Priority for this week: Schedule PDS, Find Rakudo slowdown, GSoC Mentoring | Post closed tickets in your report. | Note: This channel is for our Tuesday status meetings (at 20:30 UTC); you probably want #parrot instead. | irclog: irclog.perlgeek.de/
Set by moderator on 7 May 2011.
00:08 darbelo left 01:43 whiteknight left 03:06 contingencyplan left 03:07 contingencyplan joined 06:59 ShaneC left 09:11 mikehh left 09:44 contingencyplan left 11:50 mikehh joined 11:54 darbelo joined 12:50 bluescreen joined 13:55 Coke left 14:07 bluescreen left 14:23 bluescreen joined 14:53 bluescreen left 15:07 bluescreen joined 15:32 darbelo_ joined, darbelo left 15:33 darbelo joined, darbelo_ left 15:38 contingencyplan joined 17:16 darbelo_ joined, darbelo left 17:39 pmichaud joined 17:45 darbelo_ left 17:54 atrodo joined 17:55 lucian joined
pmichaud What I did recently: 18:19
* Migrated rakudo.org from Drupal to new home with WordPress
- aka rakudoperl.org
- We once again have working RSS feeds!
* Built a benchmark / regression suite for Rakudo+Parrot
- github.com/pmichaud/rakbench
- See github.com/pmichaud/rakbench/tree/...er/results for
example results
* Did lots of timing comparisons among Rakudo releases
* Located a regression in Parrot's ms2 GC that was causing significant
(50%+) slowdowns since 2011.01
* Summary: I'm not seeing significant overall performance improvements since 2011.01
- we are seeing good improvements in some cases for core.pm and rx.t
* Tested various patches and tuning suggestions from #parrot
* Got a new desktop machine (kiwi); can now run Rakudo spectests in < 5 mins
What I plan to do:
* More benchmarking and testing
* Minor improvements to Rakudo build system
EOR
18:47 particle left 18:54 particle joined 19:02 darbelo joined 19:13 atrodo is now known as atrodo_ 19:14 atrodo_ is now known as atrodo, lucian_ joined 19:18 lucian left
mikehh What I did since my last report: 19:19
* building and testing parrot on amd64/i386, with gcc/g++
* some fixes
* building and testing rakudo and winxed against latest parrot
* trying to track down some regressions regarding rakudo timings
What I intend to do in the next week:
* testing and fixing
.eor
19:47 kid51 joined
kid51 kid51's report 19:48
* Am on vacation, jetlagged, won't be at psketch today.
* Dealt with Smolder/Trac problems with OSU OSL
* Advise everyone to click on the "Forgot Your Password" link and get a new Trac password.
* Testing: As usual, uncovered GC-related problems when testing code on Darwin/PPC. bacek++ has solved some of these. 19:49
*EOR
19:50 kid51 left, plobsing joined
tcurtis What I did: not much; scav hunt interfered. What I plan: give lightning talk on GSoC project at the Chicago GSoC meetup Thursday. More paper reading. More blogging. 19:52
plobsing What I DId: 19:54
* merge tt1931 branch
* work on prototype of locked GC (no results yet) 19:55
What I Plan:
* get a proof of concept locked GC up and running, benchmark costs of this strategy
Blockers:
* trac won't let me fill in the deprecation data for TT #1931
EOR
19:58 NotFound joined 19:59 bubaflub joined
atrodo .done 20:01
* Got more ops->lasm code written, but nothing stable yet
.todo
* isparrotfastyet is going to get some love this week. Watch for it
.end
20:06 ShaneC joined
Util # Nothing done; traveling. 20:07
7-day ticket report: 2 new.
.end
20:07 whiteknight joined 20:08 benabik joined, dukeleto joined
NotFound What I did: 20:15
-parrot
* Minor fixes.
* Improve a few error messages.
-winxed
* Use 'using extern' for module imports.
* Improve stage 0 prefix ++ and -- operators.
* Allow 'null' in conditional operator.
* Look for functions in accesible namespace, avoid the need of
'using' and 'using static' in most cases.
* Several fixes, cleanups and refactors.
* Restarted work in winxedxx. Now is able to run some non-trivial programs.
What I wiil do:
Probably more work in winxed scope lookups and in winxedxx.
