"Tuesday at 20:30 UTC"
Set by moderator on 25 June 2010.
02:06 ash_ joined 04:20 ash_ joined 06:34 eternaleye joined 06:48 eternaleye joined 09:39 moritz joined
moritz What I've done (parrot-ish stuff): 09:39
* Worked with tcurtis++ on getting some static analysis into 09:40
Rakudo, using the new optimization framework.
I have to praise tcurtis for being responsive, helpful
and all in all very pleasant to work with!
* tried to write a FakeDBD::pg driver using Pg.pir
* Exercised Pg.pir that way: tt#1692, tt#1695, tt#1696
Most of those shold be low-hanging fruits.
If somebody fixes them soon, our chances for shipping
postgres database support with Rakudo Star are pretty good.
(I'll probably not make it to today's meeting)
EOR
10:46 mikehh joined 12:57 bluescreen joined 18:44 tcurtis joined 19:15 darbelo joined
darbelo DONE 19:20
- Blog post.
- Plumage resurrection.
- NFG iterators and literals.
- NFG freeze and thaw. Not finished yet.
- Hate packfiles.
- Got architectural input.
- Hate packfiles.
TODO
19:20 NotFound joined
darbelo - Hate packfiles. 19:20
19:20 Chandon joined
darbelo - Finish freeze and thaw. 19:20
- Hate packfiles.
- More transcoding work.
- Hate packfiles.
EOR.
cotto_work #did: 19:32
- met with khairul to discuss his gsoc project
- he's a little behind but not so badly that his slack weeks won't help
- hacked on PIRATE
- implemented keys and strings (mostly)
- started reading and understanding the linear scan register allocator
- started a couple deprecation pages (ParrotDeprecations and HowToDeprecate (hijacked from dukeleto++))
- got helpful feedback from Austin++ and others
- OSUOSL has a trac/git test site up at trac.parrot.org/parrottest/browser
- it works but is really slow 19:33
- my preference would be to host on github with the trac/github plugin
- didn't have tuits for writing docs about GPG signing of releases
#will do:
- implement linear scan register allocator in PIRATE
- deprecation policy work
- tuits likely to be limited
#eor
qq
q2q
19:34 darbelo_ joined
NotFound What I did: 19:37
-parrot
* Added more PMC tests 19:38
* Implemented isatty method in Handle, deprecated is_tty
* Miscellaneous fixes and cleaning
* Allow usage of function-like objects in Parrot_pcc_invoke_from_sig_object
-winxed
* Namespace scoped 'using extern' meaning '.loadlib'
* Automatic loading of math ops when math predefs are used
* Changed handling of +, - and * when both operands are var
* Updated and improved some examples
What I will do:
No plan
EOR
Coke Done: 19:39
- attended YAPC, including a mini BOF about rakudo's pain^Wuncomfy
points. Kid51 posted the minutes from that, but chromatic or I
should probably do a writeup.
Partcl:
- tricked Matt Diephouse into committing code to partcl-nqp!
- Matt then tricked me back. LOTS of improvements to partcl this week.
Will Do:
- Releasing next versions of parrot /and/ Rakudo compiler.
- for parrot, want to make sure HTML docs are fixed. About 40% of the
way there. If you want to improve docs, concentrate on adding POD,
while I hack on the html converter borkage.
- Need to review experimental items before the release. let's plan on
next week.
- Need to review existing deprecation announcements. Anything that
doesn't have an existing alternate and explicit instructions on how to
cope with the change should be considered for rejection.
- Want to do a news announcement that tracks since last supported
release, not last release, including:
- committer names, patches from; diffstat; expanded, non-core-hacker
friendly descriptions
.EOR
19:40 darbelo left, darbelo joined
mikehh What I did since my last report: 19:41
* building and testing parrot on amd64/i386, with gcc/g++
* various fixes
* branch testing and some fixes
* tested rakudo, partcl-nqp, pir/PIRATE and some plumage
* somehow got my name put down as release manager for 2.7.0
What I intend to do in the next week:
* testing and fixing
* put some effort in going through the release documentation
.eor
19:45 ash_ joined
Chandon Done: 19:45
* Found new and interesting Parrot compiler errors.
