#perl6-dev on 27 April 2018 - Raku Programming Language Log

MasterDuke

b2gills: none here with master/master/master on Arch Linux. what OS, etc are you using?

00:09

b2gills

I wrote a shell script which uses NQP_REVISION and MOAR_REVISION, so it should be the latest pull

00:11

s/pull/bump/

00:19

Ubuntu 16.04.4 LTS

00:20

MasterDuke

hm, i think i once found a problem with an older gcc. what version of that do you have?

00:38

also, i know nothing about nativecall, so i'd say just make an issue and tag nine in it

00:40

b2gills

I just blew away /opt/rakudo … and I only have `moar` `nqp` & `nqp-m` in /opt/rakudo/bin.

00:55

So I was apparently using an old revision

Makefile:568: recipe for target 'm-coretest5' failed

00:58

Makefile:620: recipe for target 'm-install' failed

gcc version 5.4.0 20160609 (Ubuntu 5.4.0-6ubuntu1~16.04.9)

00:59

GNU Make 4.1

01:00

Kaiepi

what is Block in nqp?

01:10

timotimo

Block, like a class in the hierarchy of Routine and such?

01:11

Kaiepi

i'm guessing so

01:12

i can't figure out how to use stuff like nqp::for

timotimo

you're writing actual nqp code for the nqp compiler to consume?

Kaiepi

though i probably don't really need to use it

no, that's why i say that

also when do i need to use nqp::decont?

01:15

timotimo

whenever you get a container in and you want to do stuff with the value in it instead, which is almost always

01:16

in perl6 code, you get deconts code-genned everywhere, and moarvm is good at removing them when they are unnecessary

01:21

because deconting something that isn't a container is a no-op

Kaiepi

how do you tell if something has a container?

01:22

timotimo

nqp::iscont :)

01:23

m: use nqp; say nqp::iscont(my $foo); say nqp::iscont(my \blop = 1)

camelia

1
0

Kaiepi

ohh

so sigils give vars containers?

01:24

timotimo

yup

Kaiepi

what about self in class methods?

timotimo

iscont is also what different parts of the code decide whether to iterate something or not

not sure, you can try it out :)

01:25

MasterDuke

m: use nqp; say nqp::iscont(my $foo := 1);

camelia

timotimo

right, binding replaces containers

assigning goes into containers

01:26

Kaiepi

ohh

01:27

ok now it makes more sense

thanks

b2gills

make install fails with `./perl6-m tools/build/install-core-dist.pl /opt/rakudo/share/perl6␤Too few positionals passed; expected 2 arguments but got 1␤ in block <unit> at tools/build/install-core-dist.pl line 17`

01:37

Geth

roast: 432a5531bb | MasterDuke17++ | S02-types/array.t
Test reification (or not) of zen/whatever slices

01:40

Zoffix

Kaiepi: self is always deconted. One exception (both to self and what iscont gives) is Proxy containers which remain conted (since deconting them would just fetch the value and discard the likely-important Proxy)

01:54

Kaiepi

Zoffix

huggable: HLL iscont

huggable

Zoffix, say $v.VAR ~~ Scalar|Proxy ?? "conted" !! "deconted"

Zoffix

^ the non-NQP way to check if something's is conted

huggable: decont

huggable

Zoffix, Article on containers and decont: perl6advent.wordpress.com/2017/12/...oneandonly

Zoffix

^ decent article on containers

"./libmoar.so: undefined reference to `MVM_getrandom'" :S

02:22

timotimo

need to Configure.pl again

Zoffix

That's on my 32-bit box

Ah. Thanks

Geth

nqp/nummy-num: b46ca01ca1 | (Zoffix Znet)++ | t/nqp/115-nums.t
Use original string num in test desc

For easier debugging of failures

02:39

nqp/nummy-num: eb50ac8e7e | (Zoffix Znet)++ | t/nqp/115-nums.t
Test we can handle fancy unicode nums

02:44

nqp: zoffixznet++ created pull request #438:
Nummy num

03:05

nqp/master: 4 commits pushed by (Zoffix Znet)++

Zoffix

uhhh.. ."Could not execute (./perl6-m t/spec/S10-packages/precompilation.rakudo.moar): open3: exec of ./perl6-m t/spec/S10-packages/precompilation.rakudo.moar failed: No such file or directory at /home/cpan/perl5/perlbrew/perls/perl-5.24.0/lib/5.24.0/TAP/Parser/Iterator/Process.pm line 165."

