|
Parrotsketch Every Tuesday @ 18:30 UTC | IRC Log at irclog.perlgeek.de/parrotsketch/today Set by moderator on 24 May 2009. |
|||
|
00:28
particle joined
00:43
amuck joined
05:35
davidfetter joined
06:07
masak joined
11:56
Whiteknight joined
15:48
amuck joined
16:33
pmichaud joined
16:53
Whiteknight joined
17:05
jhorwitz joined
17:13
grantbow joined
17:24
NotFound joined
17:40
Util joined
17:52
PacoLinux joined
18:00
davidfetter joined
18:03
barney joined
18:22
allison joined
18:25
particle joined,
fperrad joined
18:28
masak joined
|
|||
| masak | \\o/ | 18:30 | |
| japhb | o/ | ||
| pmichaud | \\o_ | ||
|
18:30
clinton joined
|
|||
| pmichaud | (my right shoulder is a little sore) | 18:30 | |
| er, left. | 18:31 | ||
| Whiteknight | hello | ||
| japhb wonders what we're spelling in flag letters | |||
| Util | hello | ||
| barney | hi | ||
| pmichaud | "time for #parrotsketch" | ||
| allison | hi | ||
| grantbow | hello | ||
| fperrad | hello | ||
|
18:33
chromatic joined
|
|||
| chromatic | allison? | 18:33 | |
| allison | - Finished editing the book. Will check it into Parrot trunk today. | ||
| - Have to miss YAPC::NA this year (I have a concussion, doctor won't let me travel). | 18:34 | ||
| - On the plus side, that should give me some extra time to work on building languages from an installed Parrot. | |||
| EOR | |||
| grantbow | allison: what happened? | ||
| chromatic | barney? | ||
| barney | Pipp: fixed setting of default values for class attributes | 18:35 | |
| .eor | |||
| allison | grantbow: I walked into a large glass door at a hotel (jetlag, sun shining in my eyes, etc) | ||
| masak | :/ | 18:36 | |
| chromatic | I fixed some bugs and plugged some memory leaks. | ||
| grantbow | allison: ouch, I hope you feel better soon. | ||
| chromatic | I helped edit the PIR book. | ||
| I'll give allison's talk at OS Bridge on Thursday. | |||
| fperrad? | |||
| fperrad | * French Perl Workshop : give a talk about "Markdown on Parrot" | ||
| available on fperrad.googlepages.com/MarkdownOnParrot.pdf | |||
| * see a great talk about visualization of Perl ecosystem | |||
| maps available on cpan-explorer.org | |||
| EOR | |||
| chromatic | japhb? | 18:37 | |
| japhb | * Helped Tene with PIR libs exporting to HLLs | ||
| * Updated OpenGL to support PIR->HLL export | |||
| * Cleaned up OpenGL examples, especially Perl 6 variants in light of above | |||
| * Worked around Gentoo Mesa shipping Windows GL headers | |||
| chromatic | jhorwitz? | ||
| jhorwitz realizes #ps started without him | |||
| ahem... | |||
| fixed mod_parrot & mod_perl6 to work with the recent HLL namespace changes | 18:38 | ||
| have not had any time to work on PDD10 (embedding API) for 1.3 release. q1q regarding that | |||
| EOR | 18:39 | ||
| chromatic | masak? | ||
| masak | submitted 12 rakudobug since last week. worked a bit on my usual projects, mostly Web.pm. >>EOR<< | ||
| chromatic | moritz? | ||
| NotFound? | |||
| particle? | 18:40 | ||
| particle | prepping for yapc::na keynote and pvmw | ||
| ...not enough time... | 18:41 | ||
| .end | |||
| chromatic | pmichaud? | ||
|
18:41
Coke joined
|
|||
| pmichaud | General: | 18:41 | |
| * Rakudo now passing 11,535 spectests (+60 since last week) | |||
| * I now have a grant for 1 Rakudo day per week from Vienna.pm | |||
| * Took an unplanned trip for the weekend, so little done then | |||
| * Proposed detailed schedule for pvmw in Pittsburg, comments welcomed | |||
| * Worked substantially on building Rakudo from installed Parrot | |||
| ** Still quite a bit more to do there, need testers | |||
| ** Not likely to be working in time for Parrot 1.3 / Rakudo #18 releases | |||
| * Fixed unicode and iso-8859-1 string concatenation problems | |||
| * Did some troubleshooting of GC issues causing memory leaks in Rakudo | |||
| * Did troubleshooting of readline_interactive | |||
