AlexDaniel timo: hey, you've been using neo layout, right? 00:00
timo: does it work in chromium?
like shift+arrows and stuff like this?
timo yeah 00:01
AlexDaniel :-|
timo something wrong?
AlexDaniel yea, my layout is completely useless in chromium, a lot of things just don't work 00:02
timo oh crap :(
AlexDaniel and I've been ignoring this for years by simply labeling chromium as garbage
timo some IDEs are unhappy with the neo2 arrow keys
especially on windows 00:03
AlexDaniel now a lot of things use electron (which is based on chromium as far as I can see) and I'm screwed
MasterDuke_ timo: that's very well written, looks great
only one minor suggestion, and one minor thought 00:04
timo (and before sending it off i'll remove the hour counts again)
MasterDuke_ i think it's a little unclear what an end user would use the heap snapshot profiler for
timo ah, sure, i can write more about that 00:05
haha, yeah, now that i look at it again, it's many words saying very little
MasterDuke_ and i believe Devel::NYTProf is pretty popular in the Perl 5 world, maybe some comparison to it? 00:06
timo NYTProf is line-oriented, right?
MasterDuke_ but i don't really have concrete suggestions about that thought, so feel free to ignore 00:07
believe so
timo how do you like the initial part to the heap snapshot section?
MasterDuke_ good 00:11
timo the sampling profiler jnthn has been talking about will be more similar to NYTProf as it will give line-by-line accuracy, but it'll be probabilistic rather than measuring everything 00:13
so it'd be a lot like perf
MasterDuke_ btw, between 2017.04 and 2017.05, a --profile-compile of stage parse of the setting increased in size by 50mb, and the compressed size went from 50mb to 115mb
timo wow
the heap snapshot analyzer will also want a bit more memory use tweaking %) 00:14
i'm glad i still have about 21h until the deadline hits 00:18
i should send a mail to "take back" my application so the first version doesn't get published
done 00:27
MasterDuke_ afk for a while, good luck 00:30
timo i'll go to bed first :) 00:31
Geth rakudo: Gnouc++ created pull request #1159:
use 'time' as default seed for srand		 02:44
AlexDaniel what was the issue with make spectest on jvm? 03:00
like, individual files are passing ok, but not make spectest 03:01
Geth nqp: Gnouc++ created pull request #371:
add document for srand op		 03:08
roast: Gnouc++ created pull request #319:
add test for srand has default value		 03:17
cuonglm m: say srand; 03:18
camelia 5===SORRY!5=== Error while compiling <tmp>
Calling srand() will never work with declared signature (Int $seed --> Int:D)
at <tmp>:1
------> 3say 7⏏5srand;
AlexDaniel buggable: tag JVM 03:27
buggable AlexDaniel, There are 49 tickets tagged with JVM; See fail.rakudo.party/t/JVM for details
AlexDaniel o/ see you all in ≈8 hours 04:54
lizmat Files=1223, Tests=67805, 295 wallclock secs (11.02 usr 4.78 sys + 1953.74 cusr 208.60 csys = 2178.14 CPU) 06:18
bartolin for the records, I've seen the following issues with the jvm backend (but didn't have the time to take a closer look) 06:39
1) 'make' does not work before 'make install': I believe that happens, because we use the EvalServer -- and that doesn't look for BOOTSTRAP.nqp in ./blib 06:40
2) 'make test' fails a lot of nativecall tests. afaik that happens because the op 'nativecallinvoke' was introduced with commit cd7dc4ce93 in lib/NativeCall.pm6 (but isn't known on the JVM backend) 06:41
3) a lot of tests for $! seem to fail. i think that happens since commit 1a4d94930c (there was a call to nqp::setmessage removed from method throw) 06:43
4) the EvalServer seems to leak memory. it's no longer possible to run 'make spectest', even with -Xmx6000m 06:44
