🦋 Welcome to the IRC channel of the core developers of the Raku Programming Language (raku.org #rakulang). This channel is logged for the purpose of history keeping about its development | evalbot usage: 'm: say 3;' or /msg camelia m: ... | Logs available at irclogs.raku.org/raku-dev/live.html | For MoarVM see #moarvm
Set by lizmat on 8 June 2022.
[Coke] ls 01:02
... oops
MasterDuke huh, looks like 3853a0f broke the jvm build: 02:51
+++ Preparing installation
Could not find Cool::CompUnit::Repository::Staging in:
    CompUnit::Repository::JavaRuntime<47086169> 02:52
  in throw (gen/jvm/CORE.c.setting:54064)
nine MasterDuke: but that commit only touches RakuAST code? 07:22
lizmat MasterDuke: feels more likely one of my commits did that 08:13
08:15 samebchase left 08:16 samebchase joined
lizmat nine: running into something really weird 08:47
working on ConcreteRoleHOW, I realised that once it is composed, it doesn't need to be composed again 08:48
that appears to save 2 CPU seconds on "make spectest": not a lot, but why not I'd say 08:49
(1250 -> 1248)
however, it consistently breaks the reproducible-build tests
( it was already a bit of a flapper, failing about 1 in 20 times or so) 08:50
any ideas? 08:55
09:40 sena_kun joined 10:05 sena_kun left
Geth rakudo/main: 72c491f89e | (Elizabeth Mattijsen)++ | src/Perl6/Metamodel/ConcreteRoleHOW.nqp
Streamline Metamodel::ConcreteRoleHOW

  - give Collision its own new using positionals, and use native attributes
   where possible
  - make several methods that change non-scalar attributes threadsafe
  - only compose once in the lifetime of the object
  - make sure the typecache only contains unique types, which should help
   in memory usages, and runtime for failing nqp::istype checks
Also left some debugging code in there (but commented out), because iti produced some really weird results.
lizmat greppable6: RoleToRoleApplier 10:54
greppable6 lizmat, Found nothing!
lizmat greppable6: .role_to_role_applier_type
greppable6 lizmat, Found nothing!
releasable6 Next release in ≈3 days and ≈7 hours. There are no known blockers. Please log your changes in the ChangeLog: github.com/rakudo/rakudo/wiki/ChangeLog-Draft 11:00
Geth rakudo/main: 2b6cfaa9b4 | (Elizabeth Mattijsen)++ | 2 files
Don't bother to call applier if there are no roles

The applier logic also had a test for that, but since it is only being called in one location, it made more sense to move the check there. Also, since no one is interested in the return value, don't bother to set anything.
11:18 |Tux| left 11:28 |Tux| joined, Tux__ joined, Tux__ left 12:18 MasterDuke left
lizmat notable6: weekly 16:39
notable6 lizmat, No notes for “weekly”
nine lizmat: no idea. It's been a long time since I have looked at that code. But I wonder why we composed twice in the first place. That doesn't sound healthy. 16:43
lizmat well, it turned out that the compose method on a ConcreteRoleHOW is *not* called more than once 16:44
but in some debugging output I saw the same role name pass by several times, and therefore assumed it was being called more than once 16:45
but apparently the core has more than one ConcreteRoleHOW for "Callable" and "Iterator", apparently
still looking into that 16:46
[Coke] O_o; 17:13
nine Ah, that explains it then. Without composing the order of things probably won't be fixed 17:27
lizmat no, I think my change tickled an underlying issue 17:31
it's gone back to failing 1/20 times or so
17:38 patrickb_ joined 17:40 djinni`_ joined 17:46 djinni` left, patrickb left, patrickb_ is now known as patrickb
lizmat and yet another Rakudo Weekly News hits the Net: rakudoweekly.blog/2024/03/25/2024-...ing-again/ 18:24
18:32 Xliff joined
Xliff Does raku limit reading from /proc in any way? 18:32
lizmat I would be surprised 18:34
Xliff #5534 updated. See last comment. 18:43
.tell nine Can you read the last my last comment on #5534 and see if you have any ideas? Thanks. 18:52
tellable6 Xliff, I'll pass your message to nine
18:53 Xliff left 19:24 sena_kun joined
jdv i think there's something breaking the Parameterizable module 20:02
i don't think i can bisect it easily cause it just chews through mem
(had to kill it while running blin cause it was about to kill my box) 20:03
lizmat: looks like its all yours... github.com/rakudo/rakudo/issues/5539 22:19
[Coke] the raku.land links in that blin ticket are wrong. raku.land/?q=Parameterizable and raku.land/zef:dumarchie/Parameterizable work, but the link is modules.raku.org/dist/Parameterizable 22:40
probably a holdover from the switch to raku.land
lizmat jdv: will look at them tomorrow 22:41
[Coke] jdv++ lizmat++ 22:45
22:55 sena_kun left
jdv thanks 23:10
can the auth from the meta6.json always be used to create that url? 23:14
looks like no 23:16
maybe fallback to ?q= form if can't
[Coke] github.com/rakudo/rakudo/issues/4525 - can we close this? perf issue on old new-disp 23:18