00:11 HarmtH left, HarmtH_ joined 00:13 HarmtH_ left, HarmtH joined 01:14 HarmtH left, HarmtH_ joined 01:15 Kaiepi left, Kaiepi joined 01:45 MasterDuke left 02:09 kvw_5_ joined 02:12 kvw_5 left
tbrowder .tell codesections new release of Math::FractionalPart has routine as discussed...comments welcome, never too late for changes prior to v1. 02:28
tellable6 tbrowder, I'll pass your message to codesections
japhb 🎉‭🎉‭🎉‭ It's out, folks! Heavily based on Cro and Red, the first major milestone release of MUGS, 0.1.0 (codename Proof-of-Concept) is now out! :-) :-) :-) See github.com/Raku-MUGS/MUGS/blob/mai.../v0.1.0.md for details, or ask on #mugs if you'd like to know more. :-) 03:07
05:59 softmoth left
nwc10 So, we seem to have settled on "macosx" (not "darwin"). 07:22
Which means that this breaks expectations:
./rakudo-m -Ilib -e 'say $*DISTRO.name'
(and fails t/spec/S32-str/utf8-c8.t which skips on "macosx" but not "macos") 07:23
and I'm not sure what the right fix is. "blame Apple", but they're bigger than us, so what do *we* do?
This is: Darwin GCC-Farm-MiniMac.local 20.3.0 Darwin Kernel Version 20.3.0: Thu Jan 21 00:06:51 PST 2021; root:xnu-7195.81.3~1/RELEASE_ARM64_T8101 arm64
07:26 frost-lab joined 07:28 domidumont joined 07:42 frost-lab left 08:01 frost-lab joined 08:09 frost-lab left 08:20 frost-lab joined 08:21 sxmx left 08:30 [Tux] left 08:45 [Tux] joined, Xliff joined 08:50 MasterDuke joined 08:52 sena_kun left 09:01 sena_kun joined
MasterDuke japhb: i did a run with the 2021.02, 2021.03, and master tags. how do i generate an html plot comparison of their results? 10:09
japhb: oh, nm, i needed just the `--compare` option to ./analyze 10:16
japhb: next question, is there an easy way to have nqp/rakudo use a moarvm branch? i.e., compare rakudo with moarvm@master and rakudo with moarvm@branch 10:37
10:46 Xliff left 10:55 sxmx joined
MasterDuke japhb: also, just running timeall isn't working, getting `Run command failed to spawn: <path/to/executable/>` for all variants 11:38
nwc10 [Coke]: you might be amused by my comments around 07:22 UTC about "macos" without the "x" 11:58
ugexe nwc10: $*KERNEL.name should give darwin... maybe just change the test 12:10
nwc10 it does give darwin 12:12
I only have access to this particular OS X, er macOS box, so I can't confirm that the test passes on "older". But I guess that Travis CI *does* 12:13
12:14 MasterDuke left 12:31 MasterDuke joined
nwc10 ugexe: yes, but the chain of if/elseif is all $*DISTRO.name - so it will work, but it would look a bit strange. It more feels sort-of-like we should standardise $*DISTRO.name to something consistent on all "OS X"/"macOS" versions (or whatever the branding police say I must call them) 12:47
ugexe my concern is once we do that for this that in 20 years we have 1000 lines of other special cases for various other systems. the nice thing about $*KERNEL stuff is it is unamed-based so it should be more consistent than what $*DISTRO is trying to do 12:51
nwc10 that makes sense. The other specific here is that (I *think* - can't test this) it's really about filesystem - HFS+ (and whatever came next) enforce (some frozen) version of NFC in UTF-8 for their filenames 12:53
or is it the APIs?
