Parrot 3.4.0 released | parrot.org | Log: irclog.perlgeek.de/parrot/today
Set by moderator on 17 May 2011.
ttbot Parrot 1b290a46 cygwin-thread-multi-64int make error tt.taptinder.org/cmdinfo/11376 00:03
dalek rrot/leto/embed_grant: cb718b1 | dukeleto++ | t/src/extend_vtable.t:
[t] Parrot_PMC_set_string_keyed_str
 00:05
dukeleto soh_cah_toa: wazzup 00:07
soh_cah_toa dukeleto: gsoc work
dukeleto: i've got the runcore working now i'm just searching through this massive api for things that may be helpful 00:08
bubaflub soh_cah_toa++ 00:09
soh_cah_toa fortunately, i just discovered the amazing power for 'git grep'
dalek rrot/leto/embed_grant: 3047227 | dukeleto++ | t/src/extend_vtable.t:
[t] Parrot_PMC_set_string_keyed_int
 00:15
kid51 soh_cah_toa: Consider adding something to your weekly blog about 'git grep', e.g., how it differs from 'grep' or 'ack'. 00:17
soh_cah_toa kid51: sure
hey, does anybody know if you have to register/pay to get into yapc::na? 00:25
dukeleto soh_cah_toa: yes 00:26
soh_cah_toa: but, as far as confs go, it is very cheap. I think $100 or something like that
dalek rrot: 7633217 | jkeenan++ | config/gen/makefiles/root.in:
Revert "Apply patch submitted in trac.parrot.org/parrot/ticket/540.

This reverts commit 1b290a4658a8d2037855cbeed57b09b847a163f7. We will need to diagnose a build error on Cygwin reported here:
  tt.taptinder.org/cmdinfo/11376.		 00:27
soh_cah_toa dukeleto: wow, i'm glad i asked. that could've been bad...
dukeleto soh_cah_toa: submitting a talk gets you in free :) 00:33
soh_cah_toa dukeleto: lucky you :) 00:35
dalek TT #540 reopened by jkeenan++: installed versions of dynext/*.so still link to -lparrot in build ... 00:37
TT #540: trac.parrot.org/parrot/ticket/540
kid51 soh_cah_toa: Travel and room costs are always the significant parts of attending YAPC. But even those are less than other conferences. 00:39
soh_cah_toa: Where are you located, terrestrially speaking?
soh_cah_toa kid51: new jersey. yapc::na will be about a 13 hour drive for me 00:40
kid51 soh_cah_toa: Consider also FOSSCON in Philadelphia Sat July 23. 00:41
I'm hoping pmichaud and whiteknight will attend.
soh_cah_toa kid51: i'll be there too
kid51 dukeleto: Is it now the case that we have no information as to what git commit we are at until parrot has been built? 01:05
i.e., we no longer have that information as a result of Configure.pl?
01:37 AzureSto_ left
dalek rrot: 462c4ce | petdance++ | include/parrot/compiler.h:
remove the /*@null@*/ from SHIM args
 01:42
02:14 kid51 left 02:18 AzureStone joined 02:23 ShaneC joined 02:24 soh_cah_toa left
cotto ~ 04:22
04:36 fperrad joined
dukeleto msg kid51 cotto changed some stuff regarding sha1 caching, but you can always just do "git rev-parse HEAD" to get the current sha1 04:36
aloha OK. I'll deliver the message.
04:49 bubaflub left
cotto sleeps 05:09
05:17 theory left
dalek rrot: fe2ae81 | petdance++ | src/gc/alloc_resources.c:
make the loop var make what it is comparing to
 05:24
rrot: 121807c | dukeleto++ | t/src/extend_vtable.t:
Merge branch 'leto/embed_grant'
 06:18
06:46 mj41 joined 07:35 SHODAN joined 08:00 slavorgn left 08:05 dod joined 09:38 PacoLinux left
bacek ~~ 10:17
10:17 daniel-s joined
bacek msg benabik you have to implement nqp-setting bootstrapping. E.g. commit generated nqp-setting.pir into repo. Otherwise we'll have circular dependencies nqp->pct->nqp. 10:18
aloha OK. I'll deliver the message.
