|
00:19
colomon joined
00:40
vendethiel joined
01:10
vendethiel joined
01:33
colomon joined
01:48
ilbot3 joined
|
|||
| JimmyZ | japhb: I don't know , but msdn.microsoft.com/en-us/library/d...85%29.aspx FYI | 04:10 | |
|
05:58
vendethiel joined
06:26
flaviusb joined
06:50
zakharyas joined
06:52
FROGGS joined
|
|||
| nwc10 | jnthn: what helgrind makes of it all: paste.scsys.co.uk/482933 | 06:52 | |
| FROGGS | o/ | 07:05 | |
| nwc10 | \o | 07:06 | |
|
07:08
Ven joined
|
|||
| jnthn | Helgrind makes so much noise it's hard to extract the false positives from the interesting... | 07:45 | |
| nwc10 | 1381 lines - this time it wasn't massive | 07:46 | |
| anyway, that was more "if it's useful" | |||
| jnthn | *nod* | ||
| nwc10 | there's clearly a problem, as it's repeatable | ||
| but I really don't have any good idea how to dig further. | |||
| jnthn | So, in bed last night while waiting to sleep, I figured out the cheap way to fix the serialization data race, and also that there'd have been a way to trigger it without laziness anyway if you were creative enough. | ||
| nwc10 | good stuff | 07:47 | |
| jnthn | I don't know how much of the other issues that will fix | ||
|
07:47
vendethiel joined
|
|||
| jnthn | I think some of what we're seeing now is the result of previously hidden-by-a-core-under-load things coming to light | 07:48 | |
| The deserialization race is certainly one of those. It's always been there, but rarely showed up | |||
| FROGGS | I also got a segfault in SCRef.c when use'ing a module in en EVAL which was installed into a CURLI | 07:50 | |
| an* | |||
| jnthn | I highly doubt that's the same issue. You need multiple threads to get a race. :) | ||
| FROGGS | ahh | ||
| it failed on GC_WORKLIST_ADD IIRC | |||
| jnthn | there's a debugging mode for that | 07:51 | |
| FROGGS | yeah, need to look at that again, since now is post-release | 07:52 | |
| (I took the weekend off since the release was kinda painful) | |||
| jnthn | (time off)++ | 07:54 | |
| FROGGS | :o) | ||
| now I am back and pick up full fart :P | 07:55 | ||
| jnthn | :D | 07:59 | |
|
08:27
vendethiel joined
09:03
vendethiel joined
09:30
brrt joined
09:43
vendethiel joined
|
|||
| brrt | \o | 09:53 | |
| nwc10 | o/ | 09:54 | |
| FROGGS | -o- | 09:57 | |
| brrt | :-) | 10:12 | |
| so, what's today's plan for making perl6 faster | 10:13 | ||
| and more stable :-) | 10:24 | ||
| FROGGS | my plan is not to make it faster or stable, but I'm trying to make a certain slang this evening, or at least explore its limits :o) | 10:58 | |
|
10:59
vendethiel joined
11:12
Ven joined
|
|||
| brrt | sounds like a decent plan to me | 11:29 | |
| FROGGS | yeah, I am eager to start, since it is about EVALing a 'safe' subset of Perl 6, that is, a subset that is perhaps based on the restricted setting | 12:00 | |
| so that you can "EVAL" stuff passed by commandline, which may contain regexes and other stuff, but does not wipe your hard disk in a BEGIN block etc | 12:01 | ||
|
12:02
ggoebel joined
12:22
vendethiel joined
12:38
Ven joined
12:50
vendethiel joined
13:15
vendethiel joined
13:16
dalek joined
|
|||
| brrt | right. that would be bad | 13:48 | |
| nobody likes wiped hard disks | |||
|
13:51
Ven joined
14:07
vendethiel joined
|
|||
| dalek | arVM: 5ddab9a | FROGGS++ | src/core/nativecall.c: add dlerror() in errmsg for nativecall for clarity |
14:22 | |
|
14:41
vendethiel joined
14:50
colomon joined
15:14
vendethiel joined
|
|||
| japhb | JimmyZ: From msdn.microsoft.com/en-us/library/d...s.85).aspx (linked from the page you sent), it looks like the Windows "Unicode" functions push UTF-16 pretty hard, though apparently there is some support for translating UTF-7 and UTF-8 via pseudo-codepages. | 15:36 | |
.oO( UTF-16: The World^W^WUCS-2 is Not Enough ) |
15:37 | ||
|
15:43
colomon joined
|
|||
| JimmyZ | japhb: yeah. | 15:45 | |
|
16:39
camelia joined
17:38
vendethiel joined
17:40
FROGGS joined
18:16
Peter_R joined
|
|||
| timotimo wonders about jnthn's status | 19:36 | ||
|
19:49
colomon joined
|
|||
| lizmat | I think jnthn is travelling a lot in anticipation of the coming weekend | 19:49 | |
|
19:51
brrt joined
|
|||
| masak | I can confirm this. | 19:51 | |
| nwc10 | and I just thought that he was fleeing Eurovision 2016 | 19:52 | |
| lizmat | 2016 ??? | ||
| ah, yes | |||
| La Suede :-) | |||
| timotimo | ah | 19:59 | |
| i thought jnthn was going to develop and push over the course of the day | 20:05 | ||
| hm, that also means jnthn isn't going to be able to tell me more about how to figure out if a given call target wants to handle junctions on its own | 20:07 | ||
|
20:48
colomon joined
21:17
colomon joined
21:32
brrt left
21:41
colomon joined
22:11
vendethiel joined
22:19
colomon joined
|
|||
| timotimo | brrt, it seems like param_sp being used is keeping us from jitting next, last, and redo | 22:27 | |
| why must that thing be so complicated :\ | |||
| though on the other hand, we'd want to avoid p6argvmarray anyway | |||
| because that'll generate a BOOTArray every time we call next/last/redo even if the number of elements is bound to be 0 | 22:28 | ||
| can we use captureposelems without forcing the allocation of some object before that? | 22:32 | ||
| timotimo just goes ahead and tries | 22:34 | ||
| i just want a cheap way to figure out whether our block with signature -> | { ... } was called with arguments or without | 22:35 | ||
| but i suppose it really wants to be a multi | |||