you technically could get all sorts of test failures if you have a case insensitve FS mounted on linux
or, I think, told your NTFS to be case sensitive (I think it can do this) 12:54
or just NFS mounted something else onto Win32
13:06 frost-lab left 13:20 tyil left, tyilanmenyn joined 13:32 b2gills left
lizmat notable6: weekly 13:33
notable6 lizmat, 8 notes: gist.github.com/37a01f6c9534573b39...2e4fb164e6
13:39 MasterDuke left
lizmat notable6: weekly reset 13:39
notable6 lizmat, Moved existing notes to “weekly_2021-03-22T13:39:46Z”
13:40 MasterDuke joined
lizmat and another Rakudo Weekly News hits the Net: rakudoweekly.blog/2021/03/22/2021-...mes-begin/ 14:49
15:00 b2gills joined
[Tux] lizmat, did you see my question yesterday? 15:09
lizmat looks back 15:13
[Tux]: because there's a missing \" at the end of line 28? 15:19
$fh.say ("j,\"k\n", $cs, "k\""); 15:20
If that's it, I'll push a fix
[Tux] Well spotted! Thank you 15:23
lizmat pushed 15:25
[Tux] Thanks. FWIW I am working on github.com/nightwatchcybersecurity.../issues/24 15:33
sena_kun lizmat++ 15:36
15:36 sivoais left, sivoais joined 16:31 Kaeipi joined 16:32 Kaiepi left 16:54 Kaeipi left 16:57 Kaiepi joined 17:19 softmoth joined
[Tux] Rakudo v2021.03-1-gbccb2ce2d (v6.d) on MoarVM 2021.03-6-gc10e33d46
csv-ip5xs0.826 - 0.833
csv-ip5xs-207.880 - 8.127
csv-parser26.616 - 27.048
csv-test-xs-200.376 - 0.387
test6.236 - 8.066
test-t1.994 - 2.125
test-t --race0.903 - 0.930
test-t-2033.749 - 36.251
test-t-20 --race9.398 - 10.503
japhb MasterDuke: OK, sounds like there's some work to be done to bring it back up to stable and fix some bitrot. Thank you for letting me know. 18:08
18:09 domidumont left 18:17 softmoth left 18:23 sena_kun left 18:26 sena_kun joined 18:35 HarmtH_ left 18:41 krunen left 18:42 krunen joined
nwc10 OK, I did github.com/Raku/roast/pull/724 but I'm still not sure if it's the best plan 18:53
and it's only tested on one machine, and it's running 20.3.0 18:54
nine nwc10: I think it makes perfect sense to check a more technical name instead of the marketing one. The difference between file systems and operating systems was as much of an issue before as it is after your change. So in total your change is an improvement 19:47
nwc10 nine: the difference between file systems is a conceptual mistake that File::Spec makes, so probably that has been around for decades.
file systems and OSes
nine The flapping t/spec/S17-supply/return-in-tap.t is quite a riddle. It's pretty clearly spesh related. MVM_SPESH_BLOCKING helps provoke it (but not all the time). MVM_SPESH_LIMIT makes it go away, even if set to a ridiculously high value so it shouldn't make any difference at all. 20:06
Ooooh....after some 15 minutes of several different setups running in loops, the one with MVM_SPESH_LIMIT=256 failed 20:09
MasterDuke japhb: github.com/japhb/perl6-bench/pull/31 20:32
simonwillison.net/2021/Feb/19/office-hours/ an interesting idea (tl;dr: OP enjoyed and thought it was useful to have some schedulable hours for one-on-one vtcs with users) 20:36
nine++ is on a roll with the tricky bug finding+fixing 20:42
nine I just noticed. The test in question tests behavior of throwing an exception in a tap. Wouldn't it be funny if this turned out to be the mis-spesh I already identified? 20:49
MasterDuke it'd be great if fixing that mis-spesh took care of a bunch of bugs 20:50
nine Of course that would require an actual fix and not just a diagnosis :) 20:55
japhb MasterDuke: PR reviewed, thank you so much! Couple nits to address, but otherwise looks like you made a lot of very helpful fixes.
lizmat japhb: maybe also time to rename the repo to something more raku like ? 20:57
and change the About: Benchmark and compare Perl 6 implementations against perl5
nudge nudge :-) 20:58
japhb lizmat: *snrk* 20:59
All right, let me get this PR incorporated, then I will.