bacek seen benabik
aloha benabik was last seen in #perl6 3 hours 8 mins ago saying "sorear: Ah. Less useful then, although that is the reason I use git-archive. :-D".
bacek msg benabik Simplest solution is to generate nqp-setting.pir in C<master> branch, copy it across to <nqp_pct> branch and fix makefiles. I can update upstream nqp-rx after. 10:20
aloha OK. I'll deliver the message.
dalek rrot/gc_tuning: bee51f5 | bacek++ | src/gc/gc_gms.c:
Disable manual freeing of pmc/strings. It brakes invariant about single copy of header in list. And we can't remove it from objects/strings list cheaply
 10:33
10:37 daniel-s left
bacek seen pmichaud 10:42
aloha pmichaud was last seen in #parrot 3 days 9 hours ago saying "I'm able to access trac now".
bacek sigh. 10:43
msg pmichaud if you'll have time can you rebenchmark gc_tuneup branch? And I have idea how to improve performance of PCT in gms
aloha OK. I'll deliver the message.
bacek msg pmichaud Basically adding Capture.ro_list and .ro_hash will reduce number of WBs by big huge number. 10:45
aloha OK. I'll deliver the message.
11:02 Psyche^ joined 11:07 Patterner left, Psyche^ is now known as Patterner 11:41 SHODAN left 11:47 mtk left 11:52 ambs joined 11:53 mtk joined 11:54 lucian joined 12:18 bacek left 12:22 bluescreen joined 12:47 SHODAN joined 12:54 bubaflub joined
bubaflub ~~ 13:01
13:05 daniel-s joined 13:42 hercynium left 14:05 whiteknight joined 14:07 PacoLinux joined
whiteknight good morning, #parrot 14:12
sorear hello whiteknight 14:17
whiteknight hello sorear 14:18
14:22 hercynium joined, dod left
Coke_ . 14:26
14:33 Andy joined
Andy is downloading the latest GCC 4.7 snapshot, with the -Wmaybe-uninitialized flag 14:37
14:39 ambs_ joined, ambs left, ambs_ is now known as ambs
Coke_ or maybe-not. 14:47
Andy gcc does take a little while to build. 14:55
chugga chugga
NotFound Have you seen my last candy? winxed.org/misc/testwinxedgtk1.png
cotto ~
NotFound Gtk via NCI. 14:56
cotto NotFound++
14:58 theory joined 15:00 SHODAN left 15:03 dod joined 15:04 daniel-s left
xenoterracide particle1: prod 15:05
whiteknight NotFound++ 15:07
gtk bindings would be quite awesome indeed
if only we had good GObject bindings to go with it...
lucian yeah, gobject-introspection seems to be the way to go 15:09
NotFound I don't care about that for a now, I'm just working in easisly doing GUIs.
lucian i'd expect using g-i to be easier than a hand-written binding 15:10
NotFound Maybe, but that one works now. 15:11
lucian sure, you're getting results 15:13
and i STILL have to revise for my stupid exams
NotFound I want the posibility of writing GUIs right now, don't care if the engine gets obsolted soon. 15:16
whiteknight I would like to be able to add some GUI Views to some Rosella libraries 15:17
a nice GUI for the REPL would be fun
NotFound Is not too much effort to bind the most used widgets.
whiteknight: What is exactly the REPL? You haven't write about it, or I failed to see it. 15:18
whiteknight oh, it's just a prototype right now. I can't write any posts because I have no internet access at home
github.com/Whiteknight/Rosella/blo...epl.winxed 15:19
that's the frontend to it
github.com/Whiteknight/Rosella/blo...epl.winxed
and that's the main library file
NotFound What is it? Some sort of pluggable perser? 15:24
whiteknight oh, it's like a command-line interface. You write one line and eval it
the input comes from the commandline instead of a file
NotFound Oh, nice.
whiteknight What I need to do is override some parts of the winxed compiler to save variables between lines, and search for variables in a different way, but that will happen later 15:25
(That's why I want Winxed compiler to use a namespace for it's code, so I can inject my own things into that namespace) 15:26
lucian that sounds like incremental compilation
NotFound Sounds good. I'm also thinking about allowing plugins or optional parts in the winxed compiler, for the benefit of winxedxx.