| Other Rakudo stuff: | |||
| * Fixed () to return Nil | 18:42 | ||
| * Added various unicode ranges to Str increment/decrement | |||
| * Rakudo day: Working on refactoring readonly checks to be read-write checks | |||
| q1q | |||
| eor | |||
| chromatic | Tene? | ||
| Util? | |||
| Util | * Fixed TT#688 (pbc_to_exe too slow) for the remaining compiler (MSVC). | 18:43 | |
| Unless someone reports a slow Rakudo perl.exe build platform, I will close the ticket. | |||
| * Poked at TT#691 (pbc_to_exe fails when installed). | |||
| Finding multiple points of failure; total issue complexity is unclear. | |||
| # I am out of tuits from now until the Parrot VMW; prepping for YAPC talk. | |||
| * Stalled: | |||
| * perl6-examples - Shootout benchmarks, and the un-ticketed bugs from them. | |||
| .eor | |||
| pmichaud | queue another q | 18:44 | |
| chromatic | Whiteknight? | ||
| Whiteknight | * Fixing bugs and closing tickets | ||
| * Lots of testing for the release. can fulltest now in about 7min, so much less painful (was > 1hr last month) | |||
| * Practicing the release procedure. | |||
| * Will start the release right after #ps, unless there are last minute blockers. | |||
| * Good job this month everybody! | |||
| chromatic | Did I miss anyone? | ||
| Coke | ~~, I'm late. | ||
| chromatic | Coke? | ||
| Coke | Parrot - minor cleanups for release. | ||
| Partcl - - now passing 2,972 spec tests. | |||
| - started to think about eliminating custom call chain and using parrot's. Need documentation. Would be a big win for interoperability and avoiding lots of handrolled code. | |||
| - fix [exit] in partcl's REPL so it actually does that. | 18:45 | ||
| - eliminate helper .sub 'toList' and instead put a method on various pmcs/classes - avoids a lot of explicit "look up this code" and MULTIS. | |||
| - eliminate some unnecessary code (overrides from years ago that are no longer needed) | |||
| - remove TclObject; it was only used to avoid a very small amount of otherwise duplicated code, and the multiple PMC inheritence headache wasn't worth it. | |||
| - Handle "\\udead" processing, which isn't valid unicode, but is valid Tcl. (Not handling it correctly from Tcl's standpoint yet, but can now parse a spec test file we couldn't before.) | |||
| - created a branch to switch TclList from a PMC to a class. 'make test' works, but 'make spectest' has many regressions. Goal is to avoid crossing the PBC/C boundary quite as often, nice side effect of having the code more readable in PIR. | |||
| . | |||
| chromatic | Anyone else? | ||
| Alright, let's talk questions. jhorwitz? | |||
| jhorwitz | embedding API was supposed to be completed for the 1.3 milestone today. it was my task and i've had zero tuits to work on it. should we just bump the milestone? | 18:46 | |
| Coke | Yes. | 18:47 | |
| Do you think you'll get to it for 1.4? | |||
| jhorwitz | yes -- this month is much better for me | ||
| Coke | (if so, bump there; otherwise just remove it) | ||
| jhorwitz | it needs to get done | ||
| chromatic | It can be parsimonious; it doesn't have to be complete. | ||
| It's easier to add features than to remove them. | 18:48 | ||
| allison | yes, go ahead and bump the milestone on the ticket | ||
| jhorwitz | very well. next! | ||
| chromatic | pmichaud? | ||
| Coke | q1q | ||
| pmichaud | Util: how comfortable are you with the TT#688 change that was just made? | 18:49 | |
| i.e., has it been fairly well tested on various platforms? | |||
| Util | Quite comfortable. Win32/MSVC is very tested, including Rakudo spectest. GCC codepath is solid. The only possible problem is if I erred in DRY refactoring across the default codepath. | 18:50 | |
| pmichaud | okay. | ||
| just flagging the usual warnings about significant changes immediately before a release :-) | |||
| (because if the release goes out and Rakudo can't build/run against it, we're hosed :-) | |||
