Priorities for this week: all tests to pass on Win32; exciting branches, boring master; no dogs on fire | Post closed tickets in your report. | Note: This channel is for our weekly status meetings (Tuesdays at 19:30 UTC); you probably want #parrot instead. | irclog: irclog.perlgeek.de/
Set by moderator on 12 July 2011.
00:39 lucian left 00:40 darbelo left 03:37 bluescreen left 08:53 contingencyplan left 09:22 particle1 joined 09:26 particle left 12:20 bluescreen joined 13:40 Coke left, Coke joined 14:00 darbelo joined 14:07 darbelo left 14:09 lucian joined, darbelo joined 14:19 Coke left, Coke joined 14:29 zloyrusskiy joined 14:32 zloyrusskiy left 14:40 darbelo_ joined, darbelo left 14:41 darbelo joined, darbelo_ left 14:46 darbelo left, darbelo joined 14:47 PacoLinux joined 14:50 PacoLinux left 15:05 lucian left 15:15 contingencyplan joined 15:30 darbelo_ joined 15:35 darbelo left 15:46 darbelo_ left 15:51 bluescreen left 15:58 bluescreen joined 16:13 whiteknight joined, benabik joined 16:16 darbelo joined
whiteknight you there? 16:25
benabik nobody here but us chickens.
Felipe :)
whiteknight benabik: awesome. I don't know if you saw my email earlier. You got graded. PASS 16:26
so, money in the bank, and now we can expect more work out of you!
benabik Saw it just as I started talking on IRC. :-)
I'd probably keep working even without a PASS, although delays would occur with the whole "needing money to eat" bit. 16:27
whiteknight I quickly looked through your commits list in prep for the evaluation. It's looking pretty good
benabik I'm pretty sure the core bits of PAST is working well. Less sure about the less-tested bits pmichaud added.
whiteknight I haven't really looked at the code in a while and haven't tried to build it recently either. I will try when I get home. What should I expect? 16:28
benabik At the moment, the nqp_pct branch should function 100% like master. It has some framework for direct PBC compilation, but I'm just starting to hook it all up. 16:29
the nqp_pct branch in my fork has it wired up to a couple of tests and it's showing some issues.
whiteknight your fork is ahead of the branch in the parrot repo?
benabik I need to diagnose the test I've wired up and see why it's attempting to compile strings instead of PCT nodes... 16:30
Yes. Because of the HEAD commit: "Test PBC stage - horribly broken"
whiteknight oh yes, stands out like a red herring
benabik While keeping branches 100% working isn't required, I like it at least somewhat working. 16:31
Although "horribly broken" really just refers to the tests.
test.
Coke wrong window? 16:32
benabik Coke: We're over here so it's logged but not taking up everyone else's conversation. (I think.)
whiteknight yeah, that was my motivation 16:36
be logged, but be elsewhere
benabik: okay, so the new POST is working exactly the same as master for producing PIR, and you're still wiring up the pbc-generation capabilities for it? 16:37
benabik Yes. PAST->POST->PIR should be 100%
whiteknight awesome, that's the number we like to see 16:38
benabik And I've wired together PAST->POST->PBC, but it's erroring out in odd places.
github.com/Benabik/parrot/commit/a...d3b81851ce
^^ That commit should be the needed infrastructure to output PBC, but as the "horribly broken" commit that follows states, it's not working.
whiteknight okay. 16:40
do you have a handle on the failure? That is, do you know what you need to do to fix it?
benabik I know where it is, but am not sure why it's happening. I think I need to wire up a simple HLL like Squaak so I can poke it more directly. Tests are great for automation but LTA for exploration. 16:42
whiteknight Can you jam in some debugging print statements, and spit out values around where the bug is happening?
benabik Yes. And, actually, the test leaves the PIR behind so I can explore it directly instead of recompiling everything. 16:44
benabik is poking it with a stick right now.
whiteknight okay. Last question from me: Is there anything you need help with, any external support you need? Anything need to change in Parrot or elsewhere outside the scope of your work to support what you are doing?
benabik There's nothing that's a total blocker right now, although there are some notes in the code I'm working with that I don't know are up to date. 16:47
whiteknight okay.
benabik It says "Currently there is no way in Parrot to generate Packfile and attach it to Interp via existing API". IIRC, you were working/are going to work on that.
whiteknight It doesn't need to be attached to the interp now, I don't think
I'll have to look back at that. plobsing may have some insight as well 16:48
benabik Right now it writes the packfile out to /tmp/temp.pbc and uses load_bytecode.
Which is LTA, but I don't know of a better way.
whiteknight oh, so you need some kind of API to take a bytecode stream in something like a STRING literal, and pump that into a packfile?
er, Packfile PMC? 16:49
benabik I have a Packfile PMC. 16:50
I need a way to run the bytecode in it.
Right now it writes it out to a temp file, uses load_bytecode and find_sub_not_null to get the main sub. 16:51
(This technique is stolen from bacek's PIRate work so has no knowledge of anything done recently.) 16:52
whiteknight Packfile, or PackfileView? 16:53
We can probably add a method to Packfile to get a PackfileView. That is what IMCC returns, and you can query subs from it
benabik And those subs can be run directly? That would simplify matters. 16:54
whiteknight yes
There is a .main_sub() method that returns the main sub, an .all_subs() method to get an array of all subs, and a few other related methods 16:55
and the interface there is going to improve a little bit more after the 3.6.0 release too
benabik Yes, that would de-ugly a lot of the pbc HLLCompiler stage. It currently uses a static temp file and returns a closure that handles loading and searching for the main sub.
If I could just do (the moral equiv of) return PackfileView.new(packfile).main_sub(), that would be excellent. 16:56
whiteknight probably be sub = packfile.view.main_sub() 16:57
benabik hand-waves
Technically, it would all be PIR b/c I don't want to spend time re-writing another class instead of getting generation working correctly.
However it's spelled, it sounds more awesome than now. 16:58
I might even be able to remove some main sub tracking code out of POST::Compiler. Hm. 17:00
whiteknight I'll hack that method up in the whiteknight/pbc_pbc branch. I'm going to get that branch cleaned up and probably merge shortly after 3.6.0 is out 17:04
if you can wait that long
benabik Since PBC generation isn't working correctly yet, I'd say there's no rush. :-)
whiteknight yay! I like not being rushed 17:05
Okay, if bacek doesn't show up, let's plan on another meeting late next week. We can work out the details as we get closer 17:06
and keep up with the blogging too, it's always very interesting
benabik Sounds like a plan. 17:08
I feel like I've talked to bacek about once a week, but it might be closer to every other, so late next week is a good idea.
benabik 's concept of time doesn't always match reality.
whiteknight okay. I know he's been busy, so I don't want to you to be without proper mentoring. We can always chat between now and then too, of course 17:14
benabik Everyone on #parrot has been helpful. :-) 17:15
17:20 lucian joined
benabik And my wife is agitating for me to get going, so unless there's anything else...? 17:23
whiteknight nope. Have fun 17:33
17:44 benabik left 18:00 Coke left, Coke joined 18:36 darbelo left 18:37 darbelo joined 18:58 whiteknight left 21:15 bluescreen left 21:37 darbelo left 21:39 lucian left 22:04 contingencyplan left

Powered by Cro and the Raku® Programming Language.

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

Thank you!