| Geth | rakudo/main: 266e714f30 | (Nick Logan)++ | t/12-rakuast/xx-fixed-in-rakuast.rakutest Keep operator wraps from leaking outside their test The wrap tests for infix:<|> and postfix:<++> wrapped a setting operator with a handler that does not call the original, asserted with is-deeply, and only then removed the wrap through a separate EVAL. Everything that ran in between, including the compilation of the unwrap EVAL itself, executed with the operator broken for the ... (13 more lines) |
00:16 | |
| rakudo/main: d2f6b7947d | (Nick Logan)++ (committed using GitHub Web editor) | t/12-rakuast/xx-fixed-in-rakuast.rakutest Merge pull request #6260 from ugexe/ugexe/wrap-test-narrow-window Keep operator wraps from leaking outside their test |
|||
| [Coke] | have managed to kill blin 3 times now before it actually started testing anything | 00:37 | |
| Geth | setup-raku/dependabot/npm_and_yarn/esbuild-0.28.1: 0e7fa42646 | dependabot[bot]++ (committed using GitHub Web editor) | 2 files Bump esbuild from 0.28.0 to 0.28.1 Bumps [esbuild](github.com/evanw/esbuild) from 0.28.0 to 0.28.1. - [Release notes](github.com/evanw/esbuild/releases) - [Changelog](github.com/evanw/esbuild/blob/main...NGELOG.md) - [Commits](github.com/evanw/esbuild/compare/v.....v0.28.1) ... (8 more lines) |
01:03 | |
| setup-raku: dependabot[bot]++ created pull request #56: Bump esbuild from 0.28.0 to 0.28.1 |
|||
| rakudo: ugexe++ created pull request #6262: RakuAST: capture bracketed groups as a whole |
|||
| rakudo: ugexe++ created pull request #6263: RakuAST: accept lexical regexes in the grammar sanity check |
01:24 | ||
| [Coke] | it's been hanging try to pull build 2f9a672e30e807a2e6dff51b8a25cca76725aafe | 01:35 | |
| linkable6 | (2026-05-23) github.com/rakudo/rakudo/commit/2f9a672e30 [release] Bump VERSION to 2026.05 | ||
| [Coke] | c: 2f9a672e30e807a2e6dff51b8a25cca76725aafe 3.say | ||
| committable6 | [Coke], ¦2f9a672: «3» | ||
| [Coke] | m: d2f6b7947 2.say | 01:36 | |
| camelia | ===SORRY!=== Error while compiling <tmp> Undeclared routine: d2f6b7947 used at line 1 |
||
| linkable6 | (2026-06-12) github.com/rakudo/rakudo/commit/d2f6b7947d Merge pull request #6260 from ugexe/ugexe/wrap-test-narrow-window | ||
| [Coke] | c: d2f6b7947 2.say | ||
| committable6 | [Coke], ¦d2f6b79: «2» | ||
| [Coke] | yah, keeps hanging on Attempting to fetch 2f9a672e30e807a2e6dff51b8a25cca76725aafe… | 02:36 | |
| I foolishly did an apt-get update today | 02:39 | ||
|
03:27
timo left
|
|||
| ugexe | well the us government just banned the model i've been using the last week :/ | 03:45 | |
| Geth | rakudo: ugexe++ created pull request #6264: RakuAST: release compiler services after composition |
03:50 | |
|
06:30
librasteve_ joined
|
|||
| Geth | rakudo/main: 8c2682d231 | (Nick Logan)++ (committed using GitHub Web editor) | 4 files RakuAST: release compiler services after composition The compiler services handed to the metamodel on compose end up in the slurpy named arguments of any compose method along the way, and those live in the method's frame. A custom metaclass that creates a closure during compose and installs it as a method, the way OO::Monitors wraps every method with lock handling, puts that frame ... (11 more lines) |
07:04 | |
| setup-raku: f6d278afea | dependabot[bot]++ (committed using GitHub Web editor) | 2 files Bump esbuild from 0.28.0 to 0.28.1 (#56) Bumps [esbuild](github.com/evanw/esbuild) from 0.28.0 to 0.28.1. - [Release notes](github.com/evanw/esbuild/releases) - [Changelog](github.com/evanw/esbuild/blob/main...NGELOG.md) - [Commits](github.com/evanw/esbuild/compare/v.....v0.28.1) ... (9 more lines) |
08:07 | ||
|
12:26
timo joined
|
|||
| [Coke] | bah. takes *minutes* to get to "attempting to fetch ..." - put in a dd statement there to capture the link it's retrieving... no output. | 13:19 | |
