|
00:24
AlexDaniel left
|
|||
| Geth | rakudo/4b68e9471-but-build-rakudo-with-rakuast: f5caad1027 | (Will Coleda)++ | src/main.nqp Reverse meaning of RAKUDO_RAKUAST, introduce RAKUDO_NO_RAKUAST just so we can build a whateverable-built rakudo for Blin to use where it's easier to have an entirely different rakudo build that turns rakuast on than it is to get a rakudo into Blin sideways ... |
00:36 | |
| [Coke] | alexdaniel++ | 00:37 | |
| kicked off a new ast blin run | 00:51 | ||
| export OLD=4b68e9471 | 00:52 | ||
| export NEW=f5caad102 | |||
|
00:54
donaldh left,
MasterDuke joined
|
|||
| SmokeMachine | ugexe: I just saw your commit adding the ast optimisation on a different pass, thank you very much for that! | 01:04 | |
| [Coke] | (that one is included in the testing, fwiw) | 01:06 | |
| m: 6/139.say | 01:33 | ||
| camelia | WARNINGS for <tmp>: 139 Useless use of "/" in expression "6/139.say" in sink context (line 1) |
||
| [Coke] | m: (6/139).say | ||
| camelia | 0.043165 | ||
| [Coke] | m: (5/143).say | 01:35 | |
| camelia | 0.034965 | ||
| [Coke] | ugexe: ^^ running failure %age | ||
| much better | |||
| here's the failures so far: | 01:38 | ||
| github.com/coke/raku-ast-blin-results | |||
| m: (10/173).say | 01:48 | ||
| camelia | 0.057803 | ||
| [Coke] | seeing a lot of "Object does not exist in serialization context" | 01:49 | |
| ugexe | yeah, there are like 50 of those. my understanding is it will require a fix at the moarvm level | 01:50 | |
| timo | may have to ensure "setobjsc" is called on the objects in question? | 02:02 | |
| i mean nqp::setobjsc, plus also the corresponding one that does the other half of the assignment | |||
| setobjsc and scsetobj | 02:03 | ||
| [Coke] | m: say 18/238 | 02:16 | |
| camelia | 0.07563 | ||
| ugexe | that isn't the case with the ones i'm specifically referencing, but then again I'm talking strictly about a RAKUDO_RAKUAST=1 built core as well. for example i can reproduce with echo 'class Grammar::Solo { }' > /tmp/Solo.rakumod + RAKUDO_RAKUAST=1 raku -I /tmp -e 'use Solo; say "ok"' (again, with a RAKUDO_RAKUAST=1 built core) | 02:32 | |
|
02:32
MasterDuke left
|
|||
| ugexe | it actually blows up in the serializer when it precompiles Solo, but the object it can't find traces back to loading the core setting | 02:32 | |
| my understanding is the stable repossession pass runs before the object pass, so a stable body can demand an object whose slot the object pass hasn't repossessed yet | 02:33 | ||
| as for why it happens so much now with rakuast, rakuast routine path applies parameterized mixins where the legacy actions didn't. the clearest case being the Callable[T] vs Callable return type mixin on routines with a return constraint | 02:45 | ||
|
02:54
japhb joined
04:04
vrurg_ joined
04:07
vrurg left
|
|||
| Geth | rakudo: ugexe++ created pull request #6344: RakuAST: Lower operator and dot assignments in the optimize stage |
04:27 | |
|
05:39
zostay left
05:53
zostay joined
06:42
releasable6 left
07:46
librasteve_ joined
08:26
finanalyst left
08:34
hurufu joined
08:53
hurufu left
09:25
tellable6 left
09:29
Voldenet left
10:09
Voldenet joined
10:20
AlexDaniel joined
|
|||
| AlexDaniel | c: HEAD say 42 | 10:21 | |
| committable6 | AlexDaniel, ¦HEAD(4b68e94): «42» | ||
| AlexDaniel | uhhh | ||
|
11:06
finanalyst joined
|
|||
| AlexDaniel | e: say 42 | 11:13 | |
| evalable6 | 42 | ||
|
11:19
evalable6 left
11:21
unicodable6 left,
greppable6 left,
nativecallable6 left,
bloatable6 left,
quotable6 left,
sourceable6 left,
benchable6 left,
coverable6 left,
huggable6 left,
committable6 left
11:22
finanalyst left,
finanalyst_ joined,
notable6 left,
bisectable6 left,
shareable6 left,
linkable6 left
11:23
coverable6 joined,
bloatable6 joined
11:24
evalable6 joined,
nativecallable6 joined,
linkable6 joined,
unicodable6 joined,
notable6 joined,
releasable6 joined,
