"Tuesday at 20:30 UTC"
Set by moderator on 3 August 2010.
00:04 wagle joined 01:28 wagle joined 01:39 ascent joined 11:45 wagle joined 12:15 kid51 joined
kid51 kid51's report (for several weeks) 12:24
DONE
* Posted to list report about Parrot/Rakudo increased memory footprint leading to inability to compile Rakudo Star from source on machines on which it once compiled.
* Created TT #1725 re headerizer documentation; created branch to work on this.
* Took TT #1726 created by mikehh re .pmc files lacking function documentation. Created t/codingstd/pmc_docs.t to identify such functions; merged to trunk last night.
DOING/WILL DO 12:25
* Will go through list of undocumented PMC functions, try to determine who is best to write docs on each, then contact contributors via email.
* If everybody pitches in, we can get this all done quickly.
* Have organized New York City Rakudo Star study group.
QUESTIONS/COMMENTS
* Over the last 1-1/2 years we've had a lot of discussions about improved garbage collection, LLVM, Lorito, etc.
* But my impression is that most of this work is still in the future, i.e., not much has actually been accomplished.
* Is that impression correct? When I speak to non-Parrot people about Parrot, I need talking points.
* Which of our improvement projects can we actually get done, say, by October 15?
EOR
13:18 bluescreen joined 13:41 bluescreen joined 14:55 mikehh joined 16:27 robin-gvx joined 16:57 cotto_work joined 18:10 NotFound joined 18:30 robin-gvx joined 18:35 bubaflub joined 19:18 khairul joined
khairul did: 19:23
. code cleanup to prepare move to github at the conclusion of GSoC.
. minor bugfixes.
. partial user guide at parrot.mangkok.com/?page_id=140
will do:
. figure out a apply a fix relating to vtable syncing between supervisor and child interpreter.
. finish up user guide.
. hopefully add more test cases.
eor
19:23 khairul left 19:26 atrodo joined
Coke Only did minor hacks on partcl-nqp this week. Still need to get back to the 'make html' fixes. 19:27
19:27 smash joined 19:38 mikehh joined 19:39 darbelo joined 19:40 tcurtis joined
bubaflub did: 19:45
* GSoC - parrot configures out of directory 19:46
* GSoC - parrot can run configure-time test programs in build directory
will do:
* GSoC - allow templated files to be in build (rather than source) directory
* GSoC - allow an out of directory build
tcurtis What I did: 19:50
* Extracted my test optimizations for constant-folding and tail call elimination for NQP into an indepedent module
- Here: github.com/ekiru/simple-optimizations
* Expanded the constant-folding optimization to support several ops, add, sub, mul, fdiv.
* Fixed various bugs.
* Added support for bundling up passes in modules and adding them to optimizers more conveniently.
What I will do: 19:51
* Write a blog post.
* Fix bugs.
* Improve docs.
* Improve tests.
* Prepare for "firm 'pencils down'" date Monday.
EOR
19:59 chromatic joined 20:00 Chandon joined
cotto_work #done: 20:01
- met with khairul
- started on trac plugin, got help from dafrito++ mapping between svn revisions and git hashes
#hope to: 20:02
- get the plugin into a usable state and set up a test site
- write something cool to show what Khairul's been doing
#eor
NotFound What I did: 20:03
-parrot
* Minor refactors and nano optimizations
* Implement init_int in Default, allowing usage of init_int in
HLL mapped internals.
* Added a few tests
-winxed
* Initial review of plobsing issues
What I will do:
No plan, short of available time
EOR
Chandon Did: 20:13
- Green threads with native threads for blocking IO mostly works in my branch.
Will do:
- Fix some remaining bugs before gsoc pencils down.
- Wrap blocking operations other than FileHandle#readline
- Add more tests.
END
mikehh What I did since my last report: 20:17
* building and testing parrot on amd64/i386, with gcc/g++
* some fixes
* testing rakudo, partcl-nqp, partcl, pir/PIRATE, winxed and plumage
* tracking down some documentation problems with pmc's
What I intend to do in the next week:
* testing and fixing
* prepare for and release 2.7.0
.eor
chromatic I worked on some optimizations and found more infelicities (hashes, MRO). 20:21
I also did some profiling of the gc_massacre branch (huge memory usage for NQP-rx).
I will... do what I can.
NotFound Forgot to say: I also made the worse demo video ever seen: www.youtube.com/watch?v=trtvDaLUJb8 20:24
20:28 Paul_the_Greek joined
Paul_the_Greek Hey folks. 20:28
cotto_work just a bit early 20:29
(but not much)
NotFound Hola 20:30
cotto_work hi
Util Hello
tcurtis Hi.
chromatic greetings 20:31
Chandon Hi
darbelo Hi. 20:32
DONE
- GSoC soft pencils down date yesterday.
- Worked on the unshared_buffers branch.
