|
01:31
MasterDuke joined
02:28
Kaiepi joined
|
|||
| MasterDuke | .tell brrt i spun up an ubuntu 14.04 vm and built my branch, but nqp and rakudo both built just fine | 02:40 | |
| yoleaux | MasterDuke: I'll pass your message to brrt. | ||
| MasterDuke | .tell brrt i take that back. if i do `--optimize=0` i get the same failure | 02:54 | |
| yoleaux | MasterDuke: I'll pass your message to brrt. | ||
|
02:56
ilbot3 joined
|
|||
| MasterDuke | .tell brrt don't need the vm, i get the failure on my laptop and desktop with --optimize=0. some debug info and the jit-bisect data here gist.github.com/MasterDuke17/27feb...15f1af9b17 | 02:56 | |
| yoleaux | MasterDuke: I'll pass your message to brrt. | ||
| samcv | .ask jnthn where do I need to set gc mark? not sure where it is github.com/MoarVM/MoarVM/pull/815#...3675573R40 | 04:04 | |
| yoleaux | samcv: I'll pass your message to jnthn. | ||
|
06:18
domidumont joined
06:27
domidumont joined
06:59
quotable6 joined
07:01
domidumont joined
07:37
domidumont joined,
brrt joined
|
|||
| brrt | good * | 07:38 | |
| yoleaux | 02:40Z <MasterDuke> brrt: i spun up an ubuntu 14.04 vm and built my branch, but nqp and rakudo both built just fine | ||
| 02:43Z <MasterDuke> brrt: i take that back. if i do `--optimize=0` i get the same failure | |||
| 02:56Z <MasterDuke> brrt: don't need the vm, i get the failure on my laptop and desktop with --optimize=0. some debug info and the jit-bisect data here gist.github.com/MasterDuke17/27feb...15f1af9b17 | |||
| brrt | MasterDuke++ | ||
| nine | So it's an inverse-Heisenberg bug? It appears only when debugging? | ||
| brrt | hehe, that's better right :-) | 07:43 | |
|
07:52
robertle joined
|
|||
| brrt | also nice, is that it breaks quite fast | 07:53 | |
|
08:35
FROGGS joined
08:42
lizmat joined
08:46
brrt joined
|
|||
| Kaiepi | how would you force moar to crash instead of allowing MVM_panic to bubble back up to rakudo? | 08:58 | |
| er, MVM_throw_exception_throw_addhoc` | 08:59 | ||
| *MVM_throw_exception_adhoc | 09:00 | ||
| agh **MVM_exception_throw_adhoc | 09:01 | ||
| there are some errors in IO::Socket::Async that look like they could be a problem, but i don't know how to force it to coredump | 09:02 | ||
|
09:19
FROGGS joined
|
|||
| brrt | Kaiepi: MVM_oops | 09:44 | |
| or | |||
| MVM_panic | |||
| MVM_oops is for when the stack is not corrupted, but there's still an unrecoverable error | 09:45 | ||
| MVM_panic is when you might expect a memory corruption and theres no hope of recovering | |||
| Kaiepi | thanks | 09:49 | |
| brrt | np | 09:50 | |
|
12:02
domidumont joined
12:25
dogbert17 joined
12:43
domidumont joined
12:46
domidumont joined
13:43
greppable6 joined,
brrt joined
14:05
brrt joined
|
|||
| brrt | i think i found the bug uncovered by MasterDuke++ | 14:25 | |
| well, or at least, the proximate cause; there's nothing emitted for the ZR test | 14:26 | ||
| Geth | MoarVM: ee83e78b68 | (Bart Wiegmans)++ | src/jit/x64/tiles.dasc [JIT] Test with default size of 8 We'd skip on emitting a test if the size of the operand was not given, which would happen whenever we'd have the result of a CALL node. The real bug is that CALL nodes can introduce zero-sized values into the tree, which we should fix. |
14:32 | |
| brrt | anyway, MasterDuke++, pls try travis again on your pull request | ||
|
14:53
releasable6 joined
15:09
AlexDaniel joined
15:32
Kaiepi joined
15:50
Kaiepi joined
15:56
Kaypie joined
16:03
FROGGS joined
16:33
bisectable6 joined
16:43
Kaiepi joined
17:15
Kaiepi joined
18:00
domidumont joined
18:03
domidumont joined
19:12
greppable6 joined
19:21
Kaiepi joined
19:22
Ven`` joined
19:31
Ven`` joined
|
|||
| Geth | MoarVM: timo++ created pull request #817: MoarVM Remote Debugging Support |
22:27 | |
| timotimo | oh look what's this | 22:29 | |
| jnthn | Oh my goodness, look at that :) | 22:30 | |
