08:53 [Tux] left 08:57 [Tux] joined
lizmat hmmm I seem to be missing something obvious 10:15
dd $*REPO.candidates("Foo::Bar").head # how do I get to the bytecode file from there ? 10:17
evalable6 Nil = Nil
Saw 1 occurrence of deprecated code.…
lizmat, Full output: gist.github.com/2e56e9c6623506ec1f...7a8d97d904
lizmat ugexe ^^ 10:18
10:31 sena_kun joined
lizmat (if any, of course) 10:40
I'd like to be able to use a module name as an argument to raku.land/zef:lizmat/MoarVM::Bytecode .new 11:17
Geth rakudo/main: b2dd11850e | (Elizabeth Mattijsen)++ | src/vm/js/CompUnit/Repository/FileSystemWithRecording.rakumod
Fix obsolete reference to CompUnit::PrecompilationStore::File
 11:39
lizmat wonders whether nine could give some pointers
12:24 finanalyst joined 12:33 finanalyst left 12:34 finanalyst joined
finanalyst [Coke] I want to test an update to Rainbow (the syntax highlighter) on docs-dev branch, but it is way behind main. Would you object to me deleting the docs-dev branch in order to clear it out? 12:54
Geth rakudo/main: f57a989a1f | (Elizabeth Mattijsen)++ | src/core.c/RakuAST/Fixups.rakumod
Make A, P, and Δ = Capital Greek Delta) produce AST like markup X

Fixes #5721		 13:01
nqp/main: ea21704bc7 | (Elizabeth Mattijsen)++ | tools/templates/MOAR_REVISION
Bump NQP to get gdb extensions fix
 13:20
finanalyst .tell [Coke] I want to test an update to Rainbow (the syntax highlighter) on docs-dev branch, but it is way behind main. Would you object to me deleting the docs-dev branch in order to clear it out? 13:22
guifa So I'm going to probably add one more argument to IO::Socket::Async.bind-udp, nqp::asyncudp, and MVM_io_socket_udp_async 13:40
lizmat ++guifa 13:41
guifa When compiling NQP I cannot seem to get it to recognize that new number of arguments -- I've updated things in asyncsocketudp.c, asyncsocketudp.h, interp.c 13:43
lizmat guifa: perhaps it needs tweaking in nqp/src/vm/moar/QAST/QASTOperationsMAST.nqp 13:45
guifa yeah that's where it's erroring out 13:46
lizmat lines 2284 and later
guifa I'm not seeing what I need to change there -- the only reference I can see to asyncudp anywhere is in the list of ops to be mapped. The error is coming from line 153 in that file 14:06
Trying to work backwards to figure out where it's getting the list.
lizmat maybe asyncudp is in the wrong map now ? 14:08
guifa I don't think so. Going from 6 to 7 args but returning the same value I don't think should shift that 14:39
lizmat then I'm out of suggestions 14:43
perhaps ask on #moarvm
[Tux] Rakudo v2024.12-28-gf57a989a1 (v6.d) on MoarVM 2024.12-10-g1c2ea87e7
csv-ip5xs0.259 - 0.269
csv-ip5xs-201.123 - 1.174
csv-parser1.519 - 1.520
csv-test-xs-200.142 - 0.142
test1.889 - 1.891
test-t0.412 - 0.414
test-t --race0.266 - 0.276
test-t-205.047 - 5.133
test-t-20 --race1.281 - 1.303
 16:12
tux.nl/Talks/CSV6/speed4-20.html / tux.nl/Talks/CSV6/speed4.html tux.nl/Talks/CSV6/speed.log
[Coke] finanalyst: that's fine 18:04
finanalyst [Coke] thanks. 18:05
coleman finanalyst: When your first build is pushed to docs-dev, I will need to hit a switch on the server to turn autodeploy back on. Right now docs-dev is tracking main 19:00
buildkite.com/raku/doc-website/bui...h=docs-dev is where your build will show up. 19:01
finanalyst coleman: can we have a conversation about this? I want to do what should be a relatively minor change to main, but docs-dev is definitely not in sync with main 21:42
coleman you can feel free to it with --force 21:51
that will reset it to whatever your local branch is. it won't affect production.
after that initial force push you can push normally. 21:52
alternatively, you can push your work to a branch with a different name. send me that name and I can do the hard reset stuff on your behalf at first 21:53
finanalyst: I sent you an email with a few example git commands you can run to get docs-dev in sync with the latest main. Feel free to reply in email or here on IRC. I'll see it. 22:03
finanalyst coleman: reading it now. thanks. I'll get back if I need to. 22:04
coleman sounds good
I see that docs-dev is currently building. I'll update the auto-deploy mechanism after it succeeds. 22:15
lizmat timo ab5tract about the DYNAMIC trick yesterday: I couldn't resist: raku.land/zef:lizmat/auto-dynamic 22:26
timo do you think we should make a namespace in our ecosystem called SCNR? 22:27
lizmat Selective Non-Catalytic Reduction 22:28
?
ab5tract Super Nominative Coalescing Revolution! 22:29
Oops, got the c and r swapped 22:35
*c and n
timo Superb Capybaras and Normal Rhinoceri 22:36
finanalyst coleman: I followed your instructions, seems to have worked on my end. 22:40
Just created a PR to make the small changes to the syntax highlighter 22:41
timo lizmat: i actually meant "Sorry, Could Not Resist"
Geth nqp/main: 9a843dc6ed | (Timo Paulssen)++ (committed by timo) | src/vm/moar/QAST/QASTCompilerMAST.nqp
Verify some sizes in qast compiler mast