EOR
dukeleto What I did:
* Some GSoCy stuff (one of our students has gone missing already)
* M0 work on the m0-spec and m0-prototype branches
* Some grant work on leto/embed_grant branch
* In the process of adding PL/Parrot to PGXN.org
What I will do:
* More of the same
Blockers:
* Lack of infinite time
.EOR
NotFound: can you define winxedxx? 20:16
whiteknight WHAT I DID
* Helping to track down the Rakudo performance regressions. Thanks to everybody involved for the continued efforts
* Talking about threading/concurrency with plobsing++ et al. Lots of good ideas floating around
* Digging into 6model code. I have a better understanding of it now and a huge list of questions for jnthn++
* Looking hard at possible refactors and improvements to PCC. It is appearing to be a big performance drain.
* Did a few timings/benchmarks for vtable overrides vs methods vs built-in vtables. The results are bad (and are posted on my blog).
* Starting to learn Ruby.
* Keeping up with GSoC.
WHAT I WILL DO
* Follow along with jnthn's 6model/Rakudo work. Continue planning that work.
* Keep poking around with PCC. I think there is plenty of gold to mine there, if we are willing to go after it.
EOR
NotFound winxedxx is a backend for winxed that generates c++
20:17 soh_cah_toa joined
bubaflub ETOOMANYFINALS 20:17
20:18 luben joined
benabik ETOOMANYFINALS 20:19
soh_cah_toa WEEKLY REPORT
DONE:
* Started separate GSoC blog
TODO:
* Mirror blog on parrot.org
EOR
dukeleto soh_cah_toa++ 20:22
soh_cah_toa dukeleto: yeah, i initially created it so i wouldn't pollute the parrot.org blog w/ gsoc stuff but then cotto told me you actually wanted it that way. i'm just gonna mirror the two so i didn't waste my entire weekend 20:23
20:25 Coke joined
luben What I did: 20:26
* benchmarked parrot GC GMS on i386 (RAM) amd64 (4G RAM)
* picked new default parameters for it:
- nursery-size=2% of sysmem
- 4 generations are enough
.END
cotto_work *did: 20:28
- M0 progress:
-- hacked out some tests for the M0 interp once the assembler is working
-- thinking
- GSoC progress:
*eor 20:29
hello 20:30
20:31 whiteknight left
bubaflub hello 20:31
dukeleto greetings 20:32
mikehh hi there
benabik Hi. (Note: I only semi-exist, working on project.)
tcurtis hi.
cotto_work How'd we do this week?
NotFound Hola 20:33
soh_cah_toa hello
cotto_work The two goals (scheduling PDS and digging into the Rakudo slowdown) looked like they got some eyeballs.
soh_cah_toa pds is this saturday, right? 20:35
mikehh trying to track things down inn rakudo is exhausting - takes far too long to build and test
cotto_work soh_cah_toa: yes, depending on your timezone
www.doodle.com/nf5cip2yrt6a448d
soh_cah_toa ok
cotto_work any questions or comments on the past week? 20:38
mikehh we shiuld set up a codetest (maybe without perlcritic) that needs to be run before commits 20:40
IOW perlcritic tests need only be run with perl changes but the rest needs to be run before committing
dukeleto mikehh: we can use git hooks for that 20:41
cotto_work dukeleto: can jitterbug help us here? 20:42
20:42 coke_ joined
plobsing we could push to a temp/staging branch pending testing and automated testing would move over to master (and would be the only thing with perms) 20:43
soh_cah_toa that's a good idea
cotto_work plobsing: That doesn't sound too bad if we can make sure that there's not a major change to the typical dev workflow. 20:44
dukeleto jitterbug already does something like this 20:45
cotto_work Though it brings up the question about our philosophy of what "master" means.
dukeleto cotto_work: indeed.
mikehh: can you send an email to parrot-dev explaining exactly which tests need to be run before each commit? 20:46
mikehh: then we can discuss turning that into a git hook
mikehh dukeleto: 'k, will compose it later probably after I get some sleep 20:47
cotto_work If we want master to be safe, we should test code that goes into it more frequently. 20:49
If we look at master as the place where we integrate changes that are (or will be) a part of Parrot proper, making it's always passing all test is less important.
