buggable 🎺🎺🎺 It's time for the monthly Accidental /win Lottery 😍😍😍 We have 2 ballots submitted by 2 users! DRUM ROLL PLEASE!... 00:00
And the winning number is 4! Congratulations to AlexDaniel-! You win a can of WD40!
Geth rakudo/curfs-candidates-refactor: 270b355edf | (Nick Logan)++ (committed using GitHub Web editor) | src/core/CompUnit/Repository/FileSystem.pm
Improved META6 info-from-path logic

This improves the Distribution that gets created for a CURFS repo prefix that does -not- contain a META6.json file. It will guess the dist name (as the shortest module name in provides) as well as resources (trying to do a reverse platform-library-name on files in resources/libraries).
This essentially allows the following to work:
  `$ perl6 -e 'my $dist = CompUnit::Repository::FileSystem.new(prefix => $*CWD.absolute).candidates("Inline::Perl5")[0]; CompUnit::RepositoryRegistry.repository-for-name("site").install($dist)'`
BenGoldberg .u 🕷 03:41
yoleaux No characters found
Geth roast/cur-candidates-tests: c1de5852d0 | (Nick Logan)++ (committed using GitHub Web editor) | S11-repository/candidates.t
Tests for CUR candidates PRs

roast: ugexe++ created pull request #351:
Tests for CUR candidates PRs
synopsebot SPEC#351 [open]: github.com/perl6/roast/pull/351 Tests for CUR candidates PRs
Geth rakudo: 581edd58e9 | usev6++ | src/core/ThreadPoolScheduler.pm
Add bandaid for JVM: no native array in supervisor

For some reasons using a native num array when starting the supervisor thread does not work correctly on the JVM backend. It looks like a VMArrayInstance is created (instead of VMArrayInstance_n). See also
bartolin .tell pmurias: leaving a note somewhere about bugs in rakudo.js that have the same cause in rakudo.jvm would be great. I don't have a strong opinion about *where* to write it down, though. probably a short ticket/issue tagged with 'jvm' would make sense 06:28
yoleaux bartolin: What kind of a name is "pmurias:"?!
bartolin .tell pmurias leaving a note somewhere about bugs in rakudo.js that have the same cause in rakudo.jvm would be great. I don't have a strong opinion about *where* to write it down, though. probably a short ticket/issue tagged with 'jvm' would make sense
yoleaux bartolin: I'll pass your message to pmurias.
Geth rakudo: de92e34013 | (Aleks-Daniel Jakimenko-Aleksejev)++ | docs/ChangeLog
Start the changelog for 2017.11

So that others have some place to log their changes if they want to.
[Tux] Rakudo version 2017.10-66-gde92e3401 - MoarVM version 2017.10-29-g116c7a13
csv-ip5xs1.132 - 1.141
csv-ip5xs-2013.062 - 13.200
csv-parser12.266 - 12.334
csv-test-xs-200.442 - 0.512
test12.065 - 12.206
test-t3.294 - 3.338
test-t-2060.100 - 60.474
test-t-20 --race20.798 - 21.685
lizmat Files=1229, Tests=75828, 318 wallclock secs (14.62 usr 5.28 sys + 2196.31 cusr 214.70 csys = 2430.91 CPU) 09:42
pmurias bartolin: creating tickets is fine to me 10:24
yoleaux 06:28Z <bartolin> pmurias: leaving a note somewhere about bugs in rakudo.js that have the same cause in rakudo.jvm would be great. I don't have a strong opinion about *where* to write it down, though. probably a short ticket/issue tagged with 'jvm' would make sense
wander I am looking for the implementation of hyper operator "»". Anyone knows its location? 10:25
pmurias on github should I have a [JVM] in the ticket/issue title or is a JVM github label enough 10:29
? 10:36
Geth nqp: fcc0fedfd4 | pmurias++ | src/vm/js/Operations.nqp
[js] Fix 3 argument while with redo
nqp: f924509c17 | pmurias++ | t/nqp/014-while.t
Test that redo works with 3 argument while
rakudo: 6389cea9f2 | (Elizabeth Mattijsen)++ | src/core/ThreadPoolScheduler.pm
Add Worker.worked attribute

  - number of wallclock seconds the Worker actually worked
  - would be nice to get amount of CPU as well, but that's currently not possible
   - OTOH, assuming CPU intensive jobs, this is probably 1:1 with CPU
travis-ci NQP build failed. pmurias 'Test that redo works with 3 argument while' 11:55
travis-ci.org/perl6/nqp/builds/295707257 github.com/perl6/nqp/compare/e4e2f...24509c17f9
Geth rakudo: 2ec29f20b8 | (Zoffix Znet)++ (committed using GitHub Web editor) | CONTRIBUTING.md
Add a blurb about filing Issues