| Coke | I didn't care since pbc_to_exe is still broken over in tcl land. | 18:51 | |
| pmichaud | next question | ||
| I'm starting to see a lot more -G related failures again in Rakudo; | |||
| likely due to the various memory leak cleanups in Parrot. Any suggestions? | |||
| (Review: A "-G failure" indicates a program that fails when run | |||
| normally but passes when the -G flag is present.) | |||
| chromatic | No suggestions here without seeing backtraces or test cases or both. | ||
| Coke | If it really is something not getting GC'd, chromatic had some nice articles on how to debug those. | 18:52 | |
| pmichaud | the problem is usually something being gc'd early | ||
| NotFound | Sorry, been busy. Nothing to report. | ||
| pmichaud | or, that the behavior of the program completely changes when run with -G | ||
| Coke | er, "properly" is missing from my last send. | ||
| pmichaud | in the latest incarnation, I'm starting to add a few more property flags to some PMCs | 18:53 | |
| doing so is causing method calls to jump to the wrong method entirely | |||
| chromatic | Through PMC_metadata? | ||
| pmichaud | through the "setprop" opcode, whatever it does. | ||
| but if I run with "parrot -G", it runs correctly. | |||
| chromatic | That uses PMC_metadata. That could mean we're not great about managing pmc_ext. | 18:54 | |
| Though Rakudo's subclass_or_rebless (or whatever it is) opcode still bothers me. | |||
| pmichaud | agreed, that's why I'm not immediately ascribing the problem to Parrot :-) | ||
| we're also noticing "double free" errors on exit again. | 18:55 | ||
| Whiteknight | those are the worst | ||
| chromatic | pmichaud, do you have time/knowledge to write a short description of your context refcount refactoring? Specifically I'd like to know when, where, and why you incremented and decremented contexts. | 18:56 | |
| Unless it's "Wherever seemed right..." in which case you know as much as I do. | |||
| pmichaud | I can describe it here | ||
| basically "Wherever seemed right..." is a pretty good summation (more) | |||
| chromatic | It might be easier in mail to the list, where archival is better. | ||
| pmichaud | in truth, I didn't truly refactor refcounts | 18:57 | |
| what I did do was add some additional instrumentation to make it easier to find out when things were blowing up | |||
| but afaict, there's not really any overarching logic to when refcounts are increased or decreased, and there are some places that explicitly zero out the refcounts for no obvious reason. | |||
| chromatic | Maybe we should branch after the release and try to track this down. | 18:58 | |
| pmichaud | That was when I started to work on moving contexts into PMCs, but that was also proving to be far more of a pain than I had time to deal with (given that I had already spent 50+ hours on the lexicals problem in the first place) | ||
| so ultimately it came down to "when it works, stop." | |||
| where "works" means "lexicals exhibit the right behavior and we don't have any segfaults" | 18:59 | ||
| "works" explicitly did not include "and we aren't leaking contexts" | |||
| Whiteknight | (don't have any segfaults)++ | ||
| chromatic | Let's put that on the list for debugging in the near future then. | ||
| Is anyone blocking on anything besides time, energy, and magical flying candy-flavored unicorns? | |||
| Coke | Yes. | ||
| pmichaud | I'll be glad to help as much as I can there, but I got seriously burned by it before. | ||
|
19:00
clinton left
|
|||
| chromatic | Anyone besides Coke blocking? | 19:00 | |
| pmichaud | (done with my questions/comments) | 19:01 | |
| chromatic | Coke, what are you blocking on? | 19:02 | |
| pmichaud | (Coke had a question come in late) | ||