If any of the section offsets overflow the 32bit offset number that we write to the file, we throw an exception.
If any frame has more than 65534 locals or lexicals, we throw an exception. If there are more than 65534 frames, we throw an exception as well.		 22:42
coleman finanalyst: your updates to docs-dev will deploy themselves in ~15 minutes. I'll be watching to make sure it works. Your PR looks good at first glance. Now we wait for the build. 22:46
finanalyst coleman: the docs-dev build succeeded. I looked at docs-dev.raku.org, the changes have gone through correctly 23:10
[Coke]++ has reviewed them, so I am merging to main 23:11
coleman finanalyst: go for it
finanalyst coleman: thank you.
coleman What changes should we look for 23:12
the new rainbow highlighter upgrade?
23:12 sena_kun left
timo it's not actually a big impact on startup time, but nqp::numcpus or what it's called uses uv_cpu_info which gets detailed information about every CPU core you have, even though all we want is the total number of CPUs 23:18
looks funny in a syscall tracer
the more cores your cpu has, the more syscalls we do at startup, which is also a thing that that benchmark from the other day counts 23:19
coleman we were middle of the road on syscalls, though
timo well, it also depends on how many cores the person who ran the has :D 23:20
finanalyst coleman: In a code block, say the first code snippet in docs-dev.raku.org/language/pod there are `span` elements with classes like RAKUDOC_TEXT instead of POD_TEXT 23:22
Currently, I haven't colored them, so all RAKUDOC_TExT is black, but it could be styled differently 23:23
take a look at the first code block after finanalyst.github.io/samples/rakud...nt%20links 23:25
timo rakudoc now with TeX support? :) 23:31
coleman Your changes should be on docs-dev.raku.org as well as production
I triggered a prod deploy
timo gist.github.com/timo/114ec972bd746...8128bc4e68 here's the syscalls with absolute timings (and time spent inside the syscall in parenthesis next to it) 23:34
you can see the uv_cpu_info is not taking much time
hm, i'm not sure why i don't see filenames here of the moarvm bytecode files that we must be opening for reading here 23:40
i guess that's what i get from "perf trace record" vs "perf trace" 23:41
4.210 into the run time i see the statx for perl6/runtime/perl6.moarvm, 0.010 later the openat; 4.318 is ModuleLoader.moarvm, 7.257 might be the last bit of nqp's pieces, which is nqp/lib/MASTNodes.moarvm 23:45
ah, no P5QRegex and P6QRegex come a bit later 23:46
as a dependency of Perl6/Grammar.moarvm
lizmat would merging P5QRegerx and P6QRegex into a single bytecode file help with startup ? 23:47
timo 35.883 is when we look for and load the CORE.d.setting.moarvm file 23:48
i don't think it would noticeably help
Geth rakudo/main: f12ff834d1 | (Elizabeth Mattijsen)++ | tools/templates/NQP_REVISION
Bump NQP to get gdb extensions fix
 23:50
timo for some reason we are looking for/at ModuleLoader.moarvm multiple times over the course of the runtime. maybe to see if it's changed since the last time?
lizmat the whole module loader bootstrap is .... interesting 23:51
timo :)
lizmat I also wonder whether it would make sense to create a single bytecode files for 6.c / 6.c + 6.d / 6.c + 6.d + 6.e / 23:52
loading the 6.d setting over the 6.c setting needs to do the same work over and over again for every startup 23:53
timo ok, so dependencies+deserialize is generated to first loadbytecode("ModuleLoader.moarvm") 23:54
and the first thing it does is look through the libpath to see where the correct ModuleLoader.moarvm lives 23:57
that's exactly why it shows up in the syscalls
after it finds the actual path, it sees that we've already loaded it and goes on with its lif 23:58
Geth rakudo/main: 0bf65ec794 | (Elizabeth Mattijsen)++ | tools/templates/NQP_REVISION
Bump NQP to get size checks in qast compiler mast
 23:59
timo the time it takes to do that is really not much

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!