Since this document is linked to from "New Issue" page, it's helpful to include "rules" for bug reports first and place all the info about contributions later.
rakudo: c360d9a28d | (Zoffix Znet)++ (committed using GitHub Web editor) | CONTRIBUTING.md
Add more separation between sections
jnthn lizmat: I'm not so keen on 6389cea9f2; calls to time_n are not free 13:05
Also it's going to give very dubious results with non-blocking awaits 13:06
Would have to be moved into the top level of the method to be accurate 13:07
m: for ^1000000 { }; say now - BEGIN now
camelia 0.2239942
jnthn m: for ^1000000 { nqp::time_n }; say now - BEGIN now
camelia ===SORRY!=== Error while compiling <tmp>
Could not find nqp::time_n, did you forget 'use nqp;' ?
at <tmp>:1
------> for ^1000000 { nqp::time_n }; say now - BEGIN now
expecting any of:
argument list
jnthn m: use nqp; for ^1000000 { nqp::time_n }; say now - BEGIN now 13:08
camelia 0.2580639
Geth rakudo: d48e446ddb | (Zoffix Znet)++ | .github/ISSUE_TEMPLATE.md

GitHub uses it as a template for new issues.
jnthn m: for ^10000000 { nqp::time_n }; say now - BEGIN now
camelia ===SORRY!=== Error while compiling <tmp>
Could not find nqp::time_n, did you forget 'use nqp;' ?
at <tmp>:1
------> for ^10000000 { nqp::time_n }; say now - BEGIN now
expecting any of:
argument list
jnthn m: use nqp; for ^10000000 { nqp::time_n }; say now - BEGIN now
camelia 2.3208378
jnthn m: use nqp; for ^10000000 { }; say now - BEGIN now
camelia 1.8529501
jnthn m: say 1.85 / 2.32
camelia 0.797414
jnthn I think we should perhaps do that in a module and by using .wrap 13:09
Like my StuckAwaits finder
Geth rakudo: 5e158d9c6b | (Zoffix Znet)++ (committed using GitHub Web editor) | .github/ISSUE_TEMPLATE.md
Simplify wording
rakudo: b386963a37 | (Elizabeth Mattijsen)++ | src/core/ThreadPoolScheduler.pm
One last place where we can use push-worker
lizmat sorry jnthn, missed your remarks here so far because I was watching #moarvm, but they're noted 13:27
jnthn: do we have something cheaper than nqp::time_n ? Guess not :-( 13:42
jnthn No, but I don't really think profiling should be in "normal" code, we should instrument stuff from the outside 13:43
lizmat ok, removing $!worked atm 13:45
it was visible in performance for test-t --race
jnthn I guess that area's a hot path given every single scheduled task goes through it
lizmat yup 13:46
Geth rakudo: 2377624729 | (Elizabeth Mattijsen)++ | src/core/ThreadPoolScheduler.pm
Remove Worker.worked, it was too heavy to be run always