cotto_work ~~
NotFound whiteknight: That way probably will not work well because of static binding. 15:27
whiteknight NotFound: It's not a big deal. I have a state object that the user can store values into for now, and I added an ability to add a common prefix. I'm thinking about changing the code generator to force each new line to use the previous one as an :outer 15:30
that could make lookups costly, but I don't expect that to be a big problem
NotFound The variables can also be hard, given the not-so-lexical oriented winxed way. 15:31
whiteknight is there a way I can force a variable to be stored lexically, or hint to the compiler that they should be saved?
blah, nevermind. I'm getting too complicated right now
NotFound whiteknight: I can probably add something, but that will only work for var, not for int, float or string. 15:32
whiteknight yeah, no big deal. Don't worry about it
you have too many other fun projects to work on first :)
NotFound Ars longa, vita brevis. 15:33
The main problem I see in NCI right now is that callbacks don't return a result. 15:34
They use a void function, ignoring the result type in the siganture. 15:35
That is a big problem for some Gtk signals. 15:36
whiteknight We know that callbacks don't do enough stuff. We've known that for years 15:37
we might be able to hack in return values for it. Can you open a ticket?
NotFound I'll wait until I publish something of that module, to have something to work on. 15:38
cotto_work dukeleto: ping 16:23
lucian allison: ping 16:25
dukeleto cotto_work: pong
cotto_work dukeleto: privmsg 16:27
whiteknight dukeleto ping and also privmsg 16:33
dukeleto is being attacked from all sides! 16:34
cotto_work oh noes
In generated code, how common are non-constant gotos? 16:35
dukeleto cotto_work: good question. Which code? PIR, PASM, NQP ? 16:36
cotto_work as a general principle
whiteknight I would say highly uncommon
cotto_work it applies to PIR, but I'm also wondering about actual machine code too
I'm thinking that M0 might benefit if goto statements can only target fixed locations. 16:37
dukeleto: how do you feel about git config --global branch.autosetuprebase always 16:38
Is that something we should recommend? 16:39
NotFound Winxed does not use non-costant gotos, unless somone is using some trick I'm not aware of.
dukeleto cotto_work: i think i know what that does. tell me why you want it 16:41
cotto_work dukeleto: avoiding merge commits 16:42
lazily
dukeleto cotto_work: the problem with that command is that it makes it default for all git repos on the system, not just parrot.git 16:43
cotto_work: that is the --global part 16:44
cotto_work dukeleto: ok. We shouldn't recommend global settings changes without noting them as such.
dukeleto cotto_work: but I think telling people about the autosetuprebase in our docs would be a good middle ground 16:45
cotto_work dukeleto: ok 16:47
whiteknight: thanks. I'll consider it safe to only have constant gotos in M0. If it turns out to be a bad idea, nothing's set in stone yet. 16:49
whiteknight cotto: For everything else, we have Continuations
cotto_work When I get home, I should have a copy of Appel's Compiling with Continuations waiting for me. I expect a very nerdy evening. 16:52
whiteknight nerd on! 16:54
16:58 mj41 left
benabik cotto_work: Awesome. I need to get my hands on that. 17:00
whiteknight yes, that's definitely on my Amazon wish list 17:01
cotto_work benabik: It might be popular. I ordered it for delivery on Friday but it didn't get shipped until Saturday.
lta, but I got free shipping for the inconvenience. 17:02
whiteknight free shipping sounds pretty awesome 17:05
cotto_work something about this book appeals to me: www.amazon.com/Let-Over-Lambda-Doug...rhf_shvl_1
benabik In my functional class last quarter, I finally realized why I don't like Lisp. Everything looks too alike. Can't distinguish math from functions from data. 17:06
cotto_work benabik: +1. If every want to get serious about a lisp-family language, that's the first big obstacle I'll have to get over. 17:07
The second will be an urge to clip nails.
