|
02:04
vendethiel joined
02:48
ilbot3 joined
02:55
FROGGS_ joined
04:32
TimToady joined
07:37
FROGGS joined
08:39
zakharyas joined
08:52
vendethiel joined
|
|||
| dalek | arVM: 7d0e60f | FROGGS++ | src/io/procops.c: plug memory leak when shelling out without capturing |
09:11 | |
| FROGGS | jnthn / hoelzro: ^^ | ||
|
09:18
zakharyas joined
10:00
brrt joined
|
|||
| brrt | timotimo: thanks for that link on LLVM | 10:07 | |
| was interesting | |||
|
10:30
vendethiel joined
11:47
brrt joined
11:53
vendethiel joined
12:39
vendethiel joined
|
|||
| timotimo | i'm glad :) | 12:44 | |
|
14:19
vendethiel joined
15:43
brrt joined
16:34
vendethiel joined
17:04
vendethiel joined
17:05
FROGGS joined,
zakharyas joined
17:29
vendethiel joined
18:36
colomon joined
18:40
vendethiel joined
18:49
FROGGS[mobile] joined
18:57
FROGGS[mobile] joined
18:58
FROGGS[mobile]2 joined
|
|||
| FROGGS[mobile]2 | o/ | 18:59 | |
| [Coke] | Getting a hang on non-jit moar at rakudo/nom and roast/master for t/spec/S32-io/IO-Socket-Async.t on hack.p6c.org | 19:15 | |
|
19:16
patrickz joined
|
|||
| jnthn | FROGGS: Your patch fixed two of the three leaking t/nqp tests | 20:37 | |
| 86-pipes.t still has 3KB worth of leaks | 20:38 | ||
| qregex.t seems to take forever under valgrind, so skipping that for now | 20:44 | ||
| t/hll/*.t is clean. t/qast/*.t also. | |||
| t/moar/*.t has some leaks | 20:45 | ||
| timotimo | qregex.t is a pretty huge | ||
| jnthn | Yeah | ||
| timotimo | i think it does an eval per regex that comes in via the declarative file? | ||
| jnthn | yeah | 20:46 | |
| timotimo | that probably does a gigantic amount of churn memory-wise | ||
| jnthn | I may try it with cut down input | ||
| But there's other leaks "on the table" first | |||
| timotimo | good good :) | 20:49 | |
| jnthn | ooh, naughty multidim code | 20:53 | |
| dalek | arVM: 35d5ff1 | jnthn++ | src/ (3 files): Fix leak on out-of-bounds multi-dim access. In-bounds ones never leaked. Along the way, improve performance by avoiding a load of malloc/free churn. While most real-life accesses should eventually go through the cheaper 2D/3D special-case APIs, this improves things for the remaining cases, and in the interim. |
21:06 | |
| arVM: 0e96e1c | jnthn++ | src/core/ (3 files): Make sure unused continuation tags get freed. |
21:27 | ||
|
21:28
colomon joined
|
|||
| jnthn | Turns out disabling the fixed size allocator is a good way to uncover more issues. | 21:34 | |
| timotimo | oh, good point | 21:40 | |
| we don't have leak analysis for that | 21:41 | ||
| dalek | arVM: a4b683e | jnthn++ | src/ (3 files): Fix a couple of NFG cleanup flaws. Free memory allocated with fixed size allocator using it, and make sure we don't trip over the unchanged case sentinel. |
21:52 | |
| arVM: 6f9be70 | jnthn++ | src/ (2 files): Free synthetic combiner storage. Like other NFG leaks, not really interesting for long-lived programs, just missing cleanup at shutdown with --full-cleanup. |
22:02 | ||
| jnthn | Still need to walk/free the trie | ||
| jnthn decides to trie it tomorrow... :) | 22:08 | ||
| jnthn needs to place an order, to fix the "extreme low beer stock" condition here... | |||
| [Coke] | that test is still hung, btw, hours later. The memory/cpu ulimits in place aren't killing it. | ||
| [Coke] remembers beer. :| | 22:09 | ||
| jnthn | [Coke]: I'm guessing it's a deadlock... | ||
| [Coke] | it happens consistently for me on hack sans jit. | ||
| (so I don't know if it happens with jit, because non-jit always runs first. :) | |||
| jnthn | [Coke]: Possibly the really nasty GC orchestration one that I need to get the brane to tackle... | ||
| Noooo...they stopped selling the nice regional breweries ale selection :( | 22:15 | ||
| timotimo | :< | 22:26 | |
|
22:40
vendethiel joined
23:20
vendethiel joined
23:31
ilmari_ joined
23:32
[Coke]_ joined
23:39
khagan joined
23:41
TimToady joined
23:50
synopsebot6 joined
|
|||