Also move update of Worker.total to run-one
  - so we don't lose any completions not seen by supervisor yet at end
lizmat jnthn: is it correct that if we start a ThreadPoolScheduler with initial threads > 0, that we don't get a supervisor ? 14:08
jnthn Um, no 14:10
lizmat ok, I guess then we have a buglet :-)
by default, we start with 0 initial threads, right ?
jnthn Right :)
lizmat ok, so not something that would cause a problem with all of the things I tried so far :-)
ah, scratch that, it *is* doing the right thing 14:15
Zoffix Just reading the code, looks like in this code we create an initial worker in self!general-queue, but then immediatelly throw it away in the following loop and create a new replacement. i.e. we create a new IterationBuffer instead of pushing more workers into the buffer from $!general-workers: github.com/rakudo/rakudo/blob/mast...#L668-L682 14:18
jnthn Zoffix: Yes 14:20
Which we MUST do
Otherwise those lists are not immutable
And we rely on being able to read them elsewhere without a lock acquisition
bbi1h 14:22
Zoffix &push-worker doesn't re-create workers. 14:23
And just clones the buffer; so we could clone the buffer with one worker up in it and add more instead of paying for 1 worker we never use vOv 14:24
Zoffix &
lizmat Zoffix: good point about push-worker 14:29
AlexDaniel pmurias: I think just using the 「JVM」 label is enough, no need to duplicate in the title 14:39
[Coke] yes, -1 on [] tags on github issues. 14:40
er, on [] *subject* tags.
Zoffix lizmat: my point was more about these two lines: github.com/rakudo/rakudo/blob/mast...#L670-L671 Instead of doing `nqp::create(IterationBuffer);` we could do `nqp::clone($!general-workers)` and change `my int $i = -1;` to `my int $i = 0;` and we save one `GeneralWorker.new` 14:41
And this in submethod BUILD and we already writing to $!general-workers there without a lock: github.com/rakudo/rakudo/blob/mast...er.pm#L682 so I don't think the restriction for immutability applies there, so perhaps we can do away with nqp::clone entirely and *just* push 14:42
Like we're still building teh scheduler object, there won't be multiple threads accessing things yet, right? 14:43
lizmat Zoffix: that's for creating the ThreadPoolScheduler object
indeed, so we could just start at 0 :-)
lemme write up some code for that :-) 14:44
Zoffix yeah, probably can. Just need to also add `$!general-queue := nqp::create(Queue);` `self!maybe-start-supervisor();` calls that self!general-queue does and then we avoid paying for the lock it's doing too. 14:45
Zoffix re-&s
Geth rakudo: 5d0ccf73a4 | (Elizabeth Mattijsen)++ | src/core/ThreadPoolScheduler.pm
Use a dedicated first-worker

Instead of cloning an empty buffer and then pushing with push-worker.
rakudo: bfcc43ec9a | (Elizabeth Mattijsen)++ | src/core/ThreadPoolScheduler.pm
Easify ThreadPoolScheduler creation with initial_threads > 0
rakudo: 87e8720202 | (Elizabeth Mattijsen)++ | src/core/ThreadPoolScheduler.pm
Simplify ThreadPoolScheduler creation a bit further
rakudo: 5eeb72a9f5 | (Elizabeth Mattijsen)++ | src/core/Range.pm
Make Range.sum sensible for -Inf/Inf endpoints
lizmat not 100% sure that (-Inf .. Inf).sum should be 0 or NaN 16:10
perlpilot shouldn't it be Inf? 16:12
lizmat perlpilot: assuminng -Inf and Inf have the same distance from 0, I would think not
Zoffix Definite NaN for me there, 'cause you're summing up end points and those are NaN per IEEE
Like endpoints are included in the summing 16:13
m: say -Inf + Inf
camelia NaN
lizmat adjusts accordingly
Zoffix lizmat: well, as I commented on the ticket, that also leaves out excluded Infs :) 16:14
And I don't know what those should be
Zoffix adds another comment
lizmat m: Inf == Inf + 1 16:15
camelia WARNINGS for <tmp>:
Useless use of "==" in expression "Inf == Inf +" in sink context (line 1)
lizmat m: say Inf == Inf + 1
camelia True
lizmat by that logic, endpoints don't matter when Inf is concerned (at least in my book :-)
*exclusion / inclusion of endpoints 16:16
Zoffix Yeah, agreed.
So then (NaN..<anything>).sum => NaN; (<anything>..NaN).sum => NaN; (-Inf..Inf) => NaN; (-Inf..<anything else> .... ummm 16:17
What about -Inf..-Inf? :) 16:18
lizmat Inf
hmmm NaN probably
Geth rakudo: 21efe96ffb | (Elizabeth Mattijsen)++ | src/core/Range.pm
-Inf .. Inf should be NaN, not 0
Zoffix Or Inf..Inf; both are empty ranges. And that makes exclusion points come into play, since Inf^..Inf is non-empty 16:19
m: say -Inf + -Inf
camelia -Inf
lizmat I think the distance between Inf and Inf is Inf, really
Zoffix shrugs; feels like throwing an exception in these cases as OP proposed would save some headache :)
m: say eager Inf..Inf 16:20
I would've thought that'd return an empty list
ZofBot: the worms! They're escaping the can! 16:21
ZofBot Zoffix, {*} is a "magic" token that means "do the next bit of the multi-dispatch here"
camelia (timeout)
lizmat FWIW, these Range.sum changes don't cause any test breakage 16:22
so I assume there simply aren't any tests for it
lizmat goes back to hacking on concurrency stuff 16:24
Zoffix goes back on playing with the Intro P6 brochure :) 16:25
timotimo Zoffix: your own or the one wendy maintains? 16:26
lizmat woolfy smiles 16:28
Zoffix timotimo: both: Wendy's brochure but with my own take on it. It might end up the new Wendy's brochure or might not. So far I'm treating it just as a design project for fun. 16:36
timotimo OK! 16:39
i just happen to have a introducing perl6 brochure now, too
.oO( they're spreading! :)
buggable New CPAN upload: JSON-Fast-0.9.6.tar.gz by TIMOTIMO cpan.metacpan.org/authors/id/T/TI/...9.6.tar.gz 16:56
lizmat still think that URL should point to a modules.perl6.org page, rather than the tarball 17:00
Geth rakudo: 9254396270 | (Elizabeth Mattijsen)++ | src/core/ThreadPoolScheduler.pm
Improve readability of ThreadPoolScheduler
rakudo: a2ae00edef | (Elizabeth Mattijsen)++ | lib/Telemetry.pm6
Remove some superfluous code