| Do we have any way to make sure the "mothership" is responing? | 13:20 | ||
| Geth | Blin/coke/cleanup: 8b9e9e556d | (Will Coleda)++ | 2 files Fix Essential handling |
13:35 | |
| Blin/coke/cleanup: ee89a67236 | (Will Coleda)++ | 2 files also track fail.json |
|||
| Blin: 1aa0628f3e | (Will Coleda)++ | lib/Blin/Tester/Zef.rakumod This is the number of hours after which to force an update 0 == no hours, don't force. |
13:50 | ||
| ugexe | What is the mothership? | 17:38 | |
| Geth | rakudo: ugexe++ created pull request #6265: RakuAST: interpret pairs, hash composers, and contextualizers |
18:17 | |
| rakudo: ugexe++ created pull request #6266: RakuAST: don't adopt a setting builtin's WHO when a package shadows it |
18:18 | ||
| ugexe | I can get most of the dependency chain of Cro::HTTP installed under RakuAST, but Cro::HTTP itself has bigger issues | 18:19 | |
| notably spesh issues that are exposed with rakuast because it doesn't do lexical to local lowering | 18:20 | ||
| we could add lexical to local lowering to rakuast and potentially get around it, but it still feels bad that code only works because of optimizations | 18:21 | ||
| Notably moar needs to reset per invocation cloned lexicals ahead of each entry into the inlined body | 18:23 | ||
| ab5tract | Yeah ideally we would fix that. In fact, theoretically we might find some interesting things from a blin run that disables opt and s | 18:25 | |
| *spesh | |||
| ugexe | i've hit a couple | 18:26 | |
| ab5tract | This is another one that needs some moarvm changes, right? | ||
| ugexe | yeah | 18:27 | |
|
18:57
Pixi left
19:04
finanalyst joined
19:31
Pixi joined
|
|||
| timo | i'm a little confused by that | 20:18 | |
| the optimizer wouldn't lower lexicals to locals if they are referenced in an inner block | |||
| what does "per invocation cloned lexicals" mean exactly? | 20:19 | ||
| do blocks usually initialize their lexicals upon entry with a `setlex` instruction? does rakuast maybe skip those in some cases? | 20:31 | ||
| [Coke] | mothership is the thing that serves out the pre-compiled rakus. | 20:40 | |
| [6~[6~ | |||
| oops | |||
| timo | moths are cool | 20:42 | |
| ugexe | m: raku -e 'my $b="/tmp/b.moarvm";run($*EXECUTABLE,"--target=mbc","--output=$b","-e",q[my @a;{@a}],:!err);my @l=run("$*EXECUTABLE/../moar","--dump",$b,:out).out.slurp.lines;say "setlex=",+@l.grep(/<<setlex>>/)," getlex=",+@l.grep(/<<getlex>>/)' | 21:34 | |
| camelia | ===SORRY!=== Error while compiling <tmp> Two terms in a row at <tmp>:1 ------> raku -e<HERE> 'my $b="/tmp/b.moarvm";run($*EXECUTABLE expecting any of: infix infix stopper postfix statement end … |
||
| ugexe | m: my $b="/tmp/b.moarvm";run($*EXECUTABLE,"--target=mbc","--output=$b","-e",q[my @a;{@a}],:!err);my @l=run("$*EXECUTABLE/../moar","--dump",$b,:out).out.slurp.lines;say "setlex=",+@l.grep(/<<setlex>>/)," getlex=",+@l.grep(/<<getlex>>/) | 21:35 | |
| camelia | setlex=0 getlex=0 | ||
| ugexe | hmm im not sure that is accurate. locally its setlex=0 getlex=1 | ||
| with the point being i don't think rakuast is skipping the setlex if that legacy example isn't doing it | 21:36 | ||
|
21:52
finanalyst left
|
|||
| ugexe | on legacy the non captred my @ gets lowered to a local so theres no contvar in the frame for the inliner to mishandle. rakuast doesnt lowe r it so the contvar stays put in an inlineable frame and the inliner reuses one clone across calls | 22:03 | |
| timo | do you have a speshlog or two for me? | 22:13 | |
| lizmat | m: sub MAIN; # TIL that "unit" is optional here | 22:22 | |
| camelia | ( no output ) | ||
| lizmat | sleep& | ||
| ugexe | gist.github.com/ugexe/8bf9bb2f0742...7dd7c60bd0 I can put a MVM_SPESH_LOG up somewhere if that would be useful, but its 600k lines | 23:54 | |
| (that link has MVM_SPESH_INLINE_LOG=1 output) | |||