- Got a 10% improvement on string stress benchmark.
- 'Real world' tests (i. e. rakudo build) didn't go that well, though.
TODO
- Make compact_pool smarter. That should improve 'Real world' performance.
- Merge back to trunk if the performance improves.
- From there to the gsoc_nfg branch, which needs this more than trunk.
EOR.
chromatic Thoughts on the past week? What went right?
Coke I had little time for parrot this week, sorry. Austin++ was helpful, though.
Paul_the_Greek I learned a lot. 20:33
mikehh some improvments, couple of new tests
darbelo We had one person show up and say he liked our docs :)
cotto_work Paul_the_Greek: and got something committed
Paul_the_Greek Yes, thanks for doing that.
bubaflub all props to cotto_work for the testing and Paul_the_Greek for the code 20:34
chromatic Parrot does run a little bit faster and uses slightly less memory than the week before.
Paul_the_Greek Thanks!
mikehh nobody seemed to break anything 20:35
Paul_the_Greek Is this the forum to point people at a ticket and ask for future comments? I don't feel confident making the decision.
NotFound And some ugly parts are a bit cleaner.
chromatic Those are all good small gains.
What didn't go so well in the past week?
cotto_work Paul_the_Greek: it's not a bad place. Wait until we get to the questions though.
Paul_the_Greek Okay.
Bad thing: I cannot find Windows binaries for GMP 4.1.4 or later. I spent 2 hours looking. 20:36
mikehh I think too many people are waiting for lorito etc
darbelo We still have several outstanding branches waiting to merge.
Most seem to block on tuits, though. 20:37
chromatic There's still a lot of low-hanging ugly and slow.
Paul_the_Greek I was thinking of jumping into the PIR assembler, but I can't figure out which one to work on. What's the plan with that?
Tene Out of all the things I'm blocking, none of them are lorito.
darbelo Oh, and we broke lua. 20:38
chromatic Most of my blocking has been "Wow, I wish I had NotFound or bacek or someone to help with this code!"
NotFound chromatic: I'm short of time these days. 20:39
Coke a common lament.
chromatic So am I, and what time I have I'm short of brilliance.
Tene I have plenty of time; I'm just short of sane.
Paul_the_Greek Short of knowledge. 20:40
chromatic Any other laments before we start looking at how to fix our lacks?
NotFound darbelo: but we fixed it, I think.
Someone looking for a quick and easy task: add a test for that. 20:41
chromatic Let's move on to What Now. 20:42
Suppose I do some profiling and figure out that mro calculation is significantly expensive for Rakudo.
How do I work with a Tene or a darbelo or a plobsing to fix that?
Paul_the_Greek mro? 20:43
Util method resolution order
Paul_the_Greek Got it.
mikehh we have 1 week before the 2.7.0 release, 20:44
I would like to see html_cleanup land before then
Coke (*&@#$. Hokay. 20:45
mikehh and any merges in the next couple of days
chromatic Slushy late Saturday?
mikehh so we can test properly before the release
Tene chromatic: I might be free on thursday or friday night, or sometime on Saturday. 20:46
mikehh and I probably need lots of help as I have not done this before
cotto_work I can help with uploading the tarball and docs if needed. 20:47
It really is mostly just following the doc. 20:48
Tene If you really want me to come work on something with you, you can try asking me to work with you one of those times.
chromatic Trouble is me predicting my availability. 20:49
Util chromatic: (in general) Create a ticket/wiki/branch containing your analysis and early work, and announce on #ps/#parrot/mailing-list that you need help with it; a msg/cc to Tene/darbelo/plobsing would be great in case they miss the original announcement, but others (that you don't realize have the skills) might have time where your intended target do not.
Tene I can easily predict my availability, just not my sanity. 20:50
chromatic Is it as simple as "I need help, so please let's collaborate?"
Tene Could be. That's fairly cheap to try, at least. 20:51
chromatic Will do.
Util Isn't it?
mikehh we could try for a hackathon ovewr the weekend or something like that
chromatic A hash cleanup hackathon would be nice, and it's probably more accessible than unshared buffers or GC massacre. 20:52
Then again the make html might be more immediately useful and less risky.
mikehh html_cleanup is mostly there - just indexing and the like 20:53
chromatic Other thoughts?
darbelo With a realease this close, I'd go for less risky. 20:54
mikehh If it breaks too much we can always revert
chromatic We can always develop in a branch.
Wholesale hash replacement should be a branch anyway.
Coke +1 on doing that, and in a branch. 20:55
helping gc_massacre hit would be nice, though.