  - we don't need to stub Telemetry::Period anymore
  - now that we have snapper, we're probably going to be doing less with T
lizmat dinner& 17:22
Zoffix buggable: toast 17:40
buggable Zoffix, Between 2017.10-78-g21efe96 and 2017.10: 19 (2.14%) modules got burnt; 20 (2.26%) got unsucced; 265 (29.91%) out of 886 modules appear unusable. See toast.perl6.party/ for details.
Zoffix Will looks through the burns and PR the fixes for `readonly $/` issues. 17:41
*later today
timotimo aha, yamlish was hit with readonly $/ 17:48
this diff looks ugly 17:55
i wonder what the right fix for that kind of problem is
Zoffix: gist.github.com/timo/8c1c395114765...a8ed223e55 17:57
Zoffix :o 17:58
timotimo: just rename the method's parameter
Like: method single-quoted($_) { .make: $<value>.Str.subst(/<Grammar::foldable-whitespace>/, ' ', :g).subst("''", "'", :g); }
timotimo ah 17:59
Zoffix Like: method single-quoted($_) { .make: .<value>.Str.subst(/<Grammar::foldable-whitespace>/, ' ', :g).subst("''", "'", :g); }
timotimo well, that's easy
Zoffix &
timotimo fix pushed 18:00
Geth rakudo: 5c96d554c0 | skids++ | 3 files
Implement metamethod shorthand syntax (RT#131478)

Silence Perl5 warning on $.<!alpha>
   Create a faux non-multipart-name desigilname token
   Teach dissect_longname to digest said token
   Teach variable action to generate a p6callmethodhow
   Adjust parsing to handle the meta thrigil
synopsebot RT#131478 [open]: rt.perl.org/Ticket/Display.html?id=131478 Warning about $. when using metamethod
rakudo: e4bb219d0c | (Zoffix Znet)++ (committed using GitHub Web editor) | 3 files
Merge pull request #1178 from skids/rt131478

Implement metamethod shorthand syntax (RT#131478)
synopsebot RAKUDO#1178 [closed]: github.com/rakudo/rakudo/pull/1178 Implement metamethod shorthand syntax (RT#131478)
timotimo why do i keep getting yamlish version 0.0.3 even though i pushed 0.0.4 to github? 18:56
Zoffix zef update?
===> Fetching [OK]: YAMLish:ver<0.0.4>:auth< 18:57
timotimo i thought i did that already 18:58
maybe before the last meta update
Geth rakudo: 86d541f44e | (Elizabeth Mattijsen)++ | lib/Telemetry.pm6
Make Telemetry report configurable

  - both by the column name as well with the method name
  - add :$legend parameter to add legend of columns shown
  - show report generation time with whole second granularity
timotimo lizmat: might also want to repeat the header every n lines 18:59
with the yamlish fix you can now install cro without skipping tests
Zoffix If update don't fix it; try `zef --/cached ...` to avoid the cached version (no idea if `update` clears the cache)
timotimo on latest rakudo
AlexDaniel m: say (5..-5).sum 19:10
camelia 0
AlexDaniel m: say (5..-10).sum
camelia 0
Geth rakudo: 474feb0933 | (Elizabeth Mattijsen)++ | lib/Telemetry.pm6
Add :header-repeat = 32 parameter to Telemetry report
lizmat timotimo: your wish... :-) ^^^
travis-ci Rakudo build passed. Elizabeth Mattijsen 'Easify ThreadPoolScheduler creation with initial_threads > 0' 19:17
travis-ci.org/rakudo/rakudo/builds/295799738 github.com/rakudo/rakudo/compare/2...cc43ec9a9c
gfldex lizmat: any chance to get some RAM stats into Telemetry? 19:18
lizmat point to a fast way of getting access to that info in code, and I'll put it in :-) 19:20
hmmm... perhaps we need to ff timotimo's vmhealth work in Moar ?
Zoffix AlexDaniel: whoops :} 19:22
AlexDaniel: oh wait that's right
AlexDaniel ?
yeah, that's correct
Zoffix is tired :)
timotimo doesn't getrusage give you memory usage? 19:23
jnthn I think there's some memory stuff in there 19:25
Geth rakudo: 5e7dfe52fd | (Elizabeth Mattijsen)++ | lib/Telemetry.pm6
Further Telemetry tweaks:

  - format initialization only needs values
  - remove <s gtq ttq> from default columns
   - they're really only for deep debugging
  - change some double quotes strings to single quoted (for clarity)
  - make sure no header-repeat doesn't explode
lizmat jnthn: which "in there" ? 19:26
jnthn getrusage 19:29
lizmat m: use nqp; dd nqp::elems(nqp::getrusage) # indeed ! 19:30
camelia 18
lizmat okidoki wow
where are these documented ? 19:31
jnthn man7.org/linux/man-pages/man2/getrusage.2.html
And our mapping just returns an array of those 19:32
And there's nqp::const iirc
lizmat and with struct timeval ru_utime flattened ? 19:39
lizmat starts to wonder whether timotimo has JITted getrusage correctly 19:41
timotimo oh? 19:42
there's not really a way to mis-jit it?
what symptom are you seeing?
it's probable that some fields are simply not supported in some OSes 19:43
on my system just outputting getrusage's contents gives me (0 140815 0 16715 75916 0 0 0 14227 1 0 72 0 0 0 0 70 1) 19:44
lizmat timotimo: I seem to recall only expecting 4 return values ? 19:47
timotimo not sure what that's about
lizmat ok, then nvm
sorry for the noise 19:48
jnthn lizmat: yes, timeval flattened in 19:50
gfldex .oO( the sound of progress is quite noisy )
lizmat hmmm... how to find out whether we have a $*SCHEDULER without actually creating one 19:56
I guess looking in PROCESS
is there a specific reason why: 20:00
does not say scheduler => $*SCHEDULER ?
feels wrong to me 20:01
Zoffix I vaguelly remember (or maybe it's a false memory) jnthn saying he made it that way because people wouldn't expect a Proc (which is meant to be blocking) to depend on what the user might've configured $*SCHEDULER be as. 20:34
jnthn Correct. 20:49
We always have a scheduler, though?
lizmat yes
nowadays we do
jnthn I think we found stuff got racey if not :) 20:50
lizmat indeed
Geth rakudo: 3b4f0c6ceb | (Elizabeth Mattijsen)++ | lib/Telemetry.pm6
More Telemetry additions

  - support for max-rss and ix-rss, add max-rss to default report
  - .gist/.Str no longer special case an empty Telemetry / Period
  - only decont a/b once in infix:<->
  - auto-calculate width of "-" footer for a column
travis-ci Rakudo build passed. Elizabeth Mattijsen 'Simplify ThreadPoolScheduler creation a bit further' 21:22
travis-ci.org/rakudo/rakudo/builds/295805502 github.com/rakudo/rakudo/compare/b...e8720202c0
gfldex lizmat: tyvm :) 21:23
lizmat gfldex: more goodies coming :-)
El_Che Native call is still causing non deterministic fails on the native call tests: gist.github.com/nxadm/cd7559b6611f...test-L2936 . I thought AlexDaniel told me it was supposed to be fixed (Rakudo 2017.10) 21:33
I don't post a bug because it's a known issue
just info that it doesn't seem fixed after all
AlexDaniel that does not seem to be the same issue? 21:34
El_Che I don't remember the exact test (I can look it up), but it was nativecall ánd very random (1 out of 10 maybe)
I think I see it even less now 21:35
but still there
AlexDaniel El_Che: I don't think that this particular problem is a known issue, please file a bug report
gfldex lizmat: gtc may be a bit narrow, see gist.github.com/gfldex/d83539e9451...adb68654cc
El_Che AlexDaniel: I will do
Geth roast: 6586549d82 | usev6++ | 6 files
[jvm] Skip some tests that fail badly
rakudo: ac738b988e | usev6++ | t/spectest.data
[jvm] Don't run test files that hang regularly
El_Che AlexDaniel: I will try to gather some dat first 21:47
Geth roast: d0693d9c91 | usev6++ | S32-str/parse-names.t
Fix typo in description
AlexDaniel El_Che++ 21:53
El_Che Every commit to the rakudo-pkg repo equals 12x compiling rakudo, so we'll end up with something :) 21:55
Geth rakudo: 7144dc290c | (Elizabeth Mattijsen)++ | lib/Telemetry.pm6
Still more Telemetry goodies