benabik cotto_work: Strangely enough, Lisp's S-Expressions weren't intended to be the main way to program it. There was a semi-Haskell like format called M-Expressions. 17:08
whiteknight Do we have any motivated students interesting writing a quicksort or heapsort implementation in winxed? 17:24
dukeleto whiteknight: i think emailing parrot-users/parrot-dev would be a good place for that question
whiteknight I was planning to do something like that over the weekend, but ran out of time 17:25
17:26 ligne joined
whiteknight cotto_work: does M0 allow you to specify register set by type? I've run into a few problems recently where a problem could be solved much better if we could specify PISN register types at runtime instead of at compile time 17:26
sorting is a perfect example. If we could write one sort routine and have it work for all primitive value types without autoboxing all the intermediate results, that would be killer 17:27
if I want to sort a FixedIntegerArray, I can pass it to the sort routine with an "I" register set selector, and have it do native integer sorts 17:28
dukeleto whiteknight: what do you mean by "specify register types"
whiteknight: give an example workflow of what you would like to be able to do
whiteknight dukeleto: So instead of having "$I0 = $I1 + $I2", I could have "$X0 = $X1 + $X2" and be able to specify X at runtime 17:29
so if I wanted, that becomes a PMC operation, or a Num operation, or a String operation
dukeleto whiteknight: sounds macro-ish 17:30
whiteknight: and M0 doesn't have macros
whiteknight: but i see why you would want such a thing
whiteknight yeah, macro is not a bad way to think about it
C++ calls them templates, C# calls them Generics
dukeleto we don't really intend for anybody to every manually write M0, except during bootstrapping stages 17:31
whiteknight right, but if M0 supports those kinds of operations (PIR currently does not), we can build upon them for higher-order languages
it's a small request, I was just wondering if it had been given any thought
cotto_work whiteknight: that sounds like a higher level of magic than 0. It's worth thinking about how M0 could make that kind of metaprogramming possible though. 17:35
whiteknight like I said, it's a small request. Would make several things much much easier if we could specify the target register set to use 17:36
or, dare I say, even provide a custom or subclassed register set
NotFound whiteknight: small requests like that can compromise a full architecture. 17:39
17:40 davidfetter joined
whiteknight NotFound: true :) 17:40
I actually don't think that would be too hard to implement. If the first 3 bits of every register were the type tag, we could look up which register to use for each operation 17:42
so if type tag 000 was "unknown", we would look up the default tag from the current context and use that
then we are able to write code once, and parameterize it at runtime 17:43
dukeleto whiteknight: i think what you want could be implemented at the M1 level 17:44
whiteknight is M0 so flexible as to allow those kinds of shenanigans transparently? 17:45
or would M1 have to generate multiple versions of static M0 code?
dukeleto whiteknight: the latter
whiteknight: generated code doesn't have to be pretty
whiteknight Well, I can live with that
cotto_work There aren't any free bits in opcodes as-is
NotFound whiteknight: I think we'll be better served with higher level ways for generating and parametrizing code. If we want a low level machine, it should be really low level.
cotto_work (M0 opcodes)
actually, there are a few 17:46
probably not enough for what you want though
whiteknight I'm mostly thinking of a way to deduplicate identical code where the only difference is to operate exactly the same on multiple low-level register sets without autoboxing
NotFound whiteknight: And then reduplicate it in the jit?
whiteknight .NET does it transparently with type generics 17:47
a specializing JIT is already going to duplicate a lot of code
NotFound whiteknight: define "it"
whiteknight NotFound: .NET's JIT works just fine with their generics implementation
NotFound whiteknight: fine, but I doubt cli code has a high similiude with the M0 ideas. 17:48
whiteknight that's very possible 17:49
notably, type genericism is really not needed above the register set level, because PVM is all dynamic
NotFound Id on't doubt that you can jit anything, I doubt the benefit of decuplicating code to later reduplicate it.
whiteknight Rakudo provides it with 6model though, so we may still want something
dalek rrot: f9b5a34 | (Matt Boyle)++ | docs/book/pir/ch0 (2 files):
[docs] Fix runaway PIR_FRAGMENT regions.
 17:51
rrot: c5d1909 | dukeleto++ | docs/book/pir/ch0 (2 files):
Merge pull request #131 from ligne/book_runaway_pod

  [docs] Fix runaway PIR_FRAGMENT regions.