| Coke | my blockers are: code.google.com/p/partcl/wiki/ParrotIssues (as usual), but as I mentioned yesterday in #parrot and just now in report, docs on how to use parrot's call chain. (return continuation chain?). | ||
| I suspect "rakudo's source" is probably the best sample code, but was hoping for something a little more entry level. =-) | 19:03 | ||
| pmichaud | Coke: I'll be glad to discuss in #parrot after #ps | ||
| allison | how to use it beyond calls and returns? | ||
| Coke | pmichaud: perfect, thank you. | ||
| allison: (for things like uplevel, upvar, and throwing exceptions up N levels.) | |||
| pmichaud | allison: just a reminder that I did ask about having a way to force return from a caller | ||
| allison | pmichaud: aye | 19:04 | |
| Coke: more advanced manipulations, yes. pmichaud can give you the current set of workarounds | |||
| Coke: I can say we need to provide a better interface there | 19:05 | ||
| pmichaud | well, there aren't many workarounds yet -- I'm hoping for some underlying Parrot improvements. | ||
| Yes, exactly what allison++ said. | |||
| Coke | at this point, I'd be happy to get the stuff I already have working working the right way before adding in all the crazy stuff. | ||
| allison | the details on how you all need to use them will help in defining the future interface | ||
| chromatic | Let's do roadmap review. | 19:06 | |
| pmichaud | allison: Coke and I can discuss our mutual needs and then I'll get a report to you next week? | ||
| NotFound | I have a quick question | ||
| allison | pmichaud: sounds great | ||
| chromatic | Go ahead, NotFound. | 19:07 | |
| NotFound | Can we add a isnull string opcode after the release, or needs a RFC? | ||
| allison | adding after the release is fine | 19:08 | |
| NotFound | Ok | ||
| pmichaud | adding before release would be okay, too, since it's a new feature and not a deprecation. | ||
| allison | (it's deprecations that we have long cycles on, additions are safe) | ||
| pmichaud | but better to wait the few hours at this point :-) | ||
| NotFound | pmichaud: yes, but is late and I lacked the time to do some tests | 19:09 | |
| allison | yeah, better to wait, just in case of a test failure on a more obscure platform | ||
| and yes, definitely needs tests | |||
| NotFound | Anyway, the first usage I have in mind will need a deprecation.. | ||
| allison | okay. the deprecation will have to wait until after 1.4 | 19:10 | |
| but make sure you get the notice in before 1.4 | |||
| on to roadmap review... | 19:11 | ||
| trac.parrot.org/parrot/report/14 | |||
| NotFound | allison: is about readline_interactive, but probably a discussion too long for today, better next week. | ||
| allison | NotFound: we could discuss on #parrot right after the meeting | 19:12 | |
| NotFound | allison: ok | ||
| allison | TT #754 is marked as roadmap, but not marked with a milestone or discussed in #parrotsketch | ||
| should it be marked as roadmap? | 19:13 | ||
| and if so, which milestone? | |||
| Whiteknight | no | ||
| japhb | allison: My error, I think. I think no one was around when I created that ticket, so I guessed. | ||
| allison | okay, will change to TODO | 19:14 | |
| japhb | ... it is done, but I didn't know whether to close it before release. | ||
| allison | closing it is fine if it's done | ||
| no need to wait until after the release | 19:15 | ||
| (but you probably should include a mention of it in the deprecation notice, just for documentation) | |||
| japhb | OK, I was confused about the proper way to deal with a TT that only exists to support DEPRECATED.pod | ||
| It is correctly linked in D.pod | |||
| allison | usually the associated ticket gives the details of what needs to be done in the deprecation | 19:16 | |
| this one happens to be a ticket saying to deprecate the feature | |||
| so, agreed the ticket itself is complete | |||
| japhb | allison: remember, this ticket was the one that led to 'experimental' status. | ||