i'd be interested in suggestions how to fix 1) and 4), especially :-) 06:46
Geth nqp: ea6be29658 | (Cuong Manh Le)++ | docs/ops.markdown
add document for srand op
 07:10
nqp: b6ba7362e0 | lizmat++ (committed using GitHub Web editor) | docs/ops.markdown
Merge pull request #371 from Gnouc/master

add document for srand op
lizmat afk& 07:21
Geth rakudo/nom: c3a71acb33 | (Samantha McVey)++ | t/spectest.data
Add collation tests to spectest.data
 08:17
[Tux] This is Rakudo version 2017.08-155-gc3a71acb3 built on MoarVM version 2017.08.1-171-gcf95892e 10:06
csv-ip5xs 1.371 - 1.484
test 9.777 - 10.006
test-t 3.523 - 3.539
csv-parser 10.881 - 12.371
Skarsnik Hello 10:58
samcv releasable6, status 11:20
releasable6 samcv, Next release in ≈7 hours. 2 blockers. 122 out of 155 commits logged
samcv, Details: gist.github.com/ae02f7b965262c8c45...ba6ec9923e
Geth nqp: 3b1858d6d9 | (Moritz Lenz)++ | src/QRegex/P6Regex/Grammar.nqp
Suppress warning in /a b+/

the warning (about the space not being significant) makes sense inside a normal literal, but when the last character is quantifier, it is likely just somebody splitting off the character to emphasize the quantified atom		 12:19
moritz s/quantifier/quantified/ 12:20
too late to use the example in the book :/
timo i like this change 12:21
travis-ci Rakudo build failed. Zoffix Znet 'Merge pull request #1157 from book/book/enum-pred-succ 13:32
travis-ci.org/rakudo/rakudo/builds/275895080 github.com/rakudo/rakudo/compare/2...0156146917
buggable [travis build above] ✓ All failures are due to timeout (0), missing build log (0), GitHub connectivity (1), or failed make test (0).
lizmat wonder how Rakudo Perl 6 would be doing in this: greenlab.di.uminho.pt/wp-content/up...perSLE.pdf 13:36
jnthn Hmm, intersting paper :) 13:48
timotimo dangit, this bench either makes one single major collection or a whole ton of minor collections 13:52
i just want a program that makes like 5 snapshots %)
may want to build a "limit" option or something :D 13:53
yeah! rc-forest-fire with a 2x2 grid gives me 5 snapshots 13:56
bartolin oh, AlexDaniel++ (creating tickets from my earlier messages today). i should have done that myself 14:31
bartolin blushes
AlexDaniel RT #132101 RT #132102 RT #132103 RT #132104 14:32
synopsebot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=132101
Link: rt.perl.org/rt3/Public/Bug/Display...?id=132102
Link: rt.perl.org/rt3/Public/Bug/Display...?id=132103
Link: rt.perl.org/rt3/Public/Bug/Display...?id=132104
AlexDaniel bartolin: thanks
⚠⚠⚠ Last time I checked, nothing was blocking the release seriously. If you're seeing some issues, please contact me. ⚠⚠⚠ 14:35
if there are any volunteers to run the spectest on windows, that'd be appreciated 14:36
stmuk AlexDaniel: OK I'll try now 14:37
AlexDaniel Skarsnik_: how long does it take to run toaster for you?
releasable6: status
releasable6 AlexDaniel, Next release in ≈4 hours. 2 blockers. 122 out of 155 commits logged
AlexDaniel, Details: gist.github.com/b05b81274185737530...7dd828df2a
AlexDaniel Skarsnik_: if you have resources to do that, I think about now will be the right time :) 14:38
Skarsnik_ 4h?
AlexDaniel kick it off then just in case?
FWIW I'd probably take a look at those few modules that fail because of the buffering change, but I think we want it to be in the release anyway. The fallout is minor, and these modules are truly missing .close somewhere I think 14:40
travis-ci Rakudo build failed. Jonathan Worthington 'Don't attempt non-blocking await if holding locks 14:41
travis-ci.org/rakudo/rakudo/builds/275907719 github.com/rakudo/rakudo/compare/8...6d1e24f927
buggable [travis build above] ☠ Did not recognize some failures. Check results manually.