03:09

oh, nm, I think it was a residual process hanging about after I killed z-script initiated stresstest

03:11

s/think/fairly sure/;

t/02-rakudo/03-cmp-ok.t (Wstat: 512 Tests: 6 Failed: 2)

03:14

Failed tests: 1, 6

Non-zero exit status: 2

t/08-performance/02-qast-rewrites.t (Wstat: 1024 Tests: 5 Failed: 4)

Failed tests: 1-3, 5

Non-zero exit status: 4

t/02-rakudo/repl.t (Wstat: 256 Tests: 42 Failed: 1)

Failed test: 35

Non-zero exit status: 1

sparkling clean the second time... Weird that such a burst of sporadic failures occured :/

03:17

travis-ci

NQP build failed. Zoffix Znet 'Merge pull request #438 from perl6/nummy-num

03:25

travis-ci.org/perl6/nqp/builds/371853211 github.com/perl6/nqp/compare/d0b1b...dae9460c94

Zoffix blows everything out and re-tries the bumps

ZOFFLOP: t/spec/S11-modules/require.t

03:38

ZOFFLOP: t/spec/S07-hyperrace/basics.t t/spec/S10-packages/basic.t

03:42

Geth

nqp: aec1b1fd7d | (Zoffix Znet)++ | tools/build/MOAR_REVISION
[MoarVM Bump] Brings 13 commits

MoarVM bump brought: github.com/MoarVM/MoarVM/compare/2...2-g517e26f 517e26f Merge pull request #850 from MoarVM/nummy-num c6c0bab Fix denormals and precision issues in num parser c6aba84 don't use %b in format string 7349b42 have to use FSA consistently for this field ... (9 more lines)

¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...2-g517e26f

rakudo: 3d6e25c680 | (Zoffix Znet)++ | tools/build/NQP_REVISION
[NQP Bump] Brings 5 commits

NQP bump brought: github.com/perl6/nqp/compare/2018....9-gaec1b1f aec1b1f [MoarVM Bump] Brings 13 commits 26dae94 Merge pull request #438 from perl6/nummy-num eb50ac8 Test we can handle fancy unicode nums b46ca01 Use original string num in test desc ... (16 more lines)

¦ rakudo: version bump brought these changes: github.com/perl6/nqp/compare/2018....9-gaec1b1f

Zoffix

.tell AlexDaniel` the nqp num issues are fully fixed now. I did check on my 32-bit box and both the original failing tests and then new NQP num tests were fine.

04:05

yoleaux

Zoffix: I'll pass your message to AlexDaniel`.

stmuk

Zoffix++

06:56

[Tux]

Rakudo version 2018.04-17-g3d6e25c68 - MoarVM version 2018.04-32-g517e26f04

csv-ip5xs	0.885	-	0.906
csv-ip5xs-20	8.722	-	9.094
csv-parser	37.105	-	38.971
csv-test-xs-20	0.423	-	0.447
test	9.675	-	9.739
test-t	2.507	-	2.541
test-t --race	1.012	-	1.047
test-t-20	44.698	-	45.365
test-t-20 --race	15.996	-	16.352

09:56

AlexDaniel

10:36

AlexDaniel`

yoleaux

01:50Z <Zoffix> AlexDaniel`: gist.github.com/zoffixznet/3c6e026...0be2acaaf6

04:05Z <Zoffix> AlexDaniel`: the nqp num issues are fully fixed now. I did check on my 32-bit box and both the original failing tests and then new NQP num tests were fine.

dogbert2_

No such method 'install' for invocant of type 'Str'

10:40

in block <unit> at tools/build/install-core-dist.pl line 28

wazzup with this?

AlexDaniel

Zoffix: but you still think that we should just fudge these tests?

or “fully fixed” means fully fixed and we can ship it?

10:41

jnthn

dogbert2_: Thing that was a bug introduced by a commit in Rakudo that was already reverted

*Think

10:46

Zoffix

AlexDaniel: en.wikipedia.org/wiki/Dunning%E2%8...ger_effect

10:52

I don't know. It works and passes stresstests on my 64bit and 32bit boxes.

10:53

But I don't know C so for all I know I wrote a bunch of crap there and it'll explode the second you release a point release :P

Geth

nqp: 4d8503982c | (Tobias Leich)++ | tools/build/MOAR_REVISION
Bump moar for inlined fixed sized CArray support

10:54

¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...4-g25f165a

Zoffix

new num nqp tests pass on win10 too

10:55

Geth

rakudo: dbc527d46a | (Tobias Leich)++ | tools/build/NQP_REVISION
Bump nqp/moar for inlined fixed sized CArray support

10:56

rakudo: version bump brought these changes: github.com/perl6/nqp/compare/2018....0-g4d85039
4c3807c5fb | (Tobias Leich)++ | 4 files

The shape declaration on an attributes was parsed correctly since a while, but now this information is passed to the repr composition. The following is now supported fo CStructs and CPPStructs:
use NativeCall; ... (6 more lines)

rakudo: 7847768cad | (Tobias Leich)++ | src/Perl6/Actions.nqp
Mark the CArray dimension as nut sunk

FROGGS

I hope I didn't break anything

10:58

Zoffix

AlexDaniel: also, I'm fairly sure the test failures are not a new bug but rather the old bug (failure to choose closest representable num) exposed because our stringification got more precision. Also, I'm not aware of that bug affecting anything on rakudo level.

10:59

AlexDaniel: so yeah, I think the test should be fudged for release and the fix left out of it.

AlexDaniel

alright… that kinda makes sense, but then I barely woke up

11:00

Zoffix

But other than that yeah, by “fully fixed” I mean the bug is fixed and is covered by tests.