Of course, it's important to get our tests passing quickly. 20:50
plobsing I think there's a case to be made for having regularly checkpointed known-good states (eg: for bisecting), but I don't think we should necessarily enforce that on all commits. 20:51
mikehh I always try to run at least codetest and test after making changes to master before pushing
dukeleto my idea of master is code where the test suite passes on at least one platform, and hopefully many
cotto_work My inclination is to say that occasional test failures aren't a big deal.
dukeleto cotto_work: yes, master actually shakes out many subtle bugs. Breaking master, taking into account our release cycle, has worked well for us, so far. 20:52
mikehh it should pass tests at least on your own system
plobsing if master is always not broken, does that mean I can't locally commit breakage and then fix them in subsequent commits?
dukeleto i think we need to think about how to use tools like Jitterbug and BuildBot to test Parrot more
we have access to the GCC compile farm and the OSL Supercell, but we aren't utilizing them yet, mostly because I am the only person on the bus and I can't seem to make the time to work on that stuff 20:53
plobsing: we prefer master to be bisectable, if we can 20:54
cotto_work (I don't want anyone to interpret me as saying that we should break tests on purpose, just that if breaks occasionally slip through and get fixed, it's acceptable.)
mikehh TapToinder often catches windows bugs that are missed in linux
TapTinder
cotto_work dukeleto: indeed. We need a volunteer with some tuits to make stuff happen.
dukeleto I've heard that Git may add some metadata soon where one case say "ignore this commit when bisecting", but that is still unicorn farts, for now. 20:55
cotto_work: did I just hear you volunteer to find a volunteer?
NotFound I recommend (again) paying attention to warnings that appears in changed or new code.
cotto_work dukeleto: in the sense that I'm asking for one now, yes 20:56
plobsing I think that enforcing bisectability requirements on master may be too stiffling. I'd suggest a separate branch with merges from master whenever the automated tools detect an acceptable state.
dukeleto Some projects have tests that fail when warnings are introduced, and some projects compile with -Werror (such as Drizzle) 20:57
cotto_work drafting a parrot-dev message now
dukeleto Using -Werror is quite drastic on one hand, but imagine how much time it saves by not having to write, maintain and run tests about warnings! 20:58
NotFound Using -Werror is tricky when using compile checks for compiler features or headers. 20:59
dukeleto But yes, it will be considered stifling by some. But in the age of local git branches, is it really that stifling?
NotFound It usually gives wrong negatives.
coke_ NotFound: as long as those tests are done before auto/warnings, should be fine.
dukeleto NotFound: well, I am sure we can borrow many tuits from Drizzle, which uses -Werror
plobsing dukeleto: warnings are indications of a guarrantee that can't be made by static analysis. If your static analyser can't make a guarrantee, you should make a better one or remain silent.
dukeleto plobsing: and?
plobsing: I am only saying that it would be nice to discuss on parrot-dev some of our current testing strategies and whether we should modify some of them 21:00
plobsing: -Werror is just an example
plobsing that's my take on compiler warnings. if they could guarrantee that the code did indeed fall into the suspected pattern, it would be an error. but they cannot. so they should either work on their analysis, or stop trying. 21:01
dukeleto plobsing: we could also just have Jitterbug compile Parrot with -Werror and notify us
cotto_work They're hints that something may be wrong.
NotFound A configure option to add C flags for compile time mayl be nice. 21:02
dukeleto NotFound: please make a TT. Perhaps one of our gsoc students would like to work on that...
plobsing the key word being *may*. the issue is raised to the attention of a human capable of assessing that risk. then an informed decision about whether or not there is a problem can be made.
NotFound Failed to parse. 21:03
cotto_work dukeleto: do you have any idea what kind of time commitment would be needed to get SuperCell and the GCC Compile Farm (1) working and (2) maintained day-to-day? 21:04
coke_ NotFound: Don't we already have a configure option for adding c flags?
NotFound coke_: AFAIK is for setting, not for adding.
And is used during configure. I mean one to be used only during build. 21:05
Thus avoiding problems with compiler probes. 21:06
mikehh BTW I still can't access smolder, anything happening there? 21:12
soh_cah_toa mikehh: you me signing in or sending reports? b/c sending worked fine for me a few days ago 21:17
*mean
coke_ I can restart smolder. momentito.
cotto_work Was smolder what caused trac to explode? 21:18
mikehh it needs restarting on a regular basis, at least once a week, maybe more often
cotto_work Yes. It's a great motivator for jitterbug. 21:19
coke_ is not a big fan of continual tool swappage. 21:20
better to actually fix the tools we use at some point.
moving smolder ressurection to #parrot.
cotto_work good idea
mikehh I tried to set up a local smolder server, but there were way too many dependencies there, will probably try again 21:21
cotto_work I'll make sure to send a request for testing infratructure volunteers to parrot-dev. Are there any other questions? 21:22
21:23 coke_ is now known as Freenode_Coke, Freenode_Coke left
dukeleto cotto_work: someone other than me needs to apply to access to both, and then we can both work on documentation for Parrot's stuff on each. Start up time could vary, but once things are set up, I expect both would take less than 5 hours per week, possibly much less 21:29
cotto_work dukeleto: thanks 21:30
dukeleto cotto_work: i am still interested in both, but I don't have enough tuits to spend, by myself, to get those things on their way
21:32 Coke_ joined
cotto_work request for volunteer sent 21:32
any goals for the week?
