|
02:27
releasable6 joined
02:55
ilbot3 joined
02:57
greppable6 joined
02:58
greppable6 joined
02:59
greppable6 joined
03:53
MasterDuke_ joined
06:10
AlexDani` joined
11:21
greppable6 joined
11:44
releasable6 joined
|
|||
| nwc10 | OK, this is nuts. It used to be that it (Rakudo I think) would reliably SEGV, caused it seemed by box_i | 11:46 | |
| for referece, commit dd72e1f1ecc16ce75e78db3c813bb6cf43b574ba -- Disable box_i template for the time being | |||
| if I revert that, life no longer SEGVs | |||
| OK | |||
| so, if I check out dd72e1f1ecc16ce75e78db3c813bb6cf43b574ba^ | |||
| and then wind NQP back to the last commit that will build on MoarVM | |||
| and Rakudo back to the last commit that will build on that NQP | 11:47 | ||
| it *now* doesn't SEGV | |||
| but surely that was approximately the setup that used to SEGV reliably | |||
| this doesn't make sense | |||
| jnthn's fix in commit cf523c89c06a0f480ec43a8adbc54d0ff453ebee *would* have made sense. | 11:48 | ||
| But it does not seem to be that | |||
|
11:55
mst_ joined
12:04
releasable6 joined
13:05
robertle joined
13:34
brrt joined
|
|||
| brrt | good * | 13:35 | |
| yoleaux | 16 Feb 2018 19:20Z <samcv> brrt: that sounds like a good idea | ||
| brrt | nwc10: that.. sucks... | 13:36 | |
| because we can't repro the original issue then | |||
| :-( | |||
| i kind of also want to push the template for getlexstatic_o | 13:37 | ||
| since it now works (for me) | |||
| dogbert17 | o/ | 13:44 | |
| dogbert17 clickbaits github.com/MoarVM/MoarVM/issues/804 and hides ... | |||
| MasterDuke | brrt: fwiw, i just tried enabling box_i and getlexstatic_o, but nqp segfaults during the build | 13:48 | |
| works with just box_i enabled | 13:50 | ||
| brrt | ah | 13:53 | |
| on linux? | |||
| I'm on a mac and i'm not getting it :-( | |||
| but my suspicion is that it is threading-dependent | |||
| \o dogbert17, MasterDuke | 13:54 | ||
| MasterDuke | yeah, this is kubuntu 17.10 | ||
| brrt | oh, blimey | ||
| hmm | |||
| MasterDuke | also gets through a rakudo build with just box_i enabled | 13:55 | |
| trying now with just getlexstatic_i enabled | 13:56 | ||
| nope, nqp segfaults just about right away | |||
| brrt | yeah | 13:57 | |
| hmm | |||
| don't do that then, yet :-P | 13:59 | ||
| no, in seriousness, we should probably put it in a branch | |||
| i was doing that but you can as well, if you want | |||
| (i'm heading out in 30 minutes or so) | |||
| thanks dogbert17 | 14:00 | ||
| MasterDuke | brrt: here's a backtrace of the segfault i get gist.github.com/MasterDuke17/9f386...5fbab53376 | 14:01 | |
| hm, the segv doesn't seem to trigger under valgrind | 14:07 | ||
| ah, got one | 14:08 | ||
| gist updated with gdb and valgrind output | 14:09 | ||
| brrt | dogbert17: not bisectable :-( | 14:20 | |
| oh, spesh bisectable though | 14:21 | ||
| MasterDuke: that affirms what I found earlier | 14:26 | ||
| *confirms | |||
| so it's going wrong in the call to MVM_spesh_log_static | |||
| my suspicion is now, that the call to MVM_spesh_is_logging, returns a zero value that is smaller than 32 bit | 14:27 | ||
| MasterDuke | `MVMSpeshLogEntry *entry = &(sl->body.entries[sl->body.used]);` is the line it segfaults on | 14:29 | |
| and sl is 0 | |||
|
14:29
lizmat joined
|
|||
| brrt | this is because sl = 0 :-) | 14:29 | |
| timotimo | you mean the "is_logging" function is set to return, like, an 8bit integer and the compiler expects any caller to ignore everything past the 8th bit? | 14:54 | |
| MasterDuke | MVMint32 MVM_spesh_log_is_logging(MVMThreadContext *tc) | 15:04 | |
| github.com/MoarVM/MoarVM/blob/mast...e.expr#L10 is `(do (when (nz (call (^func &MVM_spesh_log_is_logging)`, think that's the problem? | 15:05 | ||
| Geth | MoarVM: jstuder-gh++ created pull request #805: Rakudo RT#102994: Add flag indicating HLL init status on StaticFrame |
15:24 | |
| synopsebot | RT#102994 [open]: rt.perl.org/Ticket/Display.html?id=102994 [BUG] State variables are never initialized if the first call of the containing block does not reach them | ||
|
15:31
AlexDani` joined
16:44
greppable6 joined
17:21
zakharyas joined
|
|||
| Geth | MoarVM: 720705d253 | (Timo Paulssen)++ | src/6model/reprs/P6opaque.c don't optimize getattr unless known concrete otherwise a loop of calling an accessor on a type object will segfault once it gets speshed. Fixes R#1540 |
17:33 | |
| synopsebot | R#1540 [open]: github.com/rakudo/rakudo/issues/1540 [SEGV] Some snippet that segfaults | ||
|
17:41
zakharyas joined
|
|||
| Geth | MoarVM: 8d94732aad | (Zoffix Znet)++ | src/io/syncfile.c Fix eof detection on zero-size files Including special files, like those in /proc. Fixes R#1533 github.com/rakudo/rakudo/issues/1533 When we detect seek pos <= file size, we go to check if file size is zero. Both of those are true on zero-size files and /proc files, so we always report EOF for those files, which breaks some read operations. Fix by falling back on the eof_reported flag for such files. |
18:12 | |
| synopsebot | R#1533 [open]: github.com/rakudo/rakudo/issues/1533 [RFC][regression][⚠ blocker ⚠] /proc/…/status can no longer be read line by line | ||
|
19:10
zakharyas joined
19:31
zakharyas joined
19:37
brrt joined
19:53
releasable6 joined
21:37
brrt joined
21:52
AlexDaniel joined
23:37
brrt joined
|
|||