Geth rakudo/nom: e6a695b272 | (Samantha McVey)++ | t/spectest.data
Fix spectest.data path for Collation tests
 14:43
AlexDaniel Geth: ver github.com/rakudo/rakudo/commit/06e20f80 15:01
Geth AlexDaniel, version bump brought in these changes: github.com/perl6/nqp/compare/2017....5cdbdd3... No newline at end of file
AlexDaniel fucking Geth
( github.com/perl6/geth/issues/8 ) 15:03
Geth: ver github.com/perl6/nqp/commit/9a2ba3...3a49c83396 15:05
Geth AlexDaniel, version bump brought in these changes: github.com/MoarVM/MoarVM/compare/2...2-g4a7248e
rakudo/nom: bdbb7e4aca | (Aleks-Daniel Jakimenko-Aleksejev)++ | docs/ChangeLog
Log remaining changes

Deliberately not logged: 29691b2f c4043b06 b7ab48ee fe719405 06e20f80 cc64a758 ce95e162 dea0a085 a8e0352b 3de6f338 ac8e099b c3a71acb 66015614		 15:35
AlexDaniel I think Test.pm is buffering its output needlessly 15:38
does this idea ring the bells for anybody?
I'm running parallelized tests in whateverable and I see ===( 172;53 1/? 1/? 1/? 1/? 1/? 1/? 1/? 1/? )=============== 15:39
the tests are running but I'm guessing that the result only shows up once the handle is closed 15:40
these are just random thoughts though
stmuk I may be seeing something similar on windows - at least I'm not seeing test output around S32-io 15:41
timotimo next star's gonna be at least one more month out, right? 15:51
stmuk 2017.10 hopefully
timotimo good 15:55
so i don't really have a reason to rush my extra-proflie-options feature in before the release 15:56
AlexDaniel there's also this module that passes its tests just fine but fails under prove: github.com/azawawi/perl6-ncurses/issues/16 15:58
stmuk turns off windows av 16:04
I'm seeing a few windows roast failures even oon 6.c-errata 16:21
AlexDaniel stmuk: more info? 16:22
stmuk t\spec\S02-literals\quoting.t t\spec\S02-magicals\KERNEL.rakudo.moar t\spec\S02-magicals\args.t 16:25
AlexDaniel I'm pretty sure that quoting.t was like this for a while 16:30
this issue for example: github.com/perl6/roast/issues/197
stmuk yeah and at least some of those (like the precomp failure) are fixed 16:32
AlexDaniel rakudo on windows definitely needs more love 16:34
so I guess it's alright? 16:36
stmuk SNAFU :)
timotimo Situation Needs A Fine Understanding 16:38
stmuk ha 16:40
AlexDaniel well, at least not TARFU like we had it before last release 16:43
stmuk Seaman Tarfu, a brother of cartoon character Private Snafu 16:47
AlexDaniel folks, we have a blocker 16:48
I think. Maybe.
this one: rt.perl.org/Ticket/Display.html?id=132083
and here's why
say $value; # -2147483648 16:49
say -2147483648; # -2147483648
say $value == -2147483648; # False
evalable6 (exit code 1) 04===SORRY!04=== Error while compiling /tmp/XgeG6q6Q4b
Variable '$value' is not declared. Did you mean '&values'?
