#perl6-dev on 8 November 2018 - Raku Programming Language Log

Perl 6 language and compiler development | Logs at colabti.org/irclogger/irclogger_log/perl6-dev | For toolchain/installation stuff see #perl6-toolchain | For MoarVM see #moarvm
Set by Zoffix on 27 July 2018.

00:16 MasterDuke joined, p6bannerbot sets mode: +v MasterDuke, MasterDuke left, MasterDuke joined, herbert.freenode.net sets mode: +v MasterDuke, p6bannerbot sets mode: +v MasterDuke 01:14 stmuk_ joined 01:15 p6bannerbot sets mode: +v stmuk_ 01:17 stmuk left 01:29 stmuk joined 01:30 p6bannerbot sets mode: +v stmuk 01:31 stmuk_ left 02:41 leont left 03:05 ufobat_ joined 03:06 p6bannerbot sets mode: +v ufobat_ 03:08 ufobat left 03:13 MasterDuke left 05:36 stmuk_ joined 05:37 p6bannerbot sets mode: +v stmuk_ 05:39 stmuk left 07:11 stmuk joined, p6bannerbot sets mode: +v stmuk 07:13 stmuk_ left 07:15 stmuk_ joined 07:16 p6bannerbot sets mode: +v stmuk_ 07:18 stmuk left 07:32 robertle joined 07:33 p6bannerbot sets mode: +v robertle 07:56 dct joined, p6bannerbot sets mode: +v dct 07:59 [Tux] left 08:02 dct left 08:16 [TuxCM] left 08:32 dct joined, p6bannerbot sets mode: +v dct 08:40 [Tux] joined, p6bannerbot sets mode: +v [Tux] 08:42 [Tux] left 08:43 [Tux] joined 08:44 p6bannerbot sets mode: +v [Tux]

lizmat

Files=1255, Tests=76372, 358 wallclock secs (15.86 usr 5.46 sys + 2524.06 cusr 235.99 csys = 2781.37 CPU)

09:52

this is significantly worse than a few days ago

I see a similar drop in my private test-t testing

09:53

masak

lizmat: hm, do we have any continuous benchmarking on commits?

09:57

lizmat

closest thing we have on that is test-t

10:01

and my irregular timing of spectest

afaik

10:04 AlexDaniel left 10:33 [TuxCM] joined 10:34 [TuxCM] left

Geth

rakudo: 0bc72e0504 | (Elizabeth Mattijsen)++ | 4 files
Add a Hyper class

  - inactive for now, re-configure of Rakudo is needed
  - contains the initial version of the HYPER sub refactor
  - the basic gain is that operator/dwim info doesn't need to be passed
  - also fixes use of object hashes / QuantHashes in infix hypers
  - should allow for easier optimizations in the future
... (6 more lines)

10:44

lizmat is now working to plug this into the core

10:45

[Tux]

Rakudo version 2018.10-91-gcbbe16edd - MoarVM version 2018.10-74-g2fdde4a21

csv-ip5xs	0.948	-	1.034
csv-ip5xs-20	7.408	-	7.409
csv-parser	23.139	-	23.516
csv-test-xs-20	0.435	-	0.440
test	8.510	-	8.533
test-t	1.881	-	1.898
test-t --race	0.970	-	0.980
test-t-20	31.512	-	33.648
test-t-20 --race	12.527	-	12.777

11:22

lizmat

yeah, most definitely slower :-(

11:30

nine

A slow down of this magnitude ought to be easy to pin down

11:45

lizmat

Type check failed in assignment to &!operator; expected Callable but got Sub+{is-nodal} (Sub+{is-nodal}.new)

isn't that, eh, weird ??? ^^^

timotimo

we've had that a few times, yeah

11:46

11:46 travis-ci joined, p6bannerbot sets mode: +v travis-ci

travis-ci

Rakudo build failed. Elizabeth Mattijsen 'Add a Hyper class

11:46

travis-ci.org/rakudo/rakudo/builds/452321829 github.com/rakudo/rakudo/compare/c...c72e0504ad

11:46 travis-ci left

lizmat

m: sub a() is nodal {}; dd &a ~~ Callable

11:46

camelia

Bool::True

lizmat

m: sub a() is nodal {}; dd &a

camelia

Sub+{is-nodal} a = sub a { #`(Sub+{is-nodal}|50924528) ... }

lizmat

m: sub a() is nodal {}; say &a

camelia

lizmat

m: sub a() is nodal {}; say &a.Str

11:47

camelia

Sub+{is-nodal} object coerced to string (please use .gist or .perl to do that)
a
in block <unit> at <tmp> line 1

lizmat

m: sub a() is nodal {}; say &a.^name

camelia

Sub+{is-nodal}

buggable

[travis build above] ☠ Did not recognize some failures. Check results manually.