Will do (in gsoc_threads branch):
* Get timers actually working.
* Write a blog post.
20:00 bacek joined 20:01 khairul joined 20:06 plobsing_work joined
bacek Done: 20:07
* Start reimplementing PCT based PIR compiler from scratch. Codename PIRATE.
* It's alive
* Convinced tcurtis++ to generalize PAST optimizer framework to be Tree optimizer.
* Background thinking of gc_massacre tuning.
Todo:
* Made PIRATE self-hosting. Current todo list on www.rememberthemilk.com/home/bacek...tion.tasks
* Discuss with pmichaud about migrating PCT to PIRATE's POST.
* Possibly merge gc_massacre in current stage without defaulting to new GC.
EOR
plobsing_work What I Did: 20:10
+ work on dynop_mapping
- resolves dynop loading issue
- mostly works right now (hangs one pct test, some codetest failures)
- still a little rough (eg: allocates more memory than necessary in places)
What I Plan:
+ dynop_mapping
- get fulltest passing
- merge 20:11
- improve remaining weak points
Blockers:
+ time
+ brains
EOR
20:11 bubaflub joined
tcurtis What I did: 20:15
* Refactored most of the non-PAST-specific aspects of PAST::Pattern/Walker/Transformer into superclasses(Tree::Pattern/Walker/Transformer and PCT::Pattern).
- Rewrote PAST::Walker/Transformer in NQP to make this easier.
- Now works on any Capture.
- Moved PAST::Pattern::Any/Constant/Closure to Tree::Pattern::*.
* Discovered that you can't partially inherit a multi-method
- Not sure if this is a bug or intended behavior
- Ticket: trac.parrot.org/parrot/ticket/1690
* Refactored implementation of ACCEPTSEXACTLY method for PCT::Pattern subclasses.
- Now instead of having to manually write type-checking, attribute checking, and such for each subclass, just have to have .target_class and .attributes methods.
* Wrote PCT::Pattern subclasses for POST::Node subclasses.
* Added a PAST::Pattern::Control classes(had missed it initially).
* Partially implemented $*ARGFILES for Rakudo.
* Renamed Tree::Pattern::Match.from to .orig for greater similarity to Perl 6 regex Match results.
- Kept .from as an alias for now.
* Worked on optimizations for PIRATE some.
* Blog post: 20:16
- www.parrot.org/content/generalizati...treewalker
What I will do:
* Implement a less recursive variant of Tree::Pattern matching.
- example use case from moritz++: nopaste.snit.ch/21658
* Implement options for .transform to make it more analogous to .ACCEPTS.
* Fix any other problems or bugs moritz finds in my stuff.
* Update tests and docs following the generalization.
* Continue working on optimizations for PIRATE.
* Possibly work on optimizations for other compilers.
* Hope other people start trying out Tree::Pattern/Walker/Transformer.
- Help anyone who does.
* Write another blog post.
20:16 chromatic joined
chromatic Done: 20:17
- plenty of optimizations, mostly for Rakudo startup but some runtime
- planning Lorito's security model
- thinking about new metamodel post 2.6 20:18
Will do:
- more optimizations
- working on a new STRING API, could use some assistance
- helping to merge and tune gc_massacre
Util # Done in the last 2 weeks (missed last #ps due to YAPC): 20:20
* Rakudo - Changed dynops build from deprecated ops2c.pl to new NQP-based ops2c
* Parrot - reviewed branch cfunctionsdocs
* YAPC - Attended conference, and all Perl6-related talks and BOFs.
# Plan to do:
* Perl6book additions
* Fix Win32 Rakudo problems
# Blockers:
* $WORK
.end
khairul Did: 20:26
. Blog post at parrot.mangkok.com/?p=118
. Added tests for Instrument::Probe
. Instrumented GC_Subsystem
Will Do:
. Complete the events interface for GC.
. Instrument PMC vtables.
EOR
ash_ Done, I pass all of the NCI tests now except the callbacks, (working on those), after i finish callbacks, more testing on would be nice (i have OS X 10.6, and have tested on Ubuntu 10.04), I have a windows computer, but the build instructions for libffi on windows are not very straight forward, but it should work fine in theory 20:29
chromatic Hello! 20:31
cotto_work hello
Util Hello
tcurtis Hi. 20:32
bacek aloha
chromatic How'd we do last week?
