#perl6-dev on 15 January 2019 - 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.

03:04 ufobat_ joined, p6bannerbot sets mode: +v ufobat_ 03:07 ufobat left, lizmat left 06:16 robertle left

nine

.tell patrickb Though I don't remember the thoughts behind RAKUDO_PREFIX, it seems to be a documented interface and I'd guess one used in some packaging or build scripts. I'd rather not break those without good reason.

06:27

yoleaux

nine: I'll pass your message to patrickb.

07:33 brrt joined 07:34 p6bannerbot sets mode: +v brrt 07:57 robertle joined, p6bannerbot sets mode: +v robertle 08:16 lizmat joined, p6bannerbot sets mode: +v lizmat

lizmat

Files=1265, Tests=88017, 379 wallclock secs (20.26 usr 6.21 sys + 2699.20 cusr 231.20 csys = 2956.87 CPU)

08:18

09:28 ExtraCrispy joined, p6bannerbot sets mode: +v ExtraCrispy

|Tux|

Rakudo version 2018.12-217-g54fe8a1a8 - MoarVM version 2018.12-29-g1ff55bf1c

csv-ip5xs	0.740	-	0.755
csv-ip5xs-20	6.548	-	6.671
csv-parser	22.174	-	22.968
csv-test-xs-20	0.426	-	0.437
test	7.661	-	7.984
test-t	1.842	-	1.877
test-t --race	0.837	-	0.876
test-t-20	31.219	-	31.932
test-t-20 --race	9.635	-	9.808

11:16

Geth

rakudo: 2ed2f39813 | (Elizabeth Mattijsen)++ | src/Perl6/Actions.nqp
Allow for my %h is Set[Int] = ... to work

More specifically, any "is" trait with a type and a parameterization will now perform the parameterization if the .HOW of the type has a "parameterize" method.

11:37

11:38 MasterDuke left

Geth

rakudo: eb50d8cefc | (Elizabeth Mattijsen)++ | src/core/Parameter.pm6
Update comment

12:04

12:27 brrt` joined, p6bannerbot sets mode: +v brrt` 12:29 brrt left 12:33 brrt`` joined 12:34 p6bannerbot sets mode: +v brrt`` 12:35 brrt` left

dogbert17

lizmat: should we be worried about t/spec/S14-traits/attributes.t ?

12:41

lizmat

interesting... I assumed that it was a flapper :-(

12:42

dogbert17

unfurtunately it's fails very consistently

12:43

lizmat

yeah, I see that now :-(

12:46

lizmat investigate

12:49

s

dogbert17

perhaps it's an easy fix

lizmat

looking at the diff, I don't see how I changed the code path for "is doc('barks')"

12:50

timotimo

maybe the debugserver could help by putting a breakpoint in the relevant places and seeing what triggers them

12:51

13:18 brrt`` is now known as brrt

lizmat

dogbert17: yeah, fix is easy, although one could argue that the test is iffy

13:19

because "doc" is actually a type in that file, "is doc" without any parameter does not do what you expect

13:20

even if you make the second argument on the trait_mod optional

"is Type" is just handled specifically and especially

13:23

Geth

rakudo: 8ad750ab57 | (Elizabeth Mattijsen)++ | src/Perl6/Actions.nqp
Limit is Type[OtherType] parameterization to QuantHashes

Unbreaks S14-traits/attributes.t , dogbert17++ for spotting

13:28

rakudo: 6a8f08b95f | (Elizabeth Mattijsen)++ | src/core/Rakudo/QuantHash.pm6
Show the actual value instead of its type

Improves the error message for "Set[Int}.new = <a b c>" from:
Type check failed in binding; expected Int but got Str (Str)
to:
Type check failed in binding; expected Int but got Str ("a")

14:09

releasable6

Next release in ≈4 days and ≈3 hours. 6 blockers. Please log your changes in the ChangeLog: github.com/rakudo/rakudo/wiki/ChangeLog-Draft

15:00

lizmat

afk for a bit&

jnthn

Every time I look the blockers count goes up :P

15:01

15:05 brrt left

nine

So why do you keep looking? :P

15:08

jnthn

Good question :P

15:10

Well, now I have a bit of time to see if I can help with any of 'em... :)

15:19

15:23 patrickb joined, p6bannerbot sets mode: +v patrickb

patrickb

o/

15:24

yoleaux

06:27Z <nine> patrickb: Though I don't remember the thoughts behind RAKUDO_PREFIX, it seems to be a documented interface and I'd guess one used in some packaging or build scripts. I'd rather not break those without good reason.

jnthn

m: for 1..3 { { say $_ ; last } if $_ == 2 }

camelia

True

jnthn

committable6: 2018.12 for 1..3 { { say $_ ; last } if $_ == 2 }

committable6

jnthn, ¦2018.12: «2␤»

jnthn

Huh, this one is broken even with --optimize=off

15:25

(On HEAD)

patrickb

Google just opened registration for summer of code mentoring organisations!

