Perl 6 - the future is here, just unevenly distributed

IRC log for #parrotsketch, 2013-11-05

| Channels | #parrotsketch index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
13:30 lizmat joined #parrotsketch
14:43 bluescreen joined #parrotsketch
16:35 Coke joined #parrotsketch
16:35 Util joined #parrotsketch
16:40 pmichaud joined #parrotsketch
19:09 lizmat joined #parrotsketch
19:29 Util Pre-report:
19:29 Util # Done:
19:29 Util * Merged branch 'new-packfile-api'. Much thanks to rurban++ and gerdr++
19:29 Util * Caught up with filing 10 new issue tickets from Zefram++
19:29 Util * Flagged all the old Zefram tickets with the Component and Priority|Severity that he had supplied, but I had omitted.
19:29 Util * Fixed typos in Perl 6 specs.
19:29 Util * Upgraded from OS X Mountain Lion to Mavericks
19:29 Util + To allow me to work on https://github.com/parrot/parrot/issues/1016
19:29 Util # Plan to do:
19:29 Util * GH#1016
19:29 Util .end
19:30 Util Hello! Who all is in attendence?
19:32 rurban1 joined #parrotsketch
19:32 Util No worries; we usually have such attendance issues just after the time changes.
19:37 rurban I got nothing to report. But we got some sev-high bugs now
19:39 Util True. I am tackling GH#1016 due to it bugging me, but we should start talking about severity and priority, and actually work from those.
19:39 rurban no blocker, 1 new critical, 4 old critical
19:40 Util That is not to say that Zefram's assessment of severity and priority has to become *our* assessment, but it is a very good starting place.
19:42 rurban benabik++ for doing the actual work on the new-packfile-api
19:43 Util GH#548 is marked critical, and since I need to revisit the Complex class anyway, I will tackle it after GH#1016.
19:44 Util benabik++ indeed. Thanks for pointing that out!
19:44 rurban We have Sev-* and Pri-* which is kinda mix of severities. https://github.com/parrot/parrot/issues?labels=Sev-high&state=open seems to be a good start
19:45 rurban #548 is indeed a bummer. maybe we should ask nqp/perl6 what hurts them most nowadays (besides performance)
19:46 rurban #1018 is actually a feature, no bug
19:49 Util Asking nqp/perl6 "what hurts them most nowadays" is always a good idea, IMHO
19:50 rurban I closed some of the thread bugs which are actually features
19:53 Util Last time I asked pmichaud about it, he said these were pain points in Parrot:
19:53 Util 1. calling conventions
19:53 Util 2. exception handling
19:53 Util But I think those are painful because of their poor performance.
19:53 rurban 1. is not fixable for now
19:53 rurban 2. I have no idea about
19:55 Util Maybe the NQP/Perl 6 team can provide more detail, and we can make tickets for their current pain points.
19:55 rurban 1. is why I gave up on parrot, and rakudo is switching to MoarVM because of this.
19:56 rurban we should ask for fixable bugs, not pain points
19:56 Util rurban: What would *you* envision as an ideal calling convention? Just curious, as I have been trying to give it thought, and not getting very far.
19:57 rurban But I have to leave now, getting my glasses. See nothing
19:57 Util rurban: very good point about fixable bugs vs pain points
19:57 Util rurban: Thanks for attending!
19:57 rurban My ideal collaing convention is 100% opposite to chromatics vision
19:58 Util Even if it is not implementable in Parrot, I would love to hear the details sometime. At the very least, it will help clarify my own thinking about the issues and tradeoffs.
19:58 rurban I'd go for cdecl (and did it with p2) c -stack, chromatic goes for CPS in pure bytecode
19:59 rurban CPS is good for exceptions and callcc/yield, but otherwise pretty heavy
19:59 Util Are you providing continuations in p2?
19:59 rurban yes, via ucontext
19:59 Util (as a user-facing feature, I mean)
20:00 rurban callcc is via the op "here", and yield can go back to any here, or to the scheduler
20:00 rurban but I do not switch threads yet
20:01 rurban this is the tricky part
20:01 Util I see (well, mostly)
20:01 rurban this=here; do smthng, somewhere else: goto this
20:02 Util I need to look at some ucontext example code. I will put that on my TODO list, but not for this week. Thanks!
20:02 rurban we cannot use it, we do CPS. it's much easier with CPS
20:02 Util This week, I will invite NQP/Perl 6 team to speak up about fixable bugs in Parrot.
20:03 Util rurban: Yes, I understand we (Parrot) cannot use it, but Util's brain can use it.
20:03 rurban http://lifecs.likai.org/2010/11/continuation-passing-style-calling.html
20:04 Util That is enlightening; thanks!
20:04 rurban That's for my cdecl style. for parrot read up on pure CPS
20:05 rurban fperrad also uses my style (luajit) in his parrot side-project
20:05 Util Right, I already understand pure CPS, just not continuations-without-pure-CPS
20:05 rurban it's basically stack saving and switching
20:06 Util OK
20:06 rurban But a bit more than longjmp, because it needs to handle signals and threads
20:06 rurban So they invented ucontext for a better longjmp
20:07 Util Sure, like "state" instead of just "stack"
20:07 Util (saving and switching)
20:07 rurban yes, but on windows you need something else, like fibers
20:09 rurban left #parrotsketch
20:09 Util Anything else to discuss, from anyone?
20:11 Util Thank you for attending (even via lurking or backscrolling)!
20:12 Util Meeting adjourned.
20:13 Tene joined #parrotsketch
21:08 rurban1 joined #parrotsketch
21:32 rurban1 joined #parrotsketch
23:02 Tene joined #parrotsketch
23:15 Tene joined #parrotsketch

| Channels | #parrotsketch index | Today | | Search | Google Search | Plain-Text | summary