| Geth | rakudo/nom: 79553d0fc3 | (Zoffix Znet)++ | src/core/Rational.pm Finalize fix for Rational.ceiling/.floor The original fix[^1] is good; the <4/2> Rat gets created during addition of two Rats and it's not reduced there for performance reasons. So keep the reduction in .ceiling and toss it from .floor as it's not needed there. [1] github.com/rakudo/rakudo/commit/63...1a58490101 |
01:08 | |
| roast: 50e33352e9 | (Zoffix Znet)++ | 2 files Test .ceiling works for whole Rationals without 1 in denominator Rakudo fix: github.com/rakudo/rakudo/commit/79553d0fc3 |
01:10 | ||
| travis-ci | Rakudo build passed. Zoffix Znet 'Finalize fix for Rational.ceiling/.floor | 03:20 | |
| travis-ci.org/rakudo/rakudo/builds/200549770 github.com/rakudo/rakudo/compare/6...553d0fc3a0 | |||
| IOninja | so quiet... | 14:32 | |
| dogbert17 | hmm, there's still something wrong when spectesting under harness6 | 15:01 | |
| timotimo, jnthn: see anything suspicious here? gist.github.com/dogbert17/ffdabdb6...06d596434f | 15:05 | ||
| IOninja | ZOFVM: Files=1224, Tests=132886, 131 wallclock secs (22.17 usr 3.28 sys + 2638.71 cusr 267.61 csys = 2931.77 CPU) | 15:15 | |
| Geth | rakudo/nom: 266f345718 | (Zoffix Znet)++ | t/spectest.data Mark some slow tests as... slow So they get spread out through the run |
15:16 | |
| jnthn | dogbert17: Hmm, that's an interesting one... | 15:17 | |
| dogbert17 | annoyingly I don't seem to be able to do an MVM_dump_backtrace since gdb is acting up.. | 15:20 | |
| hopefully there's a least a sliver of clue there :) | 15:21 | ||
| jnthn | It's odd because it blows up freeing some memory, but that memory apparently contaiend a valid pointer which we freed successfully | ||
| Or the line number is off-by-one | |||
| MVM_free(cur_chars->chars); | |||
| MVM_free(cur_chars); | |||
| 515 is the second of those two | 15:22 | ||
| dogbert17 | don't forget that I'm running the 'fixed' orchestrate.c | ||
| jnthn | Yeah, but don't think that'd cause anything to be off in decodestream.c. :) | 15:23 | |
| dogbert17 | doh | ||
| :) | |||
| for me line 515 is MVM_free(cur_chars); | |||
| jnthn | Yeah, we too :) | 15:24 | |
| So hmm | |||
| dogbert17 | problem is that I can run the spectest several times before this happens | 15:25 | |
| jnthn | I guess no luck with ASAN? | ||
| dogbert17 | haven't tried that on this but I could give it a go | ||
| if ASAN triggers while running a spectest will I see the output or does the test harness hide it somehow? | 15:26 | ||
| ok ASAN it is then...to be continued | 15:27 | ||
| jnthn | If it's spat out by a test itself, it'd be eaten I guess, but if it's the harness itself that blows up it'll halt | 15:28 | |
| And I think it's the harness that's the trouble, so should do the right thing | |||
| dogbert17 | this might take a while ... | 15:29 | |
| also, moarvm was probably compiled without --no-optimize :( | 15:31 | ||
| jnthn | At least ASAN is only a little slowdown :) | 15:33 | |
| Compared to valgrind ;) | |||
| dogbert17 | yes, valgrinding this would be a nightmare | 15:36 | |
| I wonder if this intermittent problem is the one that lizmat has been complaining about | 15:37 | ||
| jnthn | Quite possibly | ||
| I had bad frees but in P6bigint | 15:38 | ||
| timotimo | hm, running both the harness and the tests with asan ... | 15:39 | |
| dogbert17 | it's at least running :) | ||
| jnthn | :) | 15:40 | |
| dogbert17 | hmm, I have a vague menory that there's at least one test file which hangs when ASAN i enabled, wish I could remember which one it is | 15:42 | |
| (brilliant) spelling, and here I thought that my cataract operation would make things better, grr | 15:43 | ||
| timotimo | so, we had the problem where maxrss was growing way out of proportion, but that was fixed with your PR, right? | 15:47 | |
| dogbert17 | yes I hope so :) | 15:48 | |
| timotimo | OK | ||
| if not, it might have been interesting to try getting vmhealth reports regularly | |||
| but not important any more :) | |||
| dogbert17 | maybe we should trick IOninja into trying this, doesn't he have access to a monster box with hundreds of cores | 15:49 | |
| [Tux] | This is Rakudo version 2017.01-204-g266f34571 built on MoarVM version 2017.01-49-g7ff91dbb | ||
| csv-ip5xs 2.856 | 15:50 | ||
| test 12.359 | |||
| test-t 5.033 - 5.048 | |||
| csv-parser 13.097 | |||
| timotimo | i mean regularly within one run | 15:51 | |
| dogbert17 | the problem was actually an overflow in a percentage calculation | 15:52 | |
| timotimo | yup | 15:55 | |
| dogbert17 | so there are at least two bugs left wrt harness6, the one ASAN is trying to find and the file descriptor one when running with TEST_JOBS=1 | 15:58 | |
| IOninja | buggable: speed | 16:04 | |
| buggable | IOninja, ▃▂▁▂▃▂▂▃▃▂▅▂▂▂▁▂▂▂▁▂▂▂▄▂▃▃▂▂▂▂▂▂▂▃▂▃▃█▂▃▂▂▁▁▂▁▂▁▂▂ data for 2017-01-11–2017-02-11; range: 4.808s–6.505s | ||
| IOninja | I was hopping removing the IO::Handle.lines close check for each pull-one would improve speed :( | 16:05 | |
| hoping too | |||
| s: $*PROGRAM.IO, 'lines', \() | 16:07 | ||
| SourceBaby | IOninja, Sauce is at github.com/rakudo/rakudo/blob/266f...th.pm#L511 | ||
| IOninja | s: $*PROGRAM.IO.open, 'lines', \() | ||
| SourceBaby | IOninja, Sauce is at github.com/rakudo/rakudo/blob/266f...ol.pm#L190 | ||
| IOninja | wa | ||
| s: $*PROGRAM.IO.open, 'lines', \(:close) | 16:08 | ||
| SourceBaby | IOninja, Sauce is at github.com/rakudo/rakudo/blob/266f...ol.pm#L190 | ||
| IOninja | s: $*PROGRAM.IO.open, 'Str', \() | ||
| SourceBaby | IOninja, Sauce is at github.com/rakudo/rakudo/blob/266f...ure.pm#L73 | ||
| IOninja | ah | 16:11 | |
| s: say $*PROGRAM.IO.open | 16:12 | ||
| SourceBaby | IOninja, Something's wrong: ERR: Failed to open file /home/zoffix/services/sourceable/perl6/-e: no such file or directory in block <unit> at -e line 6Actually thrown at: in block <unit> at -e line 6 | ||
| timotimo | hehe. | ||
| [Coke] wonders if we need a --pod-only that skips any phasers that would normally run code. | 19:35 | ||
| IOninja wonders if that's still unsafe, 'cause you can stuff code into table's params IIRC | 19:40 | ||
| though you may need a BEGIN block to run 'em... I think | |||
| timotimo | i think we only run DOC when pod is turned on | 19:41 | |
| potentially BEGIN | |||
| samcv | m: #| here is a thingmy $var; say $var.WHY | 20:29 | |
| camelia | rakudo-moar 266f34: OUTPUT«(Any)» | ||
| samcv | m: #| here is a thingmy $var; sub thingy { ... }; say &thingy.WHY | ||
| camelia | rakudo-moar 266f34: OUTPUT«here is a thing» | ||
| samcv | why does the first one not work | ||
| moritz | dunno if .WHY is implemented for variiables | 20:30 | |
| japhb | m: my $var; #= This is the varsay $var.WHY; say $var.VAR.WHY; | 21:11 | |
| camelia | rakudo-moar 266f34: OUTPUT«(Any)(Any)» | ||
| japhb | Dang | ||
| samcv | moritz, it's in synopsis at least | 21:36 | |
| timotimo | isn't it just the other way around? | 22:48 | |
| m: #= here is a thingmy $var; say $var.WHY | |||
| camelia | rakudo-moar 266f34: OUTPUT«(Any)» | ||
| timotimo | m: #= here is a thingmy $var; say $var.VAR.WHY | ||
| camelia | rakudo-moar 266f34: OUTPUT«(Any)» | ||
| timotimo | m: #| here is a thingmy $var; say $var.VAR.WHY | ||
| camelia | rakudo-moar 266f34: OUTPUT«(Any)» | ||
| timotimo | *shrug* | ||
| samcv | timotimo, both are allowed i think | ||
| m: #= this thing is great!sub great-thing { }; say &great-thing.WHY | 22:49 | ||
| camelia | rakudo-moar 266f34: OUTPUT«Nil» | ||
| samcv | m: #= this thing is great!sub great-thing { }; say &great-thing.WHAT | ||
| camelia | rakudo-moar 266f34: OUTPUT«(Sub)» | ||
| samcv | hmm | ||
| m: my $var; #= here is a thing say $var.WHY | 22:50 | ||
| camelia | rakudo-moar 266f34: OUTPUT«(Any)» | ||
| samcv | m: my $var; $var.WHY = 'this is a thing' | ||
| camelia | rakudo-moar 266f34: OUTPUT«Cannot modify an immutable Any in block <unit> at <tmp> line 1» | ||
| timotimo | both are allowed, but one means "annotate the thing before the comment" and the other "annotate the thing after the comment" | 22:52 | |
| samcv | yeah that's it |