lizmat Files=1233, Tests=75966, 326 wallclock secs (14.89 usr 5.26 sys + 2258.54 cusr 226.08 csys = 2504.77 CPU) 09:42
moritz: looking at the P6A schedule, isn't it tradition to only go up to 24, and not include 25 ? 09:43
moritz lizmat: I think we had a 25th once 09:46
lizmat well, last year we had 2 on the 24th :-)
moritz and I'd rather have a bonus day than reject an eager contributor
lizmat or I could take out mine and publish it as a separate blog post 09:48
moritz no, it sounds like a good post for the 24th 09:51
|Tux| Rakudo version 2017.11-120-g3405001da - MoarVM version 2017.11-43-g93a77dc16
csv-ip5xs1.058 - 1.074
csv-ip5xs-2012.621 - 12.683
csv-parser11.598 - 11.847
csv-test-xs-200.434 - 0.500
test11.270 - 11.306
test-t2.898 - 2.912
test-t --race1.124 - 1.206
test-t-2052.798 - 52.899
test-t-20 --race17.288 - 17.810
 10:02
tbrowder .tell AlexDaniel i’m discovering more pod issues i don’t believe are documented. i can put bandages on some of it in Pod.nqp, but grammar i think will have to be changed for other issues. i will keep plugging on triaging in Pod.nqp for now. 12:18
yoleaux tbrowder: I'll pass your message to AlexDaniel.
13 Dec 2017 22:07Z <jnthn> tbrowder: No, though there is a --rxtrace option that you can pass to the compiler
AlexDaniel . 12:19
yoleaux 12:18Z <tbrowder> AlexDaniel: i’m discovering more pod issues i don’t believe are documented. i can put bandages on some of it in Pod.nqp, but grammar i think will have to be changed for other issues. i will keep plugging on triaging in Pod.nqp for now.
DrForr Wonderful, so the data structure will change before I finish traversing it :) 12:21
AlexDaniel \o/
:)
tbrowder nothing should change for users, just some things are NYI or give unexpected results if used according to S26. 12:23
such things as the pod config continuation = don’t seem to work as i expect, and nested pod comment blocks in tables don’t work. also so more LTA msgs, etc. 12:28
DrForr Oh. I was planning to finish the rewrite of Pod::To::{what I'm now calling HTMLFragment} over the weekend. 12:33
Pod::TreeWalker uses events, which is great if you're simply walking trees, but if you need to know how the individual objects are linked it's a bit suboptimal. 12:34
tbrowder i doubt many users will find existing problems are show-stoppers for most uses. pod renderers may have to change later. and no deep changes will come from me any time soon. 12:35
DrForr The problem I was thinking about last night on the way home from Star Wars was this: 12:37
tbrowder DrForr: most of what i’ve been seeing is in tables, specifically config stuff, and comment handling.
DrForr =head1 foo # creates Pod::Block::Nested( [Pod::Heading( [Pod::Block::Para( ['foo'])])]) and I don't want the paragraph to show up in the HTML. 12:39
tbrowder: Good to know; I was going to rework table handling anyway.
If I'm dealing with an event-based parser, then I'm going to get open-Nested, open-heading, open-para, text 'foo', close-para, .. 12:40
I can certainly tell the text 'foo' code to look back in the tree, and if it's <h1><p>, then remove the traiing <p>, but the close-para is harder to deal with. 12:41
tbrowder first problem that would affect you in tables is caption. right now best practice is use form with <>, e.g., :caption<Some Title> and content of caption should be very usable. 12:42
DrForr Preprocessing the pod to remove the redundant <p/> layer is possible too, but I'd prefer to be able to let the user skip it, and event-based isn't going to let me do that without tacking on a .parent attribute, and that starts to verge upon circular. 12:43
tbrowder ergh...
DrForr tbrowder: Having captioned tables would be wonderful for LaTeX renderers, as there's a ready \longtable{} macro that supports it.
So what I thought I'd do is turn it into a list of {node -> Pod::Item.new(), parent => 23, sibling => 27} tuples as an array. 12:44
tbrowder the caption property is in master now, just use the <> form for pod source and caption content should be usable. 12:45
DrForr so conditions like that become 'if @pod[$node<parent>]<node> ~~ Pod::Heading { skip }' 12:46
One catch is that siblings require pregenerating the entire list. 12:47
(or at least until you find the next sibling of a node, or the closing event.) 12:48
tbrowder time for a coffee break...i’m not thinking html right now...clutters my mind...
i would love to see a native parent property on the pod nodes. 12:49
DrForr I can certainly generate it, and it's useful in the tree form. I do stuff in Perl6::Parser that spits out an array of tokens laced with pointers to next and parent fo rwalking. 12:50
tbrowder wouldn’t the parent property on Pod::Block be useful, ease your parser work? 12:52
DrForr It's also useful to link nodes w/o parents back to themselves so that you don't have to allow parents to be undef, and can statically type it.
tbrowder: yeah, it would, which is why I'm probably going to drop the Pod::TreeWalker module. I can generate my own events easily enough, and if I put the tree in an attribute then it's easier for other methods in the class to play with it. 12:53
Also, looking at the internas of Pod::TreeWalker it munges the tree on its own, and I'd prefer to let subclasses do the munging on their own, or at least be able to avoid it via subclassing. 12:57
I.E. in my case the simple HTML I'm generating doesn't need the <p/> of <h1><p>foo</p></h1>, but someone else that knows more HTML might want to generate <div>foo</div> for CSS reasons, and if I filter out the para, then I've made them do more work. 13:01
Geth rakudo: 0a2e5172d8 | (Elizabeth Mattijsen)++ | 3 files
Replace postfix ++/-- with prefix where possible