benchable6 joined
11:25
huggable6 joined,
shareable6 joined,
bisectable6 joined,
tellable6 joined,
committable6 joined,
greppable6 joined
11:26
quotable6 joined,
sourceable6 joined
|
|||
| [Coke] | AlexDaniel: thanks for getting committable working with latest commits again | 12:48 | |
| AlexDaniel | yeah, of course. There was some other issue with the server shortly after but I'm not sure what it was | 12:49 | |
| couldn't even ssh into it while the bots were active and replying | |||
| restart helped | |||
| [Coke] | looks like the blin disk filled up, also. | 12:50 | |
| I saved whatever was there, but I'm guessing some of the later failures might be due to out of space errors. | 12:57 | ||
|
13:10
AlexDaniel left
|
|||
| [Coke] | Apparently I need to consider wiping out cached zef data more often. (blin specifically doing the caching for zef here, not zef itself) | 13:18 | |
| (I think) | 13:22 | ||
| ugexe: let me know if that's enough for you to get some more todos from, if not, I'll rerun after cleaning the disk | 13:33 | ||
| ugexe | should be good | 13:39 | |
| [Coke] | excellent. cache wiped, shutting down vm, let me know when you want to go again. :) | 13:44 | |
| librasteve_ | rakudoweekly.blog/2026/06/29/2026-...lease-194/ | 13:48 | |
| coleman | I highly recommend this cli tool for checking disk space in various folders on a Linux box github.com/byron/dua-cli#installation | 16:31 | |
| easier than remembering the raw POSIX commands for me | 16:32 | ||
|
17:09
finanalyst_ left
|
|||
| [Coke] installs dua on a few boxes. | 17:30 | ||
| ugexe | m: sub f(+@a, Int :$x) { say @a.raku; say $x }; f(:x("s")) | 19:39 | |
| camelia | [] s |
||
| ugexe | rakuast actually does the type check which is the right thing to do imo, but i *think* that is the cause of the MergeOrderedSeqs and List::Allmax rakuast failures in github.com/coke/raku-ast-blin-results | 19:42 | |
| m: run $*EXECUTABLE, q[-e], q[sub f(+@a, Int :$x) { say @a.raku; say $x }; f(:x("s"))]; BEGIN %*ENV<RAKUDO_RAKUAST>=1 | 19:44 | ||
| camelia | Type check failed in binding to parameter '$x'; expected Int but got Str ("s") in sub f at -e line 1 in block <unit> at -e line 1 The spawned command '/home/camelia/rakudo-m-inst-2/bin/perl6-m' exited unsuccessfully (exit code: 1, signal:… |
||
|
21:42
guifa_ joined
22:38
apogee_ntv left,
apogee_ntv joined
|
|||
| Geth | rakudo: ugexe++ created pull request #6345: RakuAST: fix hyper zen slice `>>[]` / `>>{}` / `>><>` selecting nothing |
23:30 | |
| rakudo: ugexe++ created pull request #6346: RakuAST: apply the sigil contextualizer to `@<foo>` / `%<foo>` captures |
23:31 | ||
| rakudo: ugexe++ created pull request #6347: Make an unrecognized `<| >` regex boundary a no-op before 6.e and an error after |
23:36 | ||
| rakudo/main: 44b9684a5f | (Nick Logan)++ (committed using GitHub Web editor) | 2 files RakuAST: Collect $=rakudoc from inside package bodies (#6341) $=rakudoc was filled by walking the compunit statement list, descending only into bare blocks. Docs declared inside a package body, as in a `unit class`, were never reached, so $=rakudoc came out empty and --rakudoc printed nothing for such files. Descend into a package or routine/block body as well, so its declarator docs and doc blocks are collected too. Fixes #6183 |
23:43 | ||
| rakudo/main: 90550fc4eb | (Nick Logan)++ (committed using GitHub Web editor) | 2 files Render a parameter's declarator doc once under --doc (#6342) A `#=` doc on a routine parameter was shown twice: once inline in the rendered signature, and again as a standalone block. The standalone form came from declarator2text treating the parameter like a class, since a Parameter object's HOW is a ClassHOW, so it fell into the `class ...` branch. Skip parameters in declarator2text; their doc already appears inline in the routine signature. Fixes #6182 |
|||