| 6guts.wordpress.com/2018/03/13/rem...or-moarvm/ :) | |||
| lizmat | wow! :-) | 22:33 | |
| yoleaux | 15:32Z <Zoffix> lizmat: FYI, I listed you as stakeholder (github.com/perl6/6.d-prep/commit/ce7fab350a) for nixing Pair.freeze in 6.d (per github.com/rakudo/rakudo/commit/c229022cb0941). Implementing that is currently blocked by R#1289 tho | ||
| synopsebot | R#1289 [open]: github.com/rakudo/rakudo/issues/1289 [6.d][severe] [6.d BLOCKER] Implement a Way to Know Caller's Language | ||
| jnthn | Blog post about it here: 6guts.wordpress.com/2018/03/13/rem...or-moarvm/ | ||
| oh, I already linked it :D | |||
| Clearly too execited :) | |||
| To type, even :) | |||
| lizmat | :-) | 22:43 | |
| jnthn: I was looking at the ThreadPoolScheduler code wrt to number of worker threads | |||
| I found that test-t --race typically uses 3 workers only | 22:44 | ||
| that's because at each interval at least 1 worker finished something, afaics | |||
| MasterDuke | brrt++ looks like his commit fixed my PR | 22:45 | |
| and timotimo++, jnthn++ for remote debugging | 22:46 | ||
| lizmat | jnthn: that feels like a wrong heuristic to me, is there a reason for that? | 22:47 | |
| jnthn: fwiw, blog post now also mentioned on Perl 6 FB group | |||
| courtesy of woolfy | |||
| timotimo | MasterDuke: couldn't have done it without jnthn's guidance :) | 22:51 | |
| timotimo does the GUI dance | |||
| lizmat | whee! | ||
| jnthn timotimo: so when will the PR be merged and revisions bumped ? | 22:53 | ||
| MasterDuke | timotimo: are you working at/for Edument now? | 23:01 | |
|
23:01
robertle joined
|
|||
| jnthn | lizmat: Are you sure that's not because :$degree defaults to 4? | 23:06 | |
| lizmat | wouldn't that imply 4 workers then ? | ||
| jnthn | lizmat: Hmm, good point | 23:18 | |
| Did you try passing :degree to see what happens? | |||
| lizmat | will do so tomorrow | ||
| samcv | jnthn: how do I gc mark this? github.com/MoarVM/MoarVM/pull/815#...3675573R40 | ||
| lizmat | jnthn: and get back to you | ||
| jnthn | As to the heuristic not making perfect decisions - of course, it's a heuristic. ;) We can explore better ones though :) | ||
| samcv: Well, directly or indirectly from gc_mark of the Decoder REPR | 23:19 | ||
| lizmat | ok, I will do that... have had some interesting effects with increasing the sleep gradually to .1 second | ||
| jnthn | Yeah, the 0.01 was picked for a reason :) | ||
| lizmat | well, the thing is that for test-t --race 20, it only made a small difference for total wallclock, but 10% less overall CPU | 23:20 | |
| jnthn | samcv: github.com/MoarVM/MoarVM/blob/mast...oder.c#L26 is currently empty | 23:21 | |
| samcv: But it'd want to be handled in there | |||
| samcv | so i do MVM_gc_worklist_add then? | ||
| jnthn | Yup | ||
| But also | |||
| Make sure you use MVM_ASSIGN_REF when storing the replacement too | 23:22 | ||
| To make sure we don't violate the inter-gen constraint | |||
| samcv | ok thanks | ||
| timotimo | lizmat: i'd like it to be in the upcoming release, which is Real Soon Now | 23:41 | |
| samcv | doing this makes MoarVM segfault. not sure if i'm doing it right? gist.github.com/42b9e67a199316656b...c0431dd5a4 | 23:47 | |
| in Decoder.c | |||
| timotimo | hm, i assume gc_mark is also called for non-concrete objects | ||
| jnthn | samcv: Cast to MVMDecoderbody | 23:48 | |
| *MVMDecoderBody | |||
| samcv | ah ok thanks | 23:49 | |
| jnthn | Also maybe guard against ds being null | ||
| About the remote debug PR - I can merge it tomorrow (I already gave it a good reviewing), if there aren't any comments against that from others. | 23:50 | ||
| timotimo | Undefined symbols for architecture x86_64: | 23:56 | |
| "_htobe16", referenced from: | |||
| _send_greeting in debugserver.o | |||
| ld: symbol(s) not found for architecture x86_64 | |||
| what fun! | |||