at /tmp/XgeG6q6Q4b:1
------> 03say 08⏏04$value; # -2147483648
-2147483648
(exit code 1) 04===SORRY!04=== Error while compiling /tmp/gJL6RZzlkm
Variabl…
AlexDaniel, Full output: gist.github.com/9d78faf157d8d41e74...3272c4e8b1
AlexDaniel shut up evalable6
anyway, these are two kinda identical numbers, but they == to False 16:50
I remember seeing something similar… hmm
which may simply be a side effect of RT #131994 16:51
synopsebot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=131994
AlexDaniel but $value * 1.0 is still -2147483648 16:52
stmuk github.com/perl6/roast/issues/320
AlexDaniel stmuk: thanks for documenting it 16:54
stmuk np 16:55
AlexDaniel we can of course say that number -2147483648 is cursed and move on, but uhh… 16:56
stmuk more positively I saw a full PASS on errata on FreeBSD 11 yesterday! 16:58
AlexDaniel I'm currently golfing that -2147483648 issue 17:03
bartolin stmuk: ooc, I'm seeing failures for roast master in S24-testing/11-plan-skip-all.t (aborted) and S29-os/system.t (test 35) on FreeBSD 10. do you have those on FreeBSD 11 as well? 17:13
stmuk: 6.c-errata is clean for me as well 17:14
stmuk bartolin: Yes I saw failures on roast master recently but I don't have a record 17:17
bartolin stmuk: thanks. no need to look again, i was just a tad bit curious 17:20
AlexDaniel okay, there we go 17:25
m: my $value = -0x7FFFFFFF - 1; say $value; say -2147483648; say $value == -2147483648;
camelia -2147483648
-2147483648
False
AlexDaniel bisect: my $value = -0x7FFFFFFF - 1; say $value; say -2147483648; say $value == -2147483648; 17:26
bisectable6 AlexDaniel, On both starting points (old=2015.12 new=bdbb7e4) the exit code is 0 and the output is identical as well
AlexDaniel, Output on both points: «-2147483648␤-2147483648␤False»
AlexDaniel :O what
commit: all my $value = -0x7FFFFFFF - 1; say $value; say -2147483648; say $value == -2147483648;
committable6 AlexDaniel, gist.github.com/78b55a0024486a7008...ce659335ee 17:27
AlexDaniel :O what
j: my $value = -0x7FFFFFFF - 1; say $value; say -2147483648; say $value == -2147483648; 17:29
camelia -2147483648
-2147483648
True
AlexDaniel okay
m: my $value = -0x7FFFFFFF - 1; say $value; say -2147483648; say $value < -2147483648; 17:30
camelia -2147483648
-2147483648
True
AlexDaniel m: my $value = -0x7FFFFFFF - 1; say $value; say -2147483648; say $value == -18446744071562067968 17:40
camelia -2147483648
-2147483648
True
AlexDaniel well, we can try bisecting it 17:48
timotimo hm, this isn't close to the smallbigint border is it?
AlexDaniel bisect old=2014.01 new=2014.02 my $value = -0x7FFFFFFF - 1; say $value; say -2147483648; say $value == -2147483648; 17:49
timotimo: that's 2³¹ I think
timotimo m: say 2³¹
camelia 2147483648
timotimo it is
AlexDaniel but negative
hello, bisectable?
timotimo that means the uppermost bit is also set, right?
AlexDaniel right 17:50
bisect: old=2014.01 new=2014.02 my $value = -0x7FFFFFFF - 1; say $value; say -2147483648; say $value == -2147483648;
bisectable6 AlexDaniel, Bisecting by output (old=2014.01 new=2014.02) because on both starting points the exit code is 0
AlexDaniel, bisect log: gist.github.com/d9e374de485f03a4c6...71cca6745e
AlexDaniel, (2014-02-08) github.com/rakudo/rakudo/commit/8e...865598e211
timotimo an nqp bump, heh. 17:51
AlexDaniel Geth: ver github.com/rakudo/rakudo/commit/8e...865598e211
Geth AlexDaniel, version bump brought in these changes: github.com/perl6/nqp/compare/2014....7-gdcc6d47
AlexDaniel Geth: ver github.com/perl6/nqp/commit/dcc6d4...22ae2da283
Geth AlexDaniel, version bump brought in these changes: github.com/MoarVM/MoarVM/compare/2...1-ge9612cb
AlexDaniel timotimo: well here's the list ^ 17:52
some of your commits there also :)
timotimo oh, could it be a problem with variable int encoding?
