Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2017-05-12

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
01:49 ilbot3 joined #moarvm
01:49 Topic for #moarvm is now https://github.com/moarvm/moarvm | IRC logs at  http://irclog.perlgeek.de/moarvm/today
01:51 geekosaur joined #moarvm
01:54 colomon_ joined #moarvm
02:18 samcv pasted this in perl6-dev but
02:19 samcv looks like moarvm doesn't have a point release. and we have 2017.04.3 but no newer tagged version of moarvm
02:19 samcv or nqp. since the fixed in .3 needed updates to moarvm
02:24 samcv do we want to make a point release so distros can update moarvm/nqp and thus get 2017.04.3? seems it would make sense
02:31 MasterDuke_ 2017.05 is only about a week away...
02:57 samcv ok
02:58 samcv so we just wait until 2017.05? i mean still it's weird for a tagged release to not be able to be installed based on other tags
03:01 samcv i mean it's easy enough to add tags for the nqp and moarvm revisions on which it depends on
04:39 colomon joined #moarvm
05:31 brrt joined #moarvm
05:45 domidumont joined #moarvm
05:50 domidumont joined #moarvm
05:54 domidumont joined #moarvm
06:03 brrt ugh, i found a new bug
06:10 samcv such is life
06:28 geekosaur joined #moarvm
06:42 brrt joined #moarvm
07:04 lizmat joined #moarvm
07:15 brrt it's always a bit dissapointing, though, when the bug turns out to be so… banal
07:29 brrt hehe, bug!
07:49 brrt i do wonder why the jit bisection script just won't work on OS X
07:56 samcv what's it written in?
08:01 Geth ¦ MoarVM/even-moar-jit: b1c982f726 | (Bart Wiegmans)++ | 4 files
08:01 Geth ¦ MoarVM/even-moar-jit: Fig call arg resolution bug
08:01 Geth ¦ MoarVM/even-moar-jit:
08:01 Geth ¦ MoarVM/even-moar-jit: CALL and CALLV nodes can use register arguments and those can conflict
08:01 Geth ¦ MoarVM/even-moar-jit: with ARGLIST parameters. So, they have to be moved out of the
08:01 Geth ¦ MoarVM/even-moar-jit: way. However, I mistakenly used the (constant value) args rather than
08:01 Geth ¦ MoarVM/even-moar-jit: the register numbers, so this didn't quite work.
08:01 Geth ¦ MoarVM/even-moar-jit: review: https://github.com/MoarVM/MoarVM/commit/b1c982f726
08:01 Geth ¦ MoarVM/even-moar-jit: 2d82ae7058 | (Bart Wiegmans)++ | 4 files
08:01 Geth ¦ MoarVM/even-moar-jit: Add decont template (uses repr function invocation)
08:01 Geth ¦ MoarVM/even-moar-jit:
08:01 Geth ¦ MoarVM/even-moar-jit: This was kind of tricky because it involves a destructive template and
08:01 Geth ¦ MoarVM/even-moar-jit: a pointer function call.
08:01 Geth ¦ MoarVM/even-moar-jit: review: https://github.com/MoarVM/MoarVM/commit/2d82ae7058
08:07 brrt perl
08:07 brrt usual old perl
08:07 brrt but it seems like stuff just goes wrong somehow
08:08 zakharyas joined #moarvm
08:23 lizmat morning!
08:23 brrt morning lizmat
08:23 brrt how's the #pts
08:24 lizmat so I want to add the information contained in nqp/gen/moar/stage1/nqp-config.nqp as part of the perl6 -V output
08:24 brrt uhuh
08:24 lizmat busy  :-)
08:24 lizmat wish I felt better  :-(
08:24 * brrt wishes you better
08:25 lizmat brrt: thanks  :-)_
08:25 lizmat probably should ask this on perl6-dev
08:34 Ven joined #moarvm
08:37 AlexDaniel joined #moarvm
09:25 robertle joined #moarvm
09:49 brrt this has to be the biggest miscompile i've had so far
09:51 brrt the broken frame nr is 1380\
09:55 robertle joined #moarvm
10:07 Ven joined #moarvm
10:20 samcv i'm gonna add info of the tags to the 2017.04.3 release page. since adding tags for a nonrelease seems no bueno. that will at least tie up that end enough for my satisfaction
10:45 Ven_ joined #moarvm
12:34 Ven joined #moarvm
12:53 Ven_ joined #moarvm
13:05 brrt joined #moarvm
13:08 Ven_ joined #moarvm
13:26 Ven_ joined #moarvm
13:53 brrt my problem looks suspiciously like a missing deopt check, or missing invokish, or something like that
13:54 Ven joined #moarvm
14:09 Ven_ joined #moarvm
14:19 Ven__ joined #moarvm
15:11 lizmat_ joined #moarvm
16:05 Ven joined #moarvm
16:42 domidumont joined #moarvm
16:44 MasterDuke_ joined #moarvm
17:11 Ven joined #moarvm
17:23 Ven_ joined #moarvm
17:30 avar joined #moarvm
17:39 Ven_ joined #moarvm
17:53 Ven_ joined #moarvm
17:56 zakharyas joined #moarvm
18:11 Ven joined #moarvm
18:21 Ven_ joined #moarvm
18:24 AlexDaniel joined #moarvm
18:28 Ven_ joined #moarvm
18:36 Ven_ joined #moarvm
19:09 Ven joined #moarvm
19:14 praisethemoon joined #moarvm
19:14 praisethemoon joined #moarvm
19:35 Ven_ joined #moarvm
19:53 Ven_ joined #moarvm
20:27 Zoffix m: use Test; like 'foo', none /bar/
20:27 camelia rakudo-moar dc5eec: OUTPUT: «not ok 1 - text matches '/bar/'␤␤# Failed test 'text matches '/bar/''␤# at SETTING::src/core/Junction.pm line 249␤#      expected: '/bar/'␤#      got: 'foo'␤»
20:39 Ven joined #moarvm
21:19 Ven joined #moarvm
21:35 samcv Zoffix, added info about the dev/escape.p6 and dev/unescape.p6 utilities to CONTRIBUTING.md for atom-language-perl6
21:35 samcv if you use those you will never have escaping problems again :)
21:35 samcv much easier to edit things when they are in *normal* form and you only have to deal with the regex escapes
21:59 lizmat joined #moarvm
22:23 timotimo i wonder if we can do better when failing to allocate more memory because we're infinitely recursing
22:24 timotimo like, figure out if we're mallocing because we're trying to have space for stack frames
22:32 timotimo it's problematic because we also allocate other stuff while infini-recursing, so it won't always be the callstack region being expanded that causes the problem
22:33 jnthn And if the frame always escapes the stack by being heap-promoted that approach won't work anyway
22:35 jnthn Not to mention that another therad may hit the out of memory condition than the one that infinitely recurses
22:35 jnthn *thread
22:38 timotimo right
22:38 timotimo we could change the out of memory panic function to inspect the TC and instance and such
22:38 timotimo to see what's currently most expensive
22:42 jnthn Maybe but it doesn't get passed tc or instance
22:43 timotimo right. we can add a function that does analysis before, and pass tc whenever we allocate in a way where it's convenient for us to pass stuff

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary