00:18 |Tux| left 00:29 |Tux| joined 00:32 bartolin left, bartolin joined 00:41 sena_kun left 00:56 sena_kun joined 01:21 Voldenet left 01:26 Voldenet joined, Voldenet left, Voldenet joined
Geth_ roast: vrurg++ created pull request #608:
Test for export of long names
02:42 sena_kun left 02:56 sena_kun joined 03:56 quotable6 left, shareable6 left, benchable6 left, nativecallable6 left, committable6 left, sourceable6 left, bloatable6 left, unicodable6 left, statisfiable6 left, releasable6 left, reportable6 left, greppable6 left, notable6 left, bisectable6 left, coverable6 left, squashable6 left 03:57 unicodable6 joined, shareable6 joined, bisectable6 joined, nativecallable6 joined, statisfiable6 joined, notable6 joined, reportable6 joined 03:58 sourceable6 joined, committable6 joined, coverable6 joined, benchable6 joined, quotable6 joined 03:59 greppable6 joined, squashable6 joined, releasable6 joined, bloatable6 joined 04:41 sena_kun left 04:56 sena_kun joined 05:35 Voldenet left 05:41 Voldenet joined, Voldenet left, Voldenet joined 06:41 sena_kun left 06:55 sena_kun joined 07:21 Xliff left 08:00 squashable6 left 08:03 squashable6 joined 08:08 Voldenet left 08:14 Voldenet joined, Voldenet left, Voldenet joined 08:18 Voldenet left 08:19 squashable6 left 08:20 squashable6 joined 08:23 Voldenet joined, Voldenet left, Voldenet joined
lizmat Files=1294, Tests=109635, 209 wallclock secs (28.23 usr 8.32 sys + 2952.08 cusr 269.24 csys = 3257.87 CPU) 08:38
08:42 sena_kun left 08:57 sena_kun joined
Geth_ rakudo: a3afde6037 | (Elizabeth Mattijsen)++ | 6 files
Re-imagine val() handling

This replaces the old val() handling by a completely re-imagined version that caters to the way that Raku is being optimized. The goal was to reach an overall speed increase of 2x, but this was not achieved. But this commit makes val handling more correct, and about 2x faster in the most common case, namely that *no* number is encountered. Which helps at every startup, as each %*ENV ... (39 more lines)
lizmat so, that's out of my system 09:20
Geth_ nqp: d69855343c | (Stefan Seifert)++ | src/vm/moar/QAST/QASTCompilerMAST.nqp
Fix "Decoder may not be used concurrently" errors with concurrent EVAL

String decoders are stateful objects, so sharing one between concurrently running compilers has never been a good idea. Give every MAST compiler its own latin1decoder to avoid concurrency issues.
09:44 AlexDaniel left, AlexDaniel joined, AlexDaniel left, AlexDaniel joined 10:09 titsuki joined 10:43 sena_kun left 10:55 sena_kun joined 11:02 squashable6 left 11:04 squashable6 joined 11:33 Voldenet left 11:38 Voldenet joined, Voldenet left, Voldenet joined
Geth_ roast: fd2c4ad9c5 | usev6++ | S17-procasync/stress.t
[JVM] Skip hanging test
12:41 sena_kun left 12:57 sena_kun joined
lizmat what would be a good way to benchmark Supply.lines ? 13:18
14:40 sena_kun left 14:57 sena_kun joined 15:05 Xliff joined 15:47 rba[m] left 15:50 lucasb joined
moritz lizmat: you could cycle through an array of strings, some containing newlines and some not, and feed them to the supplier 16:34
16:41 sena_kun left 16:56 sena_kun joined
lizmat moritz: I came up with my int $a; ("words".IO.slurp x 5).Supply.lines.tap: { ++$a }; dd $a 17:21
Geth_ rakudo: aea2a36ce0 | (Elizabeth Mattijsen)++ | src/core.c/control.pm6
Make emit about 30% faster

From what I can gather from a --profile. Before, an "emit/THROW" combo would take 588 msecs in a benchmark, and after the "emit" by itself takes 442 msecs. Since this is pretty hot code in Supply handling, thought that every little bit counts.
18:26 patrickb joined, AlexDani` joined 18:29 AlexDaniel left 18:42 sena_kun left 18:56 sena_kun joined 19:44 stoned75 left
Geth_ rakudo: a91813d017 | (Elizabeth Mattijsen)++ | src/core.c/control.pm6
Make `take foo` about 20% faster

By not using the helper THROW sub, but instead do all of the stuff inside `take`.
20:41 sena_kun left
Geth_ rakudo: 533a2ad9e9 | (Elizabeth Mattijsen)++ | src/core.c/control.pm6
Make `take-rw foo` about 20% faster

By not using the helper THROW sub, but instead do all of the stuff inside `take-rw`. Also fix containerization issue that I noticed in
  `take` while working on `take-rw`.
20:55 Xliff left 20:56 sena_kun joined
Geth_ rakudo: e1aaa83cc1 | (Elizabeth Mattijsen)++ | src/core.c/control.pm6
Make `take foo, bar` about 60% faster

By not using the helper THROW or RETURN-LIST subs, but instead do all of the stuff inside `take`. Which could be made a lot simpler, as that candidate can only be called if there is more than one parameter, so the whole check for number of elements was not necessary.
vrurg lizmat: t/spectest.data.6.c is not used anymore. All tests are in t/spec/spectest.data 21:10
I think it is time to remove the file as confusing. It was left there for transition period as some tooling might have depend on it. 21:11
vrurg has added the test into the right place. 21:13
Geth_ rakudo: bed716bed2 | (Elizabeth Mattijsen)++ | src/core.c/control.pm6
Make `take-rw foo, bar` about 60% faster

Same way as `take foo, bar`.
21:22 squashable6 left 21:24 squashable6 joined 21:33 |Tux| left
Geth_ rakudo: 63f117502a | (Elizabeth Mattijsen)++ | src/core.c/control.pm6
Make `succeed foo` 30% and `succeed foo,bar` 60% faster

A rince a repeat of the `take` optimizations
21:54 Xliff joined 22:12 lucasb left 22:27 |Tux| joined
dogbert11 not ok 146 - parsed literal == val() [.08548480145036831822e-304] 22:41
# Failed test 'parsed literal == val() [.08548480145036831822e-304]'
# at t/spec/S32-num/stress.t line 43
22:41 sena_kun left 22:57 sena_kun joined
Geth_ rakudo: ee1b9918ce | (Patrick Böker)++ | 5 files
Add MacOS binary release files and Perl 6 -> Raku renames
23:24 patrickb left
Geth_ rakudo: 5162e6f5a3 | (Elizabeth Mattijsen)++ | 7 files
Remove now useless THROW helper sub