11:49 ufobat_ left

timotimo

# Expected: all(/<<"No such private method '!bar'" \W/, /<<'RT123078_2'>>/, /<<'bar'>>/, /<<'baz'>>/)

11:50

# Got: No such private method '!!bar' for invocant of type 'RT123078_2'. Did you mean any of these?

how did that happen

the extra ! in there

lizmat

yeah, no idea:

11:59

m: class A { method foo { self!bar }; method bar { }; method baz { } }; A.new.foo

camelia

5===SORRY!5=== Error while compiling <tmp>
No such private method '!bar' for invocant of type 'A'. Did you mean any of these?
bar
baz

at <tmp>:1
------> 3class A { method foo { self!7⏏5bar }; method bar { }; method baz { …

lizmat is going to double check whether it was the Hyper addition

12:00

it does *not* appear to by my commit

12:03

(well, my last commit, that is)

wrt to slowdown, I'm adding 9457f7e540e2e959141f7 to the suspect list

12:06

and removes it again: that's compile time

Geth

rakudo: 75c9456a8b | (Elizabeth Mattijsen)++ | src/core/Hyper.pm6
Small tweaks to Hyper.pm6

  - cannot coerce Any to native int, so de-int $.dwim-left/right
  - forgot to change to instances of .do into .infix
   - the idea behind that is that we could handle other ops in the future
   as well with a different method

12:29

12:29 CIAvash left 12:36 evalable6 left

lizmat

so: with regards to the Travis breakage

12:36

the code in question is:

class RT123078_2 { method foo { self!bar }; method bar { }; method baz { } }; RT123078_2.new.foo

m: class RT123078_2 { method foo { self!bar }; method bar { }; method baz { } }; RT123078_2.new.foo

camelia

5===SORRY!5=== Error while compiling <tmp>
No such private method '!bar' for invocant of type 'RT123078_2'. Did you mean any of these?
bar
baz

at <tmp>:1
------> 3class RT123078_2 { method foo { self!7⏏5bar }; method bar …

lizmat

now, the odd thing is that this dies during compilation

12:37

timotimo

that's not odd

private methods are looked up during compilation

lizmat

indeed...

now, if I remove the actual call to the method from the test

12:38

the test fails because the code *doesn't die*

so the error that gets reported, is somehow a *runtime* error

timotimo

you mean not calling .new.foo will not make it break?

lizmat

indeed

timotimo

m: class RT123078_2 { method foo { self!bar }; method bar { }; method baz { } };

12:39

camelia

timotimo

not on camelia, apparently?

lizmat

well, and not separately either

12:40

but once it is inside t/05-messages/01-errors.t it doesn't

m: use Test; throws-like q| class RT123078_2 { method foo { self!bar }; method bar { }; method baz { } } |, X::Method::NotFound

camelia

1..2
ok 1 - ' class RT123078_2 { method foo { self!bar }; method bar { }; method baz { } } ' died
ok 2 - right exception type (X::Method::NotFound)
ok 1 - did we throws-like X::Method::NotFound?

lizmat

m: use Test; throws-like q| class RT123078_2 { method foo { self!bar }; method bar { }; method baz { } } |, X::Method::NotFound, message => all(/<<"No such private method '!bar'" \W/, /<<'RT123078_2'>>/, /<<'bar'>>/, /<<'baz'>>/)

12:41

camelia

1..3
ok 1 - ' class RT123078_2 { method foo { self!bar }; method bar { }; method baz { } } ' died
ok 2 - right exception type (X::Method::NotFound)
ok 3 - .message matches all(/<<"No such private method '!bar'" \W/, /<<'RT123078_…

lizmat

grrr

dogbert11

c: 2018.10 use Test; throws-like q| class RT123078_2 { method foo { self!bar }; method bar { }; method baz { } } |, X::Method::NotFound, message => all(/<<"No such private method '!bar'" \W/, /<<'RT123078_2'>>/, /<<'bar'>>/, /<<'baz'>>/)

12:55

committable6

dogbert11, gist.github.com/af30b4003fd3ce14d9...7a9b70ffaa

13:05 lucasb joined, p6bannerbot sets mode: +v lucasb 13:30 evalable6 joined 13:31 p6bannerbot sets mode: +v evalable6, travis-ci joined, p6bannerbot sets mode: +v travis-ci

travis-ci

Rakudo build failed. Elizabeth Mattijsen 'Small tweaks to Hyper.pm6

13:31

travis-ci.org/rakudo/rakudo/builds/452359841 github.com/rakudo/rakudo/compare/0...c9456a8b95

13:31 travis-ci left

lizmat

yeah, still bisecting ^^^

13:32

buggable

[travis build above] ☠ Did not recognize some failures. Check results manually.

lizmat

looks like it is 8a6c38bfbcb66f6b6169bc that broke it

13:36

reproduced:

13:37

$ perl6 --optimize=0 -e 'class A { method foo { self!bar }; method bar { }; method baz { } }; A.new.foo'

No such private method '!!bar' for invocant of type 'A'. Did you mean any of these?

looks like that change is causing all code to be un-optimized

13:38

at least, not have a static optimization step

Geth

rakudo: f04c6c34b8 | (Elizabeth Mattijsen)++ | src/core/ForeignCode.pm6
Revert "Make EVAL options when we didn't even read options from the command line"

This reverts commit 8a6c38bfbcb66f6b6169bc4f01285692880791b0.
It looks like this causes the static optimizer step to be ignored.