20:32 allison joined
NotFound Hola 20:33
allison hi
darbelo Hello.
chromatic How'd we do last week? 20:34
bacek gc_massacre isn't merged.
chromatic What's blocking that?
allison review, for one thing
bacek tuning. It's _slower_ on long running processes. 20:35
but consume much less memory.
darbelo bacek: do you have an example benchmark?
bacek darbelo, "time make" in rakudo
chromatic Sounds like an algorithmic thing. 20:36
bacek chromatic, gc ms2 triggers GC much more often. 20:37
chromatic Let's look into that on #parrot
How did we do removing deprecated items?
darbelo Nothing big happened there, as far as I've noticed. 20:39
chromatic Is there much left to yank?
20:40 dukeleto joined
chromatic Sounds like "We don't know." 20:41
What's the state of documentation needs?
cotto_work I have a draft summary deprecation policy up on the wiki.
review would be appreciated before we start calling it binding.
darbelo We have no easy 'delete the code and commit' deprections. We do have some deprecated behaviours.
But we have no clear 'This is what it should do instead.' 20:42
chromatic How important are those?
bacek I think pretty important. 20:43
darbelo Only, from the text in DEPRECATED.pod, I have no way of knowing.
chromatic Let's make that a priority. Can someone take them to the list, or is there a better way to discuss them?
cotto_work +1 to the list
bacek May be we need some kind op policy. Like "develop new code to replace deprecated feature in branch; document changes on wiki; merge after HLLs review/approval" 20:44
s/op/of/
chromatic +1 to more structure
cotto_work Any volunteers to start a discussion on parrot-dev? 20:45
dukeleto says 'ello, sorry I'm late 20:46
cotto_work: start a discussion of exactly what? 20:47
cotto_work the points raised by darbelo
I guess I'll take those then. 20:50
chromatic Is anyone looking into documentation bugs for 2.6?
dukeleto What I did: 20:51
* Got PL/Perl 6 on PL/Parrot working for simple things
Example: github.com/leto/plparrot/commit/6e6...2fce382024
* Created TT #1697 to deprecate open/close opcodes, eligible in 2.7
trac.parrot.org/parrot/ticket/1697
What I will do:
* Work more on PL/Perl 6 so that it works nicely with Rakudo*
Blockers:
* Time and stuff
.EOR
darbelo chromatic: kid51 had a branch for adding function docs. He's been threatening to merge it too.
chromatic I skimmed it. It was a clear improvement. 20:52
mikehh he said he was going to merge today
it clears up t/cod
chromatic How about user-level documentation?
dukeleto chromatic: we really need updated Squak/PCT docs for new users 20:53
mikehh sorry it clears up outstanding C function docs
tcurtis can work on improving PAST docs.
Coke we also need to make sure the docs we have are actually getting published. 'make html' seems subtly broken.
(I'm working on cleaning up the technical bits so others can concentrate on writing actual docs.) 20:54
mikehh Coke: thought you were working on that, need help?
chromatic You're both heroes. 20:55
Let's move on to this week's plans. 20:56
I have a couple of suggestions, mostly related to Rakudo performance.
Coke mikehh: I'm currently stuck on trying to figure out exactly how the current version is supposed to work so I can steal it. should be unstuck tomorrow.
mikehh Coke: one of the problems I had was that book/ uses different POD 20:57
Coke: seems to need PseudoPod 20:58
Coke if anyone has a problem with the current docs generation that isn't content-related, please make sure there's a ticket.
chromatic I know how that works; we can talk about it on #parrot
Suggestions for this week: dynops loading fixes, documentation, performance improvements. 20:59
Coke +1 21:00
cotto_work I'm afk for a meeting. If #ps is still happening when I get back, I'll post my questions then.
chromatic Anyone else have questions?
Let's take on cotto's then. 21:01
NotFound TT #1694
chromatic Is there someone who'd be willing to take on writing the docs for GPG-signed releases?
dukeleto chromatic: i am very interested in that, but would like some help, since I am swimming upside-down in things to do 21:02
chromatic: where would the docs for that go?
chromatic Probably in the release manager guide. 21:03
tcurtis q1q
dukeleto chromatic: i am willing to send an email to parrot-dev about it and make a TT (if there isn't one already)
chromatic Thanks!
tcurtis, go ahead.