Skarsnik m: my uint8 $a = 255; say $a++; 17:54
camelia -1
timotimo that's been a problem forever :)
Skarsnik should that throw an error ? xD 17:55
timotimo need to make dinner, almost too hungry to think ATM
native ints are supposed to behave like that i think
AlexDaniel timotimo: Maybe RT #131994 can give you some idea also? 18:00
synopsebot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=131994
AlexDaniel huggable: no release?
huggable AlexDaniel, blead build fails on windows. Options: (1). Fix it. Find a way. (2). Revert the security fix. | See irclog.perlgeek.de/perl6-dev/2017-...i_15043376
AlexDaniel huggable: no release? :is: See RT #132083 18:01
synopsebot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=132083
huggable AlexDaniel, Added no release? as See RT #132083
synopsebot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=132083
AlexDaniel timotimo: I think I can try bisecting it manually
in a hope that it will give us a better idea
but rakudo 2014.01-2014.02… uhh… 18:02
ok, bisect is starting alright 18:29
it's definitely not a change in nqp or whatever, so now bisecting moar…
Skarsnik AlexDaniel, file updated for toast 18:38
AlexDaniel link?
Skarsnik www.nyo.fr/~skarsnik/tmp/toast.sqlite.db 18:39
AlexDaniel Skarsnik: thanks, will look after I'm finished with the int issue 18:40
.in 7d teach bisectable to bisect moar, you idiot. 18:49
yoleaux AlexDaniel: I'll remind you on 23 Sep 2017 18:49Z
AlexDaniel binary search is not so binary when some commits are unbuildable… 19:08
Geth roast: d776a06e52 | (David Warring)++ | S02-types/subset.t
tests for RT #132073 - isa on a subset type object
synopsebot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=132073
timotimo the serialization function changes at either one F more or one F less in the number, fwiw 19:41
gfldex jnthn: the segfault in the long running golfed httpd seams to be gone 19:46
AlexDaniel timotimo: I really don't know if git bisect will give any meaningful result here 19:47
timotimo oof, all the commits are unbuildable?
AlexDaniel I *think* it narrowed it down to 12 commits 19:48
but it just keeps going…
probably wants to make sure that all 12 of them are unbuildable
but it feels like I've been doing it for way longer…
timotimo try "git bisect visualize" 19:50
it might require gitk 19:51
AlexDaniel I'm doing it over ssh
timotimo ah
AlexDaniel well, it's definitely not stuck… just doing something useless 19:52
timotimo i'm going to gdb around a bit in the place i'm expecting trouble
Skarsnik releasable6, status 20:05
releasable6 Skarsnik, Next release will happen when it's ready. 2 blockers. 154 out of 157 commits logged
Skarsnik, Details: gist.github.com/4b6de50e30059383f9...d54bfdba2c
AlexDaniel maybe I should just skip all the way through
timotimo: gist.github.com/AlexDaniel/6e6b040...6ad50e514a 20:08
timotimo well, crap ... 20:09
having the commit titles there would be nice 20:10
samcv good * all 20:24
timotimo good * samcv 20:25
debug output shows i'm deserializing a 2147483648 20:26
m: say 2147483648.base(16)
camelia 80000000
timotimo m: say 2147483648.base(16) - 1
camelia 79999999
timotimo oops
m: say 2147483647.base(16) 20:27
camelia 7FFFFFFF
timotimo i'm getting the value that's in the source, but + 1, so it must be the absolute value or something?
