| ugexe | i.e. a rakuast bug | 00:00 | |
| otherwise it works | 00:03 | ||
| Geth | rakudo: ugexe++ created pull request #6226: RakuAST: default object hash value type to Any when no explicit type is given |
00:05 | |
| japhb | Ah, thanks for taking a look! | 00:34 | |
| Looks like my problem with MUGS-Core and Red begins at: 5c37ccea1 - Initial completion of what became the Great QuantHash Refactor (4 weeks ago) <Elizabeth Mattijsen> | 00:35 | ||
| Actual error is: | |||
| Type check failed in binding to parameter 'key'; expected Attribute but got Str ("\%!___COLUMNS_DATA___") | |||
| at /home/geoff/foss/git/MUGS/MUGS-Core/site#sources/585CBDC6402A5D00B076838C1C3D4DE64EA5D5EC (MUGS::Server::Storage::Driver::Red::SQLite::IdentitySchema):25 | |||
| I'm going to try against actual Rakudo HEAD (I'd been bisecting within the 2026.04 and 2026.05 range) to see if it's still there | 00:36 | ||
| Geth | rakudo: ugexe++ created pull request #6227: RakuAST: drop `:D` from the element type when parameterising a sigil role |
00:37 | |
| japhb | Woah, HEAD fails relatively early in my stack trying to test Crane (which is a dependency of Config::TOML, the thing I was actually trying to install at that point in my script). | 00:52 | |
| Going to retry to see if it's a flapper | |||
| Nope, not a flapper. Failures in Crane's t/list.rakutest - looks like a couple variants of an extra itemization layer. For example: | 01:13 | ||
| [Crane] # expected: $({:path($[0]), :value("zero")}, {:path($[1]), :value("one")}, {:path($[2]), :value("two")}) | |||
| [Crane] # got: $(${:path($[2]), :value("two")}, ${:path($[1]), :value("one")}, ${:path($[0]), :value("zero")}) | |||
| Oh, and it appears to have reversed the elements as well, hmmm. | 01:14 | ||
| Exact same module list known to have worked at rakudo-moar-2026.04-60-gf2273bf17a (which was the last Rakudo commit that worked for *every* module in my stack) | 01:15 | ||
| Geth | rakudo: ugexe++ created pull request #6228: Default object hash value type to Mu under 6.e, Any otherwise |
01:29 | |
| ugexe | re CBOR::Simple, you'll want to add use v6.d since in v6.e we'll switch to the behavior the was breaking CBOR::Simple with RAKUDO_RAKUAST=1 since the new behavior does actually make sense to switch to | 01:32 | |
| but for now it'll work after the above PR | |||
| re Crane: one issue is github.com/rakudo/rakudo/commit/ee30c34ff | 01:40 | ||
| Map/Hash also do Iterable, and their iteration order isn't deterministic, so Hash cmp Hash is now unstable | |||
| japhb | ugexe: Yeah, makes sense about Hash cmp Hash being unstable. Did you see what commit caused the elements to be itemized? | 01:59 | |
| As for CBOR::Simple, I'm not sure whether I prefer doing `use v6.d` to freeze it at the old semantics, or `use v6.e` and make necessary changes. Probably the latter if I can use a syntax that works for all the `v6.e.PREVIEW` releases. | 02:01 | ||
| ugexe | it is probably a red herring | 02:04 | |
| like how .raku renders scalars | 02:05 | ||
| yeah if you can make changes that work on both that could be more ideal. i didn't look into what actual changes you'd need to make, so i can't comment on how correct that might be | 02:06 | ||
| i did see it mention something about making a distinction between Any and Mu, so it might not be easy | 02:10 | ||
| japhb | One of the things I like about CBOR is the care it takes with undefinedness and similar concepts, so I'll have to be careful, yeah. | 02:18 | |
| (SO much closer to Raku's understanding of same than most other data formats) | |||
|
05:55
vrurg_ left
06:10
vrurg joined
07:50
lizmat_ left,
lizmat joined
|
|||
| Geth | rakudo/lizmat-10: a51dbdaa53 | (Elizabeth Mattijsen)++ | src/Raku/ast/package.rakumod RakuAST: test whether some syntax issue is causing breakage Mostly done to also force another CI run |
08:13 | |
|
09:00
shareable6 left
09:03
shareable6 joined
|
|||
| Geth | rakudo/lizmat-10: 03f7535d1f | (Elizabeth Mattijsen)++ | src/Raku/ast/package.rakumod RakuAST: fix various issues in accessor generation - [0] doesn't work, use .head - $name.split("::") is too soo, use nqp::split("::",$name) - remove check for unknown traits, they will be caught elsewhere This brings the build already quite a bit further! |
09:06 | |