  - prefix is about 10% faster than postfix in nqp
  - not sure how this works out in rakudo itself
  - but every little bit helps
  - this may help unoptimizable code best		 14:35
Morfent huh 14:36
didn't know that wasn't just a c thing 14:37
lizmat Morfent: postfix needs to preserve the value of before the increment, prefix doesn't 14:38
that's in every language
Morfent ah 14:39
Zoffix Reminder: release is in 2 days and we still have 3 release blockers, if anyone can figure out how to fix them: github.com/rakudo/rakudo/issues?q=...2%9A%A0%22 14:46
lizmat ah? the last time I saw this, we had 0 so I'm unpleasantly surprised :-( 14:47
AlexDaniel` well, the last two were submitted 2 and 3 days ago 14:51
the problem with big endian systems is a bit more complicated
I didn't really manage to reproduce it in qemu
I'm assuming the submitter is able to reproduce it, but they can't bisect it :S
not sure how to make that thing move 14:52
also, today I'll run the toaster again, so maybe something will come out of it also 14:53
Zoffix++ for reminders (I had a really busy week so things were kinda slipping through the cracks) 14:54
lizmat FWIW, I'm seeing more flaky tests
lizmat runs a spectest with MVM_JIT_EXPR_DISABLE=1
AlexDaniel` will have a lot more free time starting this January, so hopefully more time will be put into the release process 14:58
DrForr AlexDaniel`: Talkie toaster? :) 15:04
Geth rakudo: 1fb5189567 | (Elizabeth Mattijsen)++ | 10 files
Replace postfix ++/-- with prefix where possible

  - but this time in HLL land		 15:38
TimToady hmm, coulda sworn we had the postfix-to-prefix optimization in already when we see sink context, but I don't see it offhand 17:50
AlexDaniel bisect: old=2017.11 gist.githubusercontent.com/AlexDan...5231312.p6 17:53
bisectable6 AlexDaniel, Successfully fetched the code from the provided URL.
AlexDaniel, Bisecting by exit code (old=2017.11 new=1fb5189). Old exit code: 0
TimToady and RAKUDO_OPTIMIZER_DEBUG seems to indicate the same 17:54
bisectable6 AlexDaniel, bisect log: gist.github.com/6180dcb75e204be85b...c64a23c8fb
AlexDaniel, (2017-11-25) github.com/rakudo/rakudo/commit/aa...51a5d32542
TimToady lizmat: you shouldn't have to change postfix to prefix in core/ if our optimizer was working right, but it doesn't seem to know that one 17:56
lizmat I'll gladly revert when it *is* handled by the optimizer :-) 18:00
AlexDaniel lizmat: speaking of reverts, what's up with github.com/rakudo/rakudo/issues/1312 ? 18:02
lizmat AlexDaniel: not sure... 18:03
AlexDaniel hmmm can't reproduce GH#1308 here :S 18:14
synopsebot GH#1308 [open]: github.com/rakudo/rakudo/issues/1308 [SEGV][severe][⚠ blocker ⚠] Frequent-ish SEQV in Perl 5 integration tests
lizmat AlexDaniel: segfaults almost always for me 18:16
s/almost//
AlexDaniel lizmat: within the stresstest or when run individually with fudgeandrun ? 18:17
lizmat perl6 t/spec/S01-perl-5-integration/basic.rakudo.moar 18:18
AlexDaniel hm… doesn't segfault here either way :S 18:21
lizmat perhaps it's an OS dependent thing, or a timing thing
Geth rakudo: 74a13ae15b | usev6++ | src/vm/jvm/runtime/org/perl6/rakudo/RakOps.java
[jvm] Check for null in p6definite