because i'm storing -0x7FFFFFFF - 1 in the module 20:28
jnthn gfldex: Ah, good; I figured that would be the same issue I uncovered with a short-running httpd hammered with requests. 20:44
Geth rakudo/nom: 24d6c66f30 | (Samantha McVey)++ | docs/ChangeLog
Update ChangeLog with more Unicode/string related MoarVM changes
 20:46
AlexDaniel samcv++ 20:47
samcv so many speedups :)
BooK AlexDaniel: the good and bad commits have branches (or at least refs pointing to them), so you can still somewhat see what happens with git log --graph and the proper revlist 20:52
AlexDaniel right
BooK if you're still trying to "see" the current bisect
AlexDaniel we really need a volunteer for github.com/perl6/whateverable/issues/40 20:53
nine AlexDaniel: maybe the simplest way would be a unified repository 20:56
AlexDaniel nine: I doubt it. One possible strategy is described in github.com/perl6/whateverable/issu...-326310996
samcv a lot of modules i think are broken. i can't `use META6:ver<0.0.4..*>` 20:57
even though i have a newer version than that installed
AlexDaniel samcv: plz more info, I have toaster results in front of me and everything looks fine
BooK AlexDaniel: I'd been interested in seeing the code doing the current bisect stuff
samcv i can do `use META6` or `use META6:auth('github:jonathanstowe')` but not `use META6:ver<0.0.4..*>` (with or without the auth)
and my friend is having the same issue
BooK nine: what do you mean by "unified"? 20:58
AlexDaniel BooK: github.com/perl6/whateverable/blob...6#L41-L125
nine samcv: ver<0.0.4..*> is a string
samcv my friend says: `zef install MongoDB::Client MongoDB::Server BSON::Document` gave me the missing META6:ver<0.0.4..*> error 20:59
i've got the same error trying to install stuff
nine, so the module needs to be changed? to what
ver(0.0.4..*) ?
AlexDaniel BooK: the point is that bisectable does not build anything, and also it does not use “git bisect run” 21:00
BooK oh
nine m: say v0.5 ~~ v0.0.4..*
camelia True
AlexDaniel BooK: we just have to make some extra builds for different nqp and moar revisions, and then teach it to handle these
BooK so you have a linear history with points in time?
nine m: say v0.5 ~~ v0.0.4+
camelia True
nine m: say v0.5 ~~ Version.new('0.0.4..*') 21:01
camelia False
nine m: say v0.5 ~~ Version.new('v0.0.4..*')
camelia False
nine m: say v0.5 ~~ Version.new('v0.0.4+')
camelia False
nine m: say v0.5 ~~ Version.new('0.0.4+')
camelia True
samcv how do i grep all modules?
AlexDaniel BooK: no, we jush have everything pre-built. We still ask git to jump between version
samcv: greppable6 ?
samcv greppable6, ver<0.0.4..*> 21:02
AlexDaniel samcv: or use github.com/moritz/perl6-all-modules/
nine samcv: so either ver<0.0.4+> or ver(v0.0.4+) or ver(v0.0.4..*)
greppable6 samcv, gist.github.com/a78573bb8fa48a36b9...5991f5c65f
AlexDaniel greppable6: ver\<0\.0\.4\.\.\*\>
greppable6 AlexDaniel, Found nothing!
AlexDaniel BooK: point is, it takes a few minutes to build moar-nqp-rakudo, and typically bisect needs about 12 test IIRC 21:03
samcv looks liek Semaphore::ReadersWriters:ver('0.2.6'):auth('Marcel Timmerman')
AlexDaniel BooK: assuming that everything is buildable and you never “skip”, which is sometimes not the case
samcv my friend is gonna PR it
thanks guys :)
AlexDaniel BooK: so you multiply 12 by 2 minutes, or even 1 minute if you're generous, and you don't get anything pretty :)
BooK: when talking about bisecting moar or nqp, I don't think building it on the fly is a good solution, because why wait 21:04
BooK sure
AlexDaniel bisectable currently has over 6000 builds IIRC, and it takes almost no space
(because we compress things in a smart way) 21:05
BooK are the builds themselves stored in git?