I've actually never changed a public repo name before, I wonder what I need to do there .... 21:00
lizmat github will redirect URLs from the old name to the new name, so no worries there
japhb Oh that's useful
lizmat I've done this with *many* repos in the raku-community-modules team 21:01
japhb Come to think of it, I wonder if this repo should move to raku-community-modules ... 21:02
lizmat well, that's definitely an option as well
japhb Any advice? You've got rights in the repo anyway IIRC, so you can work on it in either place. 21:03
lizmat I have rights, but no admin rights afaics 21:05
MasterDuke japhb: updated 21:06
japhb lizmat: Huh, I can't see how to give you admin rights. All the people I've added are shown as Collaborator access, but it's not obvious to me how to upgrade that. Maybe because it's in my personal account? 21:11
lizmat possibly
you could try moving it to raku-community-modules 21:12
I should get a notice then and be able to accept it, afaik 21:13
nine So, the error is "Cannot find method 'sink' on 'BOOTCode': no method cache and no .^find_method"
japhb MasterDuke: Merged, thanks!
nine That BOOTCode is the "identity" method. Clearly not something we want to sink in that program.
21:13 maggotbrain joined
japhb lizmat: Should I rename first, or move first? 21:13
lizmat either would be fine with me 21:14
nine Now MVM_6model_find_method_spesh tries to look up a method and if it finds one caches it in a spesh slot. The intriguing thing is that the spesh slot it would cache it in actually does contain the sink method.
So it looks like on repeated runs of this loop, the object we look up the sink method on changes to something unrelated
japhb lizmat: OK, renamed, now trying to transfer
Transfer request sent 21:16
Looks like it needs not only the description change but also a copyright update
lizmat thas arrived!
japhb Cool, I see it there now. 21:17
MasterDuke: Note that the repo just moved ^^ :-) 21:18
MasterDuke yep, just updated my fork and remotes 21:21
lizmat looks like we can drop support for niecza and perlito
japhb lizmat: I think that was part of MasterDuke's PR ... maybe there's some leftovers?
MasterDuke i didn't change the readme
lizmat I just did
21:21 HarmtH joined
japhb Thanks 21:23
lizmat Thank you! :-)
nine MasterDuke: you know what? I actually think it might be the very same mis-spesh 21:24
MasterDuke nice 21:25
japhb lizmat: If you're still in there, might want to remove the paragraph about Niecza builds in the README.
lizmat yeah... will do 21:26
and perlito?
MasterDuke nine: so all you have left is fixing the SSA generation/conversion in spesh. that's what, a 15min job?
japhb Yeah, I think so, sadly. Haven't heard anything about that one for a while either. 21:27
nine MasterDuke: it could even be. If I had a clue about what to change it to :)
MasterDuke lizmat: yeah, i removed the perlito stuff from the components/compilers files
nine MasterDuke: but for that to know I'd need to know _why_ it is the way it is right now, i.e. the design choices that were made 21:28
MasterDuke lizmat: components.json and compilers.pl could also have a bunch of names in them updated 21:34
nine: gotta lay a jnthn-trap over in #moarvm 21:35
japhb: what's the best way to test a moarvm branch? add a new rakudo-<branch> compiler (and/or component?) that does `--gen-moar=<branch>`? 21:37
or just add `rakudo-moar/branch` to @extras in the quickstart multi in bench? 21:40
japhb MasterDuke: Honestly, it's been a little too long for me to remember. I do remember that last time I touched it I felt like VM branch support was poor, and I wanted to throw some of my (then non-existant) extra time into that. 21:53
MasterDuke: I'd say pick a hack for now, and then if the project really picks up again, we can address that weakness.
MasterDuke sure. fwiw,  adding `rakudo-moar/branch` to @extras in the quickstart multi in bench doesn't do it. think i need to add a new compiler/component 21:58
22:00 cog_ joined 22:01 cog left 22:15 cog_ left 22:17 cog joined, cog_ joined 22:18 cog left, cog_ left, cog joined 22:59 lizmat left 23:43 softmoth joined