NotFound whiteknight: you are mixing high level objects with very low level registers.
dukeleto loves the merge button
benabik msg bacek Thanks for the direction. Not familiar enough with the build yet to recognize that circularity. Bootstrapped version now on github in benabik/parrot/nqp_pct 17:52
aloha OK. I'll deliver the message.
whiteknight NotFound: yes. there are two types of metaprogramming. I would like to see both
We can live with just the high-level types, but a low-level version would be nice too, I think
ligne is trac still undergoing breakage? i'm not authorised to create new tickets, even after trying the reset-password trick.
woah, that was quick. thanks dukeleto :-) 17:53
whiteknight ligne: We disabled ticket creation for most users because of heavy spam problems
17:53 dmalcolm joined
whiteknight We should probably un-disable that 17:53
dukeleto ligne: we may need to give you that bit. trac is not smart about protecting itself against spammers
ligne: keep them coming!
NotFound whiteknight: A low-level, maybe, but not so low as M0 IMHO.
dukeleto trac-- needs to learn how to captcha or something
ligne yeah, it's a shame. trac is really quite nice, but a bit simple 17:54
whiteknight I really don't like trac at all. It tries to do a lot of stuff but ends up not doing any thing particularly well 17:55
cotto_work For good or ill, it's not so bad that it's worth migrating away from. 17:56
dukeleto we could just install some plugins to our trac instance and make it not suck as much 17:57
we really need a captcha plugin
anybody want to research that?
whiteknight cotto_work: it's not so good that I use it as much as I probably should
benabik My experience with phpBB is that captchas don't help too much. :-( 17:58
whiteknight no, captchas are not a perfect solution.
dukeleto benabik: having a captcha is a barrier
and there are many kinds of captchas. Classic cat + mouse game.
but, alas, something is better than nothing. 17:59
whiteknight depends a lot of the strength of the captcha, but there are other things we could implement as well to reduce the use of bots
at least, the use of generalized bots.
benabik dukeleto: Text-based questions seem to keep the spammers away better than captchas.
dukeleto benabik: people still use phpBB ?
NotFound A sufficent annoying captcha is worse than moderating.
whiteknight benabik: IBM Watson is in ur site, decipherin' yer questions
ligne full ack about the lack of a good spam-trap. i had to delete about a dozen spam tickets from another trac the other day. not happy. 18:00
dukeleto benabik: yep. the simplest and best is to have a hidden form, which normal users can't see. Spammers will usually insert junk into all form params, and then you disallow if the hidden form is filled
benabik dukeleto: phpBB still actively developed and seems to be more featureful than the alternatives.
whiteknight dukeleto: Combine that with hidden fields which use JS to duplicate necessary information
so some hidden fields must have values, and some must not
18:01 ligne is now known as ligne-pub
NotFound If you enter the javascript-only realm there are more poweful ways: build the form with javascript with data or html code obtained with ajax. 18:03
A spam robot will need some work to duplicate that. 18:04
18:04 ShaneC left
ligne-pub anyway, the reason i wanted to create a ticket was that i've found a couple of bind()-related race conditions in the IPv6 socket tests. 18:08
i've got a patch here, but (a) i've never written any PIR before, so it's probably a bit smelly, and (b) it involves a couple of choices i'm not terribly sure about: github.com/ligne/parrot/commit/942...3b480e48e8 18:09
whiteknight ligne-pub: oh, that does sound serious 18:11
NotFound Why dow we need to bind some fixed port? 18:12
whiteknight NotFound: what is the alternative: Random ports?
NotFound whiteknight: I think so. 18:13
dukeleto ligne-pub: thanks for looking into this!
cotto_work At this point I'm willing to say that we need human-friendly trac registration. Too many people have been tricked by the existing one.
ligne++
dukeleto ligne-pub: i wrote those tests :)
NotFound: the tests attempt to bind to different ports if 1234 is not available 18:14
ligne-pub: that commit looks pretty decent for your first PIR. Nice job! 18:15
ligne-pub: if you send a pull request, I will merge it in 18:16
ligne-pub: those tests failing when running parrallel tests really sucks
ligne-pub dukeleto> thanks :-)
dukeleto> done. 18:19
dukeleto ligne-pub: also, send a pull request adding yourself to CREDITS
dalek rrot: c435b74 | dukeleto++ | t/pmc/socket_ipv6.t:
Merge pull request #132 from ligne/94251bffce8026a60af9801b38b67b3b480e48e8

Fix race condition in IPv6 socket tests and make them less likely to fail when running tests in parallel		 18:20
dukeleto ligne++ for all the pull requests 18:21
benabik just got his package from Google! 18:22
dukeleto benabik: take a pic. i would like to see what they send 18:23
18:23 PacoLinux left
cotto_work benabik: make sure the credit card number is clearly visible 18:23
benabik dukeleto: Mostly it's my prepaid card and a letter explaining it.