Or is at least bound to do so today...

15:28

I'd very much like perl6 (or tpf or whatever makes most sense) to take part again.

15:30

I'll see what I can do (even if it's only buggering the right people).

15:32

jnthn

In github.com/rakudo/rakudo/issues/2601 are we really sure that 541a4f16 is the one to blame...

15:33

committable6: 541a4f16 for 1..3 { { say $_ ; last } if $_ == 2 }

committable6

jnthn, ¦541a4f1: «True␤»

jnthn

committable6: 541a4f16~1 for 1..3 { { say $_ ; last } if $_ == 2 }

committable6

jnthn, ¦541a4f16~1: «2␤»

jnthn

How on earth can a commit that only touches Optimizer.nqp have an effect if --optimize=off

15:34

committable6: 541a4f16~1 --optimize=off for 1..3 { { say $_ ; last } if $_ == 2 }

15:35

committable6

jnthn, gist.github.com/49ccc7dfdebae3ff92...91ddb61ff3

jnthn

AlexDaniel: Is there a way to do ^^ (that is, run a commit with a command line option passed to Rakudo)?

I'm wondering if the optimizer hid a problem introduced in an earlier commit, then got fixed so it didn't hide the bug

15:36

AlexDaniel

committable6: 541a4f16~1 run <perl6 --optimize=off -e>, ｢for 1..3 { { say $_ ; last } if $_ == 2 }｣

committable6

AlexDaniel, ¦541a4f16~1: «True␤»

AlexDaniel

is this good enough?

jnthn

Yes, and confirms the hypothesis...now, can I bisset...

bisectable6: run <perl6 --optimize=off -e>, ｢for 1..3 { { say $_ ; last } if $_ == 2 }｣

15:37

bisectable6

jnthn, Bisecting by output (old=2015.12 new=6a8f08b) because on both starting points the exit code is 0

jnthn, bisect log: gist.github.com/e8f56bde082cb4bddc...ab63f9854e

jnthn, (2018-02-10) github.com/rakudo/rakudo/commit/df...671e512898

jnthn

Wow!

AlexDaniel

c: dfb6d951d7e^,dfb6d951d7e run <perl6 --optimize=off -e>, ｢for 1..3 { { say $_ ; last } if $_ == 2 }｣

15:38

committable6

AlexDaniel, ¦dfb6d951d7e^: «2␤» ¦dfb6d95: «True␤»

jnthn

committable6: 2018.12 run <perl6 --optimize=off -e>, ｢for 1..3 { { say $_ ; last } if $_ == 2 }｣

committable6

jnthn, ¦2018.12: «True␤»

jnthn

Gosh.

So what's happening is that `{ say $_ ; last } if $_ == 2` is getting compiled as if it was `if $_ == 2 { say $_ ; last }`

15:39

But it appears it has been for a while, but an optimizer flattening bug covered it up

15:40

committable6: 2018.12 for 1..3 { { say $_ ; last } if $_ == 2 }

committable6

jnthn, ¦2018.12: «2␤»

jnthn

committable6: 2018.12 for 1..3 { { my $x = 42; say $_ ; last } if $_ == 2 }

committable6

jnthn, ¦2018.12: «True␤»

jnthn

Yup; that declaration of $x was enough to prevent the old flattener from doing its work.

15:41

And probably the commit bissectable just found us was when code-gen changed so as to start respecting .count or something on a QAST::Block

15:42

And the optimizer hid the regression, and nobody wrote something complex enough in the block on the LHS to unveil the bug

15:43

AlexDaniel

heh

jnthn

So finally, I correct the block flattener to (correctly, given the AST it's looking at) not flatten that inner one because it's a block taking an arg, and...suddenly it appears. Serves me right for fixing things :P

AlexDaniel

haha :D

15:44

jnthn

So now the interesting question is how to fix it properly. :)

AlexDaniel

jnthn: it's a new button, btw

15:45

jnthn

Yeah, it's a really nice idea to have it

AlexDaniel

it appeared around the same time they introduced “Transfer issue” feature

I think

jnthn spectests a fix

15:59

15:59 cfa joined, p6bannerbot sets mode: +v cfa

cfa

👋🏽

16:00

jnthn

m: use Test; my $res; for 1..3 { { $res = $_; last } if $_ == 2; }; is $res, 2

camelia

not ok 1 -
# Failed test at <tmp> line 1
# expected: '2'
# got: 'True'

AlexDaniel relocates to another room…

jnthn

m: use Test; my $res; for 1..3 { { $res = $_; last } unless $_ != 2; }; is $res, 2

camelia

not ok 1 -
# Failed test at <tmp> line 1
# expected: '2'
# got: 'False'

AlexDaniel

neighbour above is using their rotary hammer… to drill holes in the floor??

right on top of my head…

jnthn

!!