mikehh when does GSoC start in earnest? 21:33
NotFound Is soemone looking for the opengl recent problems?
benabik mikehh: Official start date is 5/24
mikehh benabik: 'k - prep until then 21:34
cotto_work and bonding
and coding is fine too
lucian_ is very late 21:36
21:37 lucian_ is now known as lucian
cotto_work here's the goal I have: 21:38
GOAL 1: Find someone to lead testing infrastrcture planning
dukeleto cotto_work: i can design a good infrastructure, but I don't currently have to the time to implement it all
cotto_work: i can lead the planning, but I need people to implement it, manage it, set it up, etc...
cotto_work dukeleto: can you send that to parrot-dev (or put it on the wiki and link from parrot-dev)? 21:39
dukeleto cotto_work: nothing is automated enough for me to do it myself, so it goes in a vicious cycle
cotto_work: what do you want me to send?
cotto_work your plan
dukeleto 1) We need someone to help migrate Smolder to a VM on Supercell.
cotto_work: ok, will create a wiki page and send an email to parrot-dev 21:40
cotto_work What need is there for migration? All we need to to is set it up and point make smolder at it.
nm. configuration
dukeleto cotto_work: a simple matter of "making it work", which is sadisticly humorous when it comes to setting up Smolder 21:41
Coke_ be sure to cc mpeters on the request for volunteers.
cotto_work is he not on parrot-dev 21:45
?
Coke_ even if he is, cc'ing makes it obvious hint-hint style that we would like his help. or you could just ask him outright, since smolder is his baby.
cotto_work ok. Last call for further comments? 21:52
dukeleto gsoc students 21:53
cotto_work what about them?
dukeleto have all the mentors talked to their gsoc students recently?
cotto_work has
dukeleto if everyone says yes, than someone is lying
cotto_work who's mia?
dukeleto because cgaertner has gone missing
no contact since proposals were accepted
if anybody knows any way to contact him, please let him know he should check in 21:54
cotto_work has someone tried email?
dukeleto cotto_work: whiteknight++ has been trying to get ahold of him for weeks via email, iirc 21:55
as a general note, I still plan to contact all gsoc studens personally, via skype or phone, but life has gotten in the way recently.
bubaflub: we need to skype or phone talk and just touch base 21:56
same goes for all other GSoC students. Please pester me more :)
bubaflub dukeleto: sure thing. i'll shoot you an email with my schedule and we'll work something out
soh_cah_toa i still remembered
cotto_work dukeleto: ok. It's not technically a cause for concern as long as he's around by the 24th, but still lta.
bacek waves with "I hate DST" flag 21:57
dukeleto cotto_work: i would say it is definitely cause for concern now, but not cause for failure until the 24th
bacek: welcome, meatbag
bacek: please regale us with your stories
bacek: have you talked with your gsoc student benabik ? 21:58
bacek dukeleto, he is quite busy with exams atm
no parrot-related activity yet
dukeleto bacek: ok, as long as you have talked with him
cotto_work dukeleto: that's a better way to phrase it 21:59
bacek sure
dukeleto for all we know, cgaertner is in a 30ft deep final exam hole and has unplugged his computer. But he should have given us a ping about it.
this doesn't bode well for communication for the rest of the summer
21:59 bluescreen left 22:00 bubaflub left, bubaflub joined 22:01 bubaflub left
dukeleto more questions? 22:01
cotto_work GOAL 2: Contact all GSoC students 22:03
Sound like there aren't any. Let's call it a wrap. 22:04
dukeleto self.wrap() 22:08
22:08 dukeleto left
moderator Priorities for this week: Plan our Testing Infrastructure (choose lead person), Contact all GSoC students | Post closed tickets in your report. | Note: This channel is for our Tuesday status meetings (at 20:30 UTC); you probably want #parrot instead. | irclog: irclog.perlgeek.de/ 22:09
22:10 NotFound left 22:25 benabik left 22:26 plobsing left 22:30 soh_cah_toa left, ShaneC left 22:32 kid51 joined 22:38 bubaflub joined 22:48 luben left 22:52 kid51 left 22:54 lucian left 23:11 whiteknight joined 23:35 atrodo is now known as atrodo_, atrodo_ is now known as atrodo 23:55 darbelo 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!