cotto_work: :-P
dukeleto benabik: and SSN and mothers maiden name
benabik: ah, ok. I thought it was a shwag bag
18:23 PacoLinux joined
benabik dukeleto: There's a little bit of swag, and I'll take a pic of that. 18:23
18:25 ambs left
benabik Actually the card itself looks pretty neat. Will just have to blur things. 18:26
dukeleto we are now down to 6 gsoc students. cgaertner never re-appeared.
cotto_work dukeleto: he's officially out?
whiteknight officially, yes
benabik :-(
cotto_work That's disappointing.
benabik What was his project? 18:27
cotto_work GObject interop with Parrot
bubaflub bummer
atrodo that's a big shame
tadzik how officialy is he out, I mean the Mentors Community decided that because he didn't show up he's excluded? 18:28
s/Community/Comitee/, but whatever 18:29
whiteknight tadzik: his mentor (me) and the org admins(me and dukeleto) talked to Carol Smith about it
We don't know when he will come back, or if he ever will. He's been gone for over a month now 18:30
seen cgaertner 18:31
aloha cgaertner was last seen in #parrot 51 days 4 hours ago saying "btw, who would be willing to mentor the project? the proposal template ask for that information, I I don't remeber anyone actually mentioning that...".
whiteknight 51 days and 4 hours ago
so almost two months
tadzik understandable
bubaflub i hope he's alright
tadzik yeah, I hope so
whiteknight I hope so too. But there's probably no way we will know
tadzik I still have a dark vision of him appearing in a week from now saying "here, that's the code I've written in week 1st" 18:32
dalek rrot: 1d9daa5 | petdance++ | config/auto/warnings.pm:
check for GCC 4.7 new -Wmaybe-uninitialized
whiteknight tadzik: yes, I'm very afraid of that too
dalek rrot: 372ebbe | petdance++ | src/pmc/structview.pmc:
const a local pointer
cotto_work petdance++ #I just read about that option. Now I know where you get those.
Andy From the LJ guy? 18:33
cotto_work I think so.
Andy The one thing it found was something that splint already found,
but at least now GCC users can see it.
whiteknight We'll add that to the list of compile warnings that we are currently ignoring even though we know we shouldn't be 18:34
Andy whiteknight: Probably you won't on this one.
The one thing it's found looks like it's indeed a bug.
cotto_work nice
Andy line 1029 in src/pmc/structview.pmc 18:35
whiteknight Andy++
Andy It's (apparently) possible to get through the first switch without setting result.
If the default on the first switch is a That Can't Happen, then let's throw a failed assertion there, and the warning goes away. 18:36
bubaflub just got my GSoC package 18:38
18:39 ShaneC joined
tadzik I want mine too :( 18:40
benabik www.cs.rit.edu/~bcg2784/GSoC-stuff.jpg 18:43
dukeleto: ^^
whiteknight No fair! I didn't get a notebook, a pen, or a sticker 18:44
cotto_work I didn't even get money.
Andy Screw that, I'll take the credit card.
atrodo I definitely graduated college too early. Should have been on the 8 year plan 18:45
benabik atrodo: I started college in '99...
whiteknight I still have my t-shirt though. About 30% of my current wardrobe is google T-shits
t-shirts*
I'm wearing my Google Code-In 2010 shirt now
atrodo benabik> I started the year after you then
benabik Ooh, the notebook is grid paper. 18:46
whiteknight I hope mentors get one of those swank notebooks
tadzik the notebook looks supercool
whiteknight hell, I'll buy one if I have to
tadzik it looks too nice to use it
I'll ask them for a second one after the midterms :>
(assuming I get this far)
whiteknight tadzik: we have faith 18:47
tadzik thanks :)
18:51 mj41 joined
benabik Does NQP implicitly return the last value in a method? 18:54
18:55 ShaneC left
cotto_work benabik: yes 18:55
I like that idiom but it takes some getting used to.