That sounds headache-inducing...

16:01

dogbert17

AlexDaniel: are they spying on you :)

16:02

Geth

rakudo: c0f8378827 | (Jonathan Worthington)++ | src/Perl6/Actions.nqp
Fix handling of $_ in blocks to if/unless mods

The program:
for 1..3 { { say $_; last } if $_ == 2 }
Recently started to output `True` instead of `2`. This bissected to the ... (11 more lines)

16:20

roast: 9cdec2fd9c | (Jonathan Worthington)++ | 2 files
Test $_ behavior in block with if/unless mods

Covers the regression reported in
github.com/rakudo/rakudo/issues/2601.

16:21

jnthn

That's one half of #2601

16:23

cfa

jnthn: wow, that sounds a bit gnarly

16:28

nice one

patrickb relocates

16:31

16:31 patrickb left 16:36 lucasb joined, p6bannerbot sets mode: +v lucasb

jnthn

m: use Test; { $_ = 42; is (do if 42 { $_ }), 42; }

16:42

camelia

not ok 1 -
# Failed test at <tmp> line 1
# expected: '42'
# got: (Rakudo::Internals::LoweredAwayLexical)

16:45 brrt joined, p6bannerbot sets mode: +v brrt

Geth

rakudo: 623450d805 | (Jonathan Worthington)++ | src/Perl6/Optimizer.nqp
Remove an unused parameter

16:49

rakudo: 08d87f7061 | (Jonathan Worthington)++ | src/Perl6/Optimizer.nqp
Do not copy $_ upon immediate block inlining

Otherwise, in some circumstances, it can lead to lexical accesses that were meant to read outer $_ reading the lowered stub one. This addresses the second issue that was reported in #2601.

16:51

roast: 7c08f8f29a | (Jonathan Worthington)++ | S04-statements/do.t
Cover `do if 42 { $_ }` evaluating correctly

Covers a bug reported in github.com/rakudo/rakudo/issues/2601.

16:54

16:55 ilmari_ left 17:00 lizmat left 17:03 robertle left 17:04 patrickb joined 17:05 p6bannerbot sets mode: +v patrickb

AlexDaniel

jnthn++

17:05

17:07 patrickz joined, lizmat joined, p6bannerbot sets mode: +v lizmat, p6bannerbot sets mode: +v patrickz 17:10 patrickb left 17:24 dct joined, p6bannerbot sets mode: +v dct 18:07 robertle joined 18:08 p6bannerbot sets mode: +v robertle 18:16 dct left 18:24 brrt left 19:11 dct joined, p6bannerbot sets mode: +v dct 20:05 lucasb left 20:20 lizmat left 20:25 lizmat joined, p6bannerbot sets mode: +v lizmat

Geth

rakudo: bdc63f0a8d | (Elizabeth Mattijsen)++ | src/core/Rat.pm6
Revert "Introducing $*FULL-PRECISION"

This reverts commit 9f956aff4a9c0e90c223f18bbaa63603bf859d65.
Jonathan need to mull over this some more, so carry this across the release.

20:27

20:56 lizmat left 21:03 lizmat joined, p6bannerbot sets mode: +v lizmat 21:41 entonian joined 21:42 p6bannerbot sets mode: +v entonian 21:47 entonian left

Geth

roast: 73e63d5aba | (Elizabeth Mattijsen)++ | 3 files
Add basic parameterized Set/Bag/Mix tests

21:50

22:05 ufobat_ left, patrickz left

Geth

roast: 7df0786dec | (Elizabeth Mattijsen)++ | 3 files
Unskip tests for SetHash/BagHash/MixHash

Also correct some tests as they were originally written at a time when the semantics weren't as clear as they are now.

22:26

rakudo: 87b71c2758 | (Elizabeth Mattijsen)++ | src/core/MixHash.pm6
Oops, forgot to check on types in MixHash.AT-KEY

Found thanks to the power of tests!

22:45

roast: 2df8f54493 | (Elizabeth Mattijsen)++ | 3 files
Add basic parameterized SetHash/BagHash/MixHash tests

22:46

lizmat

and that concludes my hacking for today&

22:51 cfa left 22:52 dct left

lizmat just spotted a new error in S32-basics/xxKEY.t, will look at this tomorrow

22:57

afk&

Powered by Cro and the Raku® Programming Language.

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

Thank you!