AlexDaniel no
BooK cause you could build a similar dag structure, but with builds instead of source
and then use git bisect in the "binary" repo 21:06
AlexDaniel you can, but you win more free space by not doing that (and in the end it is also more flexible)
here's some info on that: github.com/perl6/whateverable/issues/23
so “quick-access” builds are compressed with zstd (these includes all commits after 2015.12 and all tags like 2014.01) 21:07
now as for everything before 2015.12, here's the info: github.com/perl6/whateverable/issues/117
basically it puts several adjacent builds in a pack of 20, and compresses it all with lrzip 21:09
samcv here's the issue gist.github.com/KingNoosh/bdbdffcd...e1-txt-L21
i think this may be a different issue
BooK ok, so the tracking is done by naming the "tarball" with the commit id
AlexDaniel yeah 21:10
BooK using git bisect run sounds like just writing down the steps once you know the commit id in a script 21:11
since git doesn't really care about what you do with the checkout it gives you, but only about the exit status
AlexDaniel and if it's in deep storage mode, it's just a symlink to a bigger lrz archive
we've been using git bisect run before, and then switched from it
git bisect run itself is just a shell script with some sugar in it 21:12
I'm not entirely sure what was the motivation behind it, but here's the commit: github.com/perl6/whateverable/comm...c96f454ef6 21:13
well, I'm not sure today, back then it was an obvious thing to do :)
BooK the commit message is light on details :-)
in fact, I jumped in the conversation with no background, so I have no idea what the current issue is, and if I can contribute anything to it 21:14
I just heard "git" and jumped in ;-)
AlexDaniel BooK: well… I just saw somebody interested in the question and started pouring out as much as I know :) 21:15
but implementing moar/nqp/rakudo bisect for bisectable requires this: 21:16
BooK there's probably a range of moar/nqp combinations that can be used with a given rakudo commit
AlexDaniel 1) create the required builds somewhere here: github.com/perl6/whateverable/blob...n/build.p6 (for example, with RAKUDOSHA-NQPSHA-MOARSHA.zst filenames)
2) teach bisectable to use these, somewhere here maybe github.com/perl6/whateverable/blob...6#L41-L125 21:17
3) …
BooK so that limits the combinatorial explosion somewhat
AlexDaniel 4) PROFIT!
BooK: typically… well, ok, *sometimes* when you bisect moar you just use whatever version is declared 21:18
you may get some unbuildable or broken builds, but that's ok
when you bisect nqp***
so there's no 2D problem with this, or so I think
BooK 2D problems? 21:19
AlexDaniel and on Moar bumps you just go through the range of moar commits, so again that's pretty much linear
trying out different nqp-moar versions
BooK I see the benefit of bisecting rakudo, but I'm not sure I understand the use case with the combination of the three 21:20
or is it: given one somewhat broken rakudo commit, find the nqp/moar that actualy broke it? 21:21
AlexDaniel yea
here's one real life example: github.com/rakudo/rakudo/commit/51...29cde39c1f
it could have had some NQP commits too, but this shouldn't matter as much 21:22
BooK ok, so rakudo keeps track of the "new minimum version expected"
AlexDaniel and nqp does too for moar 21:24
BooK but it's only minimum, so someone can compile it with newer version and see how that goes 21:25
hence the "ranges"
AlexDaniel yup. As I see it, when you're on a bump commit, you can choose any version from the range 21:27
BooK ranges
AlexDaniel ranges?
BooK nqp range, moar range
AlexDaniel As I see it, on github.com/perl6/nqp/commit/f6459b...934619da6d we can build it for every moar revision from 2017.05-100-g5ea6aaa to 2017.05-115-g369c0c5 21:28
if some of it is broken that's alright
and on rakudo bumps we can do the same, so it's not ranges
well, it is, if you want to make life harder… :) 21:29
but the goal is to get some kind of bisect to work, and make life easier :)
BooK that is only for bump commits? 21:30
I mean, if you rakudo bisect lands you on a rakudo bump commit, you want to bisect on nqp/moar?
AlexDaniel I think we should bisect nqp first, then bisect moar if we find a bump in nqp 21:31
but yeah
timotimo AlexDaniel: i just pushed something to moarvm to help with the 0x7FFFFFFF problem, could you run it against the originally problematic module?