This mirrors the behaviour of nqp::isconcrete (which returns 0 or 1, though) and makes the recently added test for RT #126972 pass.		 18:51
synopsebot RT#126972 [resolved]: rt.perl.org/Ticket/Display.html?id=126972 [LTA] error mesage about getlexrelcaller (/a/) 18:52
Geth rakudo: d98ccea278 | TimToady++ | src/Perl6/Actions.nqp
optimize postfix incr to prefix when unwanted

Somehow we never managed to do this obvious optimization even after we got explicit knowledge of sink context from all the wanted/unwanted work.		 18:57
roast: 92cbeac338 | (Zoffix Znet)++ (committed using GitHub Web editor) | MISC/bug-coverage-stress.t
Fix failing stresstest

Fixes release blocker github.com/rakudo/rakudo/issues/1312
Implementation[^1] of proper Supply.list caused our test Supply be evaluated eagerly, but since `done` is not present, we hung.
Change `.list` call to `.Seq` to restore old behaviour in the test. I double-checked on 59bfa5ab37~10 that this change still covers the RT#132343 bug, while passing on HEAD.
  github.com/rakudo/rakudo/commit/aa...51a5d32542		 19:07
synopsebot RT#132343 [resolved]: rt.perl.org/Ticket/Display.html?id=132343 [REGRESSION] better-sched and other async improvement ecosystem fallout
AlexDaniel 🤗 Zoffix++ 19:12
raschipi lizmat: now you can revert your change 😄 19:13
AlexDaniel huggable: star 19:21
huggable AlexDaniel, Estimated Rakudo Star releases for 2017: .01, .04, .07 & .10
Geth rakudo: 370073fcda | (Elizabeth Mattijsen)++ | 10 files
Revert "Replace postfix ++/-- with prefix where possible"

This reverts commit 1fb5189567bc9255aec86b45593d21baf38fcee5.		 19:42
DrForr Pod::Table isn't broken down as far as it could be in Pod::TreeWalker. Another reason to bypass it.. I think I'm going to let it pass its last few tests, then write my own tree generator. 20:02
Not discrediting autarch's work at all, I'm sure life intruded; it happens to us all. 20:03
dogbert17 ZOFFLOP: t/spec/S32-io/indir.t 21:00
AlexDaniel OK toaster results are ready 23:06
timotimo awesome 23:07
AlexDaniel 8 modules unsucced but most of them are false positives. I'll investigate later after I get some sleep 23:08
releasable6: next
releasable6 AlexDaniel, Next release in ≈19 hours. No blockers. 0 out of 125 commits logged
AlexDaniel, Details: gist.github.com/d1d7a99c20f6dac771...68e90e5953
AlexDaniel github.com/rakudo/rakudo/issues?q=...2%9A%A0%22 23:09
releasable6 doesn't yet know about rakudo blockers because github.com/zoffixznet/r6/issues/11 23:10
Zoffix Geth: help 23:11
Geth roast: 92cbeac338 | (Zoffix Znet)++ (committed using GitHub Web editor) | MISC/bug-coverage-stress.t
Fix failing stresstest

Fixes release blocker github.com/rakudo/rakudo/issues/1312
Implementation[^1] of proper Supply.list caused our test Supply be evaluated eagerly, but since `done` is not present, we hung.
Change `.list` call to `.Seq` to restore old behaviour in the test. I double-checked on 59bfa5ab37~10 that this change still covers the RT#132343 bug, while passing on HEAD.
  github.com/rakudo/rakudo/commit/aa...51a5d32542
synopsebot RT#132343 [resolved]: rt.perl.org/Ticket/Display.html?id=132343 [REGRESSION] better-sched and other async improvement ecosystem fallout
Zoffix :/
Geth: help
Geth: ver github.com/rakudo/rakudo/commit/83...091f6119df
AlexDaniel drops to bed
Zoffix meow 23:12
Something's wrong with Geth. It don't see any messages
Geth: help 23:13
stupid robot
dogbert17: yo. Teach me to debug segfaults. 23:37
MasterDuke your MoarVM is built with --debug=3? 23:39
Zoffix Yeah 23:40
I guess `bt full` is it, but it don't tell me much. 23:41
MasterDuke then i just use valgrind and gdb, usually point somewhere relevant
Zoffix Though I've just bruteforced the offending commit manually :) 23:42
MasterDuke there's also --asan, but i've never had much luck with that
oh, and --valgrind can be helpful
Zoffix Thanks. 23:43
hm... this stuff might be above my head... 23:44
timotimo: you around?
OTOH, I'm gonna play with the code first. Maybe I'll come upon the fix by chance :} 23:45
MasterDuke heh, that's how i start most of my debugging 23:46
timotimo hi there, Zoffix 23:49
what's up?
oh, segfaults, that's fun
Zoffix :) I'm gonna try to fix it on my own first :) 23:50
timotimo OK

Powered by Cro and the Raku® Programming Language.

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

Thank you!