11:03

dogbert2_

jnthn: thx

11:09

travis-ci

NQP build failed. Tobias Leich 'Bump moar for inlined fixed sized CArray support'

11:14

travis-ci.org/perl6/nqp/builds/371974446 github.com/perl6/nqp/compare/aec1b...8503982c1c

FROGGS

I guess the just reported failures of t/nqp/106-unicodenames.t and t/nqp/115-nums.t are to be expected on jvm?

11:19

because, what I did can surely have no impact on that

11:20

AlexDaniel

that's correct, yes

11:21

but what's exactly failing in 115-nums.t? :)

Zoffix guesses value drift

11:22

FWIW 115-nums.t is a new test file I added last night that covers denormals and value drift in parsing/stringification of nums

FROGGS

I see

11:23

Zoffix

.tell pmurias I think package-lock.json is something with JS backend innit? Would you know how to fix this Issue? github.com/perl6/nqp/issues/437

11:56

yoleaux

Zoffix: I'll pass your message to pmurias.

Zoffix

Oh, k, "Only users who have been granted access to vulnerability alerts for this repository can see this message."

11:57

Oh, I was wrong, it fails 'cause "not ok 7 - denormal 5e-324 is recognized and is not 0 (Uni)"

12:11

AlexDaniel

I guess it's just a matter of using a newer version of that package

Zoffix

j: say ೫e-೩೨೪

camelia

5===SORRY!5=== Error while compiling <tmp>
'೫' is not a valid number
at <tmp>:1
------> 3say ೫7⏏5e-೩೨೪

Zoffix

m: say ೫e-೩೨೪

camelia

5e-324

Zoffix

j: say ೫

12:12

camelia

5===SORRY!5=== Error while compiling <tmp>
'೫' is not a valid number
at <tmp>:1
------> 3say ೫7⏏5<EOL>

AlexDaniel

or maybe not, because there's no point release for that security issue…

12:13

and you'd have to upgrade to a newer major version

Geth

nqp: d36869cc27 | (Zoffix Znet)++ | 2 files
Fudge newly-added Unicodey tests on JVM

12:22

travis-ci

NQP build passed. Zoffix Znet 'Fudge newly-added Unicodey tests on JVM'

12:40

travis-ci.org/perl6/nqp/builds/372006759 github.com/perl6/nqp/compare/4d850...6869cc2742

dogbert2_

ZOFFLOP: t/spec/S11-modules/require.t

13:03

when I run the above script separately I sometimes get the following:

13:07

ok 13 - can load Fancy::Utilities at run time

13:08

Use of uninitialized value $repo-id of type Any in string context.

Methods .^name, .perl, .gist, or .say can be used to stringify it to something meaningful.

in any statement_control at blib/Perl6/Grammar.moarvm line 1

if I run it again the message above is gone !

brrt

dear #perl6-dev. i would very much appreciate it if someone (who is on linux, or maybe macOS), can checkout the branch `jit-stack-walker` on MoarVM and run a benchmark against master

13:22

for me, spectest passes

13:24

b2gills

So my weird problems seem to have disappeared after `git clean -dxf` and removing every remnant from previous builds I could think of (/opt/rakudo/*;~/.perl6/precomp)

14:15

lizmat

Files=1239, Tests=76311, 318 wallclock secs (15.71 usr 5.37 sys + 2198.79 cusr 220.31 csys = 2440.18 CPU)

16:01

AlexDaniel

Alright, so to me it makes sense just to fudge tests that are not passing on i386 for the point release. This way there's no need to hunt for potential new bugs and no need for moarvm point release. Especially so given that we've not seen any user-facing issues, and that the underlying bug probably existed before

16:17

if anyone disagrees, please speak up :)

and btw, from now on we'll notice this kind of issues on i386 before the release, thanks to El_Che++

16:22

AlexDaniel deletes all branches that have no commits ahead of master…

16:25

Zoffix

16:27

AlexDaniel

to be honest I feel like deleting all branches that are 10k+ commits behind

16:29

Zoffix

ZOFFLOP: t/spec/S12-class/literal.t

16:32

ZOFFLOP: t/spec/S10-packages/basic.t

16:35

AlexDaniel

Alright, this month I'll bother all of you about these stale branches

El_Che

AlexDaniel: do the canary work as intended?

16:37

does?

AlexDaniel

El_Che: no, but we will see when I'll get the point release preready

16:38

El_Che: I mean, I don't know yet :)

I've been trying to cut a virtual release from master but failed because of issues with File::Which

El_Che

AlexDaniel: we know it passes regular releases until 2018.03, so everything failing after that... :)

16:39

AlexDaniel

El_Che: but 2018.04 is failing only on i386 right?

16:40

El_Che

yes

AlexDaniel

so I guess it works then

anyway we'll see tonight

El_Che

looking forward to the dot releases. A lot of changes to my pkg repo wainting on a branch: repos for debs and rpms, support for Ubuntu 18.04 amd64/i386, script to set PATH for newbies, support for Linux on Windows for all packages exect alpine (it does not work there yet) instead of only Ubuntu

16:42

now I am reading it, it sounds like a merge nightmare, but travis passes :)