sorear Is NQP's return statement still several orders of magitude slower than falling off the end? 18:56
benabik sorear: That sounds poor. 18:58
whiteknight I never saw any benchmarks comparing the use of the return keyword 19:02
I figured it used the same mechanism as a fall off the end
NotFound I've hear nqp uses control exceptions. 19:14
davidfetter is there still a PIR? if so, is it deprecated? 19:15
whiteknight yes and no :( 19:20
dalek rrot: bbd5de8 | util++ | / (12 files):
Merge branch 'Util/TT1217'
 19:21
davidfetter whiteknight, what would you recommend i use instead?
whiteknight davidfetter: for what purpose? 19:23
davidfetter whiteknight, well, dukeleto's done yeoman work on embedding parrot in postgres
the first embedded language (PL, in postgres parlance) was PL/PIR 19:24
19:24 Coke left 19:25 Coke joined
davidfetter that's not a direction set in stone, but the idea was that HLLs would somehow compile down to PIR 19:25
19:30 Coke left, Coke__ joined
whiteknight The reality is that compiling down to PIR is still the only way to get code running on Parrot 19:30
so PL/PIR is probably an unavoidable link in the chain for now
davidfetter whiteknight, is there a "PIR Considered Harmful" somewhere so i can familiarize myself with your dislike of it? 19:35
whiteknight davidfetter: there is somewhere. It's a topic that's probably in need of a reawakening
moritz davidfetter: the gist is that it's far too complex for an assembly language, and still too low level for a high-level anguage
*language
davidfetter moritz, sounds like C ;) 19:36
tadzik :)
whiteknight and it doesn't do everything that PBC can do, and there is magic in the language which cannot currently be reproduced in other languages without compiling down to PIR
benabik davidfetter: It's too low-level to be C. :-(
whiteknight PASM can't even do everything PIR can do because of the magic
19:36 dodathome joined
lucian davidfetter: i agree with whiteknight on this, i don't think anyone should ever be writing assembly, no matter how high-level it is 19:36
19:37 Coke__ left, Coke joined
davidfetter lucian, i'm not disagreeing, generally. there are still cases in the postgresql source code (as in OS kernel code, and for similar reasons) where people code assembly 19:37
spinlocks, e.g. 19:38
Util Replacing IMCC with something more hackable should allow us to enhance PASM to allow for hand-coding the PIR-only tricks.
dalek TT #1217 closed by Util++: [PATCH] t/dynpmc/foo.t converted to PIR
TT #1217: trac.parrot.org/parrot/ticket/1217
lucian davidfetter: sure, but they're small niches, and getting smaller
davidfetter yep
dukeleto davidfetter: anything good happen at the PL/summit that I should know and/or that is relevant to PL/Parrot? 19:39
lucian and in PIR's case, it's not necessary
look at the JVM, its bytecode doesn't even have a standard textual form
davidfetter dukeleto, lessee. there was general agreement that we really need a base PL library, and some work towards implementing one in the form of a PL/NULL, which would act as a skeleton PL 19:40
dukeleto davidfetter: sounds promising. the docs could then be a description of how that works 19:41
davidfetter well, i see PL/NULL as an intermediate step to PL.h (or whatever we'd call it)
19:52 bluescreen left 19:56 jevin left 19:57 bluescreen joined 20:09 silug left, silug joined, jsut_ joined 20:14 jsut left
dalek rrot: 0624802 | dukeleto++ | / (2 files):
Merge pull request #133 from ligne/socket_test_races

fix test error when port 1234 is already bound.		 20:16
Coke_ anyone with tuits who could poke at the macport patch would be appreciates. 20:19
*d
davidfetter Coke, we normally refer to that as, "the Scottish Packaging System" out of an abundance of caution 20:23
20:40 dodathome left
benabik Yay, GSoC work: github.com/Benabik/parrot/commit/1...60ee0acce5 20:40
Coke_ davidfetter: I'm an ex-theatre geek, and *thbbbthp* 20:41
davidfetter Coke_, heh 20:42
cotto_work benabik: it makes me very happy to see progress\\ 20:51
benabik Hm. Needed to fix that merge. Old link may no longer be valid, but this one will: github.com/Benabik/parrot/commits/nqp_pct 20:52
cotto_work: Feels good to make progress. Now I need to figure out this whole "get it to build" part. :-D
20:52 soh_cah_toa joined
cotto_work benabik: that's a key step 20:55
speaking of getting things to build...
