boy, even with the 16 core Zen 3 chip, make test with GC_DEBUG=3 takes forever | 20:10 | |
Good news is that so far that's 21 files passing and no failures or segfaults | 20:11 | ||
dogbert17 | impressive | 20:16 | |
21:48
Geth left
21:49
lizmat left
21:50
lizmat joined
22:25
squashable6 left
22:27
squashable6 joined
22:35
squashable6 left
22:37
squashable6 joined
22:38
squashable6 left
22:39
squashable6 joined
|