(didn't bump the moar and nqp versions though)
AlexDaniel timotimo: sure, give me three minutes
timotimo of course 21:32
something's still wrong when i try to .base(16) the magic number, though 21:33
oh, i could have fixed that, too
cool cool.
m: say (-0x7FFFFFFF - 1).base(16) 21:36
camelia -FFFFFFFF80000000
timotimo this should also give the right answer with latest moarvm and should potentially also get put into roast
AlexDaniel yeah, that does the trick 21:41
the golfed version is fine, and all module tests are passing
timotimo++ 21:42
timotimo \o/ 21:45
glad to hear it
AFK for a bit
will you do the release tomorrow, then?
or was there another distinct blocker? 21:46
AlexDaniel I'm not sure if it's a blocker 21:47
but there's no moarvm release anyway today, so yes, tomorrow
timotimo ah, of course
AlexDaniel the buffering issue is worth another look, yes 21:48
mainly because “prove” is acting weird with what we have now
AlexDaniel AFK
c: 2014.01 say (-0x7FFFFFFF - 1).base(16) 22:59
committable6 AlexDaniel, ¦2014.01: «-80000000»
AlexDaniel OK, now we're left with RT #132030 23:02
synopsebot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=132030
AlexDaniel samcv: so is everything alright with zef problem? 23:03
samcv not sure. i don't think it's a current release problem 23:05
so ignore that for now
MasterDuke samcv: btw, MVM_string_equal shows up pretty high in perf during stage parse. any ideas up your sleeve to speed it up? 23:06
samcv well i already sped it up :) 23:08
by 1.5-2x for nonstrands
err for strands i mean 23:09
and flat as well if the types don't match
so 1.5-2.2x it got sped up
except when we use memcmp when both are flat and both are the same storage type which is much of the time not true
MasterDuke hm, maybe i need to add some fprintfs to see which branches are being taken 23:10
Geth rakudo/nom: 1ca81432af | (Timo Paulssen)++ | src/core/Rakudo/Iterator.pm
refer to "want list?" var in permutations iterator properly

since classes aren't closures, you could get the wrong type from interleaved pull-one calls on iterators with $!b set to different values.		 23:12
samcv that would be helpful MasterDuke
Geth rakudo/nom: 488ca6f067 | (Timo Paulssen)++ | src/core/Rakudo/Iterator.pm
same fix for Combinations as Permutations

so we'll always get the correct type that was set on object creation time, not whatever was last passed to the Combinations method.		 23:13
timotimo MasterDuke: you should be able to hit "a" for annotate to see which parts of the code are hot in perf 23:14
MasterDuke oh interesting 23:15
samcv you'll be looking in MVM_string_substrings_equal_nocheck
which MVM_string_equal calls
timotimo perf doesn't call how often it gets called exactly, so maybe we want to callgrind the whole thing, too 23:16
samcv you could add an fprintf to each of the three branches of MVM_string_substrings_equal_nocheck as well 23:17
and then grep it and count the number of each that appear
MasterDuke, we're talking about compiling the core setting right MasterDuke ? 23:18
timotimo (of course running the core setting compilation under callgrind will take a while ...)
MasterDuke samcv: yeah
samcv ok let me check
i'll add some fprintf's and count
MasterDuke 45.29 │ testb $0x1,0xc(%rdx) 23:20
samcv what % is that function when compiling the core setting MasterDuke 23:21
like how much time does it use
cpu
42.2% branch_1, 25.2% branch_2, 32.5% branch_3 23:26
MasterDuke hm, a perf record only showed it at 1.65%, 23:31
samcv it probably was 2x as high a month ago
or more
MasterDuke a more accuate recording puts it at 2.56% 23:34
samcv i can speed it up though 23:35
the 1st branch is 42% and let's see how much i can speed it up
MasterDuke, how much do you se MVM_string_substrings_equal_nocheck at? 23:40
MasterDuke 0.00% 23:41
samcv it must be inlining then? 23:42
hm
though weirdly
2.85% for MVM_string_equal and 0.89 for MVM_string_substrings_equal_nocheck 23:43
timotimo i'd be interested to see a distribution of string lengths we're comparing
gfldex jnthn: are the last 2 paragraphs still correct? docs.perl6.org/language/variables 23:53
[Coke] AlexDaniel: I can do a spectest run with MSVC if you need it. 23:55
AlexDaniel [Coke]: you think it can give us something more than github.com/perl6/roast/issues/320 ? 23:56

Powered by Cro and the Raku® Programming Language.

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

Thank you!