Coke (pgp) - write it on the wiki first, solicit feedback before putting in the guide.
dukeleto chromatic: i think we need sha sums as well as GPG-signed releases, just for clarity 21:04
chromatic Agreed.
tcurtis Is TT #1690 intended behavior or a bug? trac.parrot.org/parrot/ticket/1690
chromatic Looks like a bug to me. 21:05
NotFound tcurtis: to me is clearly a bug.
The question is: how many things depend on the wrong behaviour? 21:06
pmichaud very little, actually
at least, very little in my experience.
generally most people are surprised by the behavior.
chromatic I can't imagine anything thinking that's correct.
pmichaud it's been that way, like, forever.
NotFound Not intentionally, but I do care for unintentional usages. 21:07
pmichaud a multimethod defined in a subclass hides any same-named definitions in a superclass
(that's what happens now, not necessarily what should happen)
NotFound There are lost of load_bytecode .... .pir, in examples and libs.
s/lost/lots
allison it's a necessary side effect of the current implementation of multimethods 21:08
(where a multimethod is a single entity that holds an array of methods)
changing it requires replumbing multimethod dispatch
NotFound Uh, sorry, I mixed things. 21:09
chromatic Thoughts on a quick fix for the MMD thing? 21:10
allison rip out support for mmd in core parrot? 21:11
NotFound C++ uses a 'using' clause to solve that same problem.
allison leave it to the languages that actually use it?
chromatic I hope that's tongue in cheek. 21:12
pmichaud PCT uses MMD heavily. :) 21:13
chromatic Sounds like hopes of a quick fix are small. 21:14
allison yah, I'm only half-serious
chromatic allison and I talked about redoing MMD in the near future anyway.
allison but it is damnably slow
yes, what we have isn't true mmd 21:15
pmichaud I think when jnthn re-wrote mmd in rakudo, it ended up being almost as fast (if not faster) than non-mmd dispatch.
allison we should look closely at that implementation as we look at revamping parrot's 21:16
pmichaud I know we were pleased that rakudo's mmd was faster than parrot's mmd.
chromatic Anything else on this topic for #ps?
bacek PaFo organisation on github 21:17
github.com/blog/674-introducing-organizations
dukeleto does anybody know who the "parrot" github account is?
allison bacek: good idea
pmichaud dukeleto: I do, I do! 21:18
bacek pmichaud, speak! :)
pmichaud istr squatting the account last year so it wouldn't get taken by a hostile entity :)
bacek pmichaud, then convert it to organization :) 21:19
dukeleto pmichaud: thanks :) 21:20
sorear mikehh: PseudoPod and POD6 are very closely related. If you're just browsing you may have luck with App::grok
dukeleto pmichaud: i would like to put the parrot github mirror on that instead of my personal github account, we can talk more about that in #parrot or parrot-dev
chromatic Anything else for #ps today? 21:21
bacek cotto's question about using github as "primary" git repo. 21:22
chromatic ~0 from me
bacek +0.5 from me
chromatic I'd hate to migrate Trac, what a pain. 21:23
I'm not a fan of relying on a company rather than a university for hosting.
I like the fork queue though.
tcurtis chromatic: according to cotto, there's a github plugin for trac, so we could continue using trac.
chromatic Hence the ~0. 21:24
mikehh it would still require migration
chromatic Any other comments, or shall we table it? 21:27
bacek github.com/davglass/github-trac looks promising 21:28
21:28 tcurtis left, tcurtis joined
mikehh just bear in mind we have a major release coming up next 21:29
chromatic Yeah, it's polish time.
mikehh and feedback on any Rakudo* requirements 21:30
chromatic Agreed. Let's call it a week. 21:31
21:33 bacek left 21:35 darbelo left 21:48 Chandon left 22:02 bubaflub left 22:27 NotFound left 22:28 dukeleto left 23:00 whiteknight joined 23:05 PerlJam joined 23:10 pmichaud joined 23:20 Util joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!