| japhb resolves the ticket | 19:17 | ||
| allison | no need to leave it hanging around as an open task | ||
| japhb | nodnod | ||
| allison | on to 1.3 roadmap tickets: the HLL export tickets seem to be still on track | ||
| Coke | chromatic: you should review the text japhb & I wrote in DEPRECATED.pod. | ||
| pmichaud | HLL export is continuing to progress, we have some things in for 1.3 but not finalized | ||
| japhb++ and Tene++ for their efforts. | 19:18 | ||
| allison | yes indeed japhb++ and Tene++ | ||
| pmichaud | While I wouldn't necessarily call it "landed", we do have something in place where we can show good progress | ||
| allison | should some of those tickets be bumped to 1.4? | ||
| That's... | |||
| #566 \t export conventions (cross hll) | |||
| #567 \t pdd31-hll interop | 19:19 | ||
| #568 \t hll interop | |||
| pmichaud | yes, we can bump to 1.4 | ||
| allison | all three? | ||
| japhb | #566 seems to point to a wiki page that is empty | ||
| Whiteknight | do all of thsoe need to be separate tickets? | ||
| pmichaud | it's fine for them to be separate | ||
| I like them separate | |||
| Whiteknight | ok | ||
| pmichaud | anyway, bump to 1.4, list me as owner | ||
| allison | okay, will do | 19:20 | |
| #588 \t pdd10-extending/embedding/nci, define parrot api | |||
| already discussed in questions | |||
| will bump to 1.4 | |||
| end of roadmap review | |||
| japhb | ... actually, all three of the HLL interop tickets point to a blank wiki page. Something got moved? | ||
| NotFound | I think the problem is that nobody wrote that page. | 19:21 | |
| allison | that was the old roadmap page | ||
| the tickets were extracted from that page, then deleted the page | 19:22 | ||
| (it was getting silly maintaining the ticket list in a wiki page when a report does just as well) | |||
| particle | progress++ | ||
|
19:22
amuck joined
|
|||
| japhb | Ah. And the only descriptions we have are the titles? They're a little ... minimalist. | 19:22 | |
| allison | the page didn't have any more details on the task, just what's currently the title of the ticket | ||
| japhb | gotcha | ||
| allison | the roadmap tickets were all created from the Parrot Dev Summit, quick task descriptions on cards for the wall organization exercise | 19:23 | |
| chromatic | Anything else, or shall we call it a week? | 19:24 | |
| Whiteknight | I'm cutting the release nowish, | ||
| unless there are any issues outstanding | |||
| particle | checked the latest smokes? | 19:25 | |
| Whiteknight | earlier | ||
| particle | cut it now! | 19:26 | |
| Whiteknight goes on his merry little way | |||
| allison | thanks, Whiteknight! | ||
| Coke | cut cut. | ||
| particle wonders if rakudo passes more spectests that parrot does parrot tests | 19:27 | ||
| chromatic | Probably. | ||
| particle | that metric intrigues me. | 19:28 | |
| Util | Rakudo: 11672 subtests | ||
| particle | considering the subset of parrot that rakudo actually uses. | 19:29 | |
| msvc parrot passes 11797 tests, probably more on linux. | 19:30 | ||
| so... close. | |||
| moritz | so it's close ;-) | ||
| Util | Parrot: 12068 subtests | ||
| particle | ~400 tests. | 19:31 | |
| chromatic | Thank you everyone. | ||
| Util | Rakudo has the advantage of a spec, from which to write (even not-yet-compilable) spectests. | ||
| moritz | shouldn't the PDDs fill the same role? | 19:33 | |
| chromatic | They're not as well tested as they should be. | ||
| particle | io tests existed long before the implementation, they were developed along with the pdd | 19:34 | |
| anyway, parrot needs better test coverage. a pir profiler will help with that. | |||
|
19:50
fperrad left,
Util left
19:57
masak left
20:10
Coke left
20:30
grantbow1 joined,
PacoLinux left
21:01
grantbow joined
21:28
NotFound left
22:07
grantbow1 joined
22:19
chromatic left
|
|||