  - export all of the metrics as subs
   - so calling "max-rss" will give you the current max-rss value
  - added some internal documentation for clarity
  - removed protos for methods, as we don't need them outside the setting
  - multiple calls to snapper() now allow changing of period size on-the-fly
rakudo: 73e1faaa71 | (Elizabeth Mattijsen)++ | lib/Telemetry.pm6
Increase the size of gtc/ttc, gfldex++
travis-ci Rakudo build canceled. Elizabeth Mattijsen 'Still more Telemetry goodies
travis-ci.org/rakudo/rakudo/builds/295977089 github.com/rakudo/rakudo/compare/a...44dc290c48
buggable [travis build above] ☠ Did not recognize some failures. Check results manually.
gfldex lizmat: if %format would have a ref to the sub that provides a field, one could output the whole lot easily 22:03
lizmat gfldex: ??
gfldex %format = max-rss => [" max-rss", { .max-rss.fmt('%10d') }, "Maximum ...", &max-rss ], ... 22:05
%format<max-rss wallclock>[*,2]».() 22:06
lizmat I don't understand what it would bring to add the ref to the associated sub
gfldex Telemetry.new.gist is very lean 22:08
by building a connection between fields, legend entries and the subs, one can make a fat .gist with ease
lizmat ok, I see where you're going 22:09
gfldex I still got my golfed httpd around. And I want to blog about Telementry. It would be nice if easy things would be easy. :) 22:10
lizmat yeah yeah... typing as fast as I can :-) 22:11
Geth rakudo/curfs-candidates-refactor: 683df8bf6c | (Nick Logan)++ (committed using GitHub Web editor) | src/core/CompUnit/Repository/FileSystem.pm
Improve distribution lookup cache

For a single dist CUR like this it doesn't really matter, but following the same template as CURI allows run time generated CURFS lookup to work.
roast/cur-candidates-tests: 530b72e5a1 | (Nick Logan)++ (committed using GitHub Web editor) | S11-repository/candidates.t
Ensure module lookup after installation works

Check that we don't cache when .resolve'ing a missing module by immediately installing said module and seeing if .resolve returns non-Nil.
Zoffix m: .say for <a b>.any 22:26
camelia (Mu)
Zoffix oh wait, there's already a bug for that..
RT#131639 22:27
synopsebot RT#131639 [new]: rt.perl.org/Ticket/Display.html?id=131639 [BUG] Junction:D.iterator returns an iterator that .pulls-one a Mu type object
Geth rakudo: cf1742dca0 | (Elizabeth Mattijsen)++ | lib/Telemetry.pm6
More Telemetry goodness

  - add support for id-rss (which is usually 0 for me)
  - add legend by default when in END processing
rakudo: 91e00e6801 | (Elizabeth Mattijsen)++ | lib/Telemetry.pm6
Telemetry exports T, allowing for T<cpu wallclock>
lizmat gfldex: ^^^
gfldex :)
lizmat: max-rss seams to be in kb not bytes 23:16
lizmat m: use Telemetry; say T<max-rss> 23:20
camelia 92252
lizmat mine says: 84086784
argh, I guess MacOS is returning bytes rather than K's
Geth rakudo: 96751ee87c | (Elizabeth Mattijsen)++ | lib/Telemetry.pm6
Add Telemetry is-rss,min-flt,maj-flt,nswap,inblock,outblock

Which all return 0 apparently on MacOS
rakudo: 7e00908cb2 | (Elizabeth Mattijsen)++ | lib/Telemetry.pm6
Normalize all Telemetry x-rss to Kbytes

  - on MacOS reported values are bytes
  - on Linux they are apparently Kbytes
  - Kbytes seems to be the more sensible scale
travis-ci Rakudo build failed. Elizabeth Mattijsen 'Make Range.sum sensible for -Inf/Inf endpoints' 23:52
travis-ci.org/rakudo/rakudo/builds/295824731 github.com/rakudo/rakudo/compare/8...eb72a9f514
buggable [travis build above] ✓ All failures are due to: GitHub connectivity (1 failure).