(too many chainsaws in the air)--
chromatic I think that one blocks a lot on bacek.
mikehh who seems to have been very involved with $work recently 20:56
it would be nice to get some branches ready to land after the release 20:57
chromatic Can we set priorities for those? 20:58
darbelo The dynop mapping brancb seems pretty important. 21:00
mikehh html_cleanup this week, gc_masacre asap, hash_allocator and other hash asap
that as well
chromatic Oh, dynop mapping. That'd be nice for the release.
mikehh I saw quite a bit of work there, what still needs to be done? 21:01
NotFound I can try to take a look at it.
chromatic Any corrections to this order? dynop mapping, html cleanup, gc massacre, hash allocator, hash revision? 21:02
darbelo plobsing sent a mail to the list recently outlining what was missing, IIRC. 21:03
mikehh so html_cleanup and maybe dynop_mapping this week
21:03 whiteknight joined
mikehh chromatic: looks good 21:03
darbelo Found it: lists.parrot.org/pipermail/parrot-d...04582.html 21:04
chromatic Let's move on then. 21:05
Other thoughts on schedule or niceties or prioritization?
NotFound AFAIK Op and packfile PCB are used only in examples and in pirate, none of wich is a blocker for anything. 21:06
I mean PMCs
chromatic Too bad, I wanted Parrot hardware.
darbelo Also, the Op PMCs are experimental. No API concerns, there. 21:07
NotFound So I'll vote for marge and solve the remaining issues in trunk. 21:08
chromatic Coke, can you summarize the state of your branch somewhere? 21:10
Can someone get plobsing to do the same?
NotFound plobsing was active this week, he even sent a patch for winxed. 21:11
chromatic He's the one shepherding dynop mapping. 21:13
Anything else to discuss?
smash chromatic: can i just make a notice ?
Tene I might be able to merge my exceptions work sometime. 21:14
chromatic Go ahead, smash.
smash just like to remind every that that the call for nominations for this years election for board of directors of the parrot foundation is opne unitl next tuesday 21:15
more info via e-mail on parrot-members if my mail make i thorugh 21:16
chromatic: that's it, thank you
Util Paul_the_Greek had a question about bringing attention to a ticket?
Paul_the_Greek I have two questions.
chromatic Go ahead.
Paul_the_Greek First: Does anyone have a source for Windows binaries for GMP 4.1.4 or later?
We seem to have a dependency on unobtanium. 21:17
I spent two hours today with no luck. Got sources, but can't build them.
Util I do not, but I will investigate creating one with MinGW, which was used on GMP 4.1
21:17 dukeleto joined
Paul_the_Greek Yes, MinGW is what I'm using. 21:18
Thanks.
Okay, here is the ticket. If I could get a few comments over the next few days, that would be helpful.
trac.parrot.org/parrot/ticket/1731
No hurry, I'll be at the beach for two weeks.
dukeleto waves hello 21:19
mikehh hi dukeleto
chromatic Anything else? 21:20
Paul_the_Greek Util: I found binaries for 4.1.3, but Parrot requires 4.1.4 or later.
mikehh yeah smolder is still down, any ideas?
cotto_work Coke said osuosl would put smolder on parrot.org's vm 21:21
mikehh particle said something about installing it om parrot.org, but I don't know if anything has happened there 21:22
dukeleto cotto_work: one thing to note is that smolder has memory leaks, so it might not be good to run it on the same vm. perhaps if it has hard ulimits
cotto_work classy
dukeleto cotto_work: yeah, something to note. I think it is memory leaks in the dependencies of Smolder, i am not sure if it is Smolder itself that has the leaks 21:23
mikehh smolder or something like that is important to check for other platform failures
taptinder is useful. but we need something more 21:24
dukeleto q1q
chromatic Let's try to get a status report from particle on that. 21:25
dukeleto, question? 21:26
dukeleto what is the proper place to talk about board nominations? 21:27
chromatic Not sure anyone knows. 21:29
dukeleto ok, well all i have to say is that i would appreciate if somebody would nominate me for the board, because I am interested in being on it 21:30
dukeleto is much obliged 21:31
chromatic anything else? 21:33
darbelo Last week of GSoC. If you need something from your student, get it now. 21:34
Paul_the_Greek Do people announce vacations etc. via the list, on calendar, or not at all?
chromatic Are students on track to succeed?
mikehh I'll run tests on the branches in the next couple of days 21:35
Paul_the_Greek Not that I'm on any critical path.
Coke Paul_the_Greek: not that I've ever seen.
cotto_work khairul is looking good
Coke if you're collaborating with someone, no doubt they'll mention it.
chromatic tcurtis will succeed.
whiteknight Chandon is on course
darbelo still feels hope... 21:36
bubaflub laughs nervously 21:37
chromatic Let's wrap this up then. I mailed the list about branch priorities. Let's do it! 21:38
dukeleto thinks most students are on the track to pass GSoC, will know more in the next few days 21:40
21:40 dukeleto left 21:43 bubaflub left 21:57 darbelo left 23:00 Paul_the_Greek left 23:34 NotFound left 23:35 atrodo left