hi soh_cah_toa
benabik dun dun dunnnnnn 20:56
soh_cah_toa cotto_work: lol, i saw that coming :)
cotto_work: right now i'm changing all references to 'pdb' to 'hbdb'
cotto_work I'm going have trouble not subvocalizing that to "hubdub" 20:57
benabik hbdb?
soh_cah_toa benabik: honey bee debugger
cotto_work that's the name for his debugger
soh_cah_toa benabik: it's a nickname i gave to one of my cats
benabik soh_cah_toa: cats++
cotto_work It's like a parrot, but smaller
soh_cah_toa ha :)
benabik Totally is. My cat loves to sit on my shoulder even 20:58
(Well, one of them.)
soh_cah_toa benabik: yay, another cat lover
21:00 fperrad left 21:10 hercynium left
allison lucian: pong (much later) 21:20
lucian allison: what's up? i still have exams ...
i've been wondering, do you think pynie could be reused after all? 21:21
allison lucian: I was just responding to your ping
lucian: there's not much to reclaim, just NQP stuff, and a few tiny test files
lucian if i could start with the object model (next week likely), i could possibly get faster to a solution
allison lucian: pynie didn't have an object model 21:22
lucian i know
but it did have a compiler
allison lucian: CPython has a better compiler
lucian yeah, i know
i was just wondering if i could postpone writing a backend to that, to save time 21:23
whiteknight lucian: I thought writing that backend was your project
allison lucian: so, in terms of reuse, you'll save more time reusing the one that's complete, than one that's very limited
lucian whiteknight: the object model was more interesting, a compiler backend already exists 21:24
whiteknight or, a majority of your project
lucian allison: right. i guess you're right, i haven't been able to fix pynie last time i tried
allison lucian: at this point, I'd say a fresh start will be faster and far less frustrating 21:25
lucian nods
allison: right, so i'll go with that 21:27
21:27 whiteknight left
lucian hopefully i'll find some time this week as well 21:27
allison lucian: when are you done with exams? 21:29
lucian allison: last one is on the 1st
allison: i'll try to get some work done before that, too 21:31
it's very frustrating how late uni ends
last year it was much better
allison lucian: that sounds good, and yeah the schedule overlaps are tough 21:32
tadzik yeah...
21:36 mj41 left
cotto_work After reading today's tdwtf, I've come to the conclusion that cached boxing conversions are a necessary feature for the next release. 21:39
thedailywtf.com/Comments/Disgruntle...ition.aspx
benabik Hm. NQP doesn't want to compile a class named PAST::Node because one is already loaded. :-( 21:40
Or I screwed up the Makefile. Take two? 21:46
21:52 davidfetter left 22:16 Andy left 22:34 kid51 joined
mikehh dukeleto: ping 22:42
23:11 mtk left
NotFound Enjoy: github.com/NotFound/WinxedGtk 23:12
tadzik awesome 23:14
NotFound No Makefile yet, to test it you must manually do: "winxed -c -o pir/WinxedGtk.pir src/WinxedGtk.winxed" before executing setup. 23:16
And that's enough for today, bye.
cotto_work NotFound++ 23:17
aloha: clock?
aloha cotto_work: LAX: Mon, 16:17 PDT / CHI: Mon, 18:17 CDT / NYC: Mon, 19:17 EDT / UTC: Mon, 23:17 UTC / LON: Tue, 00:17 BST / BER: Tue, 01:17 CEST / TOK: Tue, 08:17 JST / SYD: Tue, 09:17 EST
cotto_work 'night
23:18 mtk joined
dalek rrot: dc7d625 | jkeenan++ | t/dynpmc/foo- (10 files):
[codingstd] Correct format of copyright notice. These are all new files, so we only need 2011 in notice.
 23:18
23:21 daniel-s joined 23:28 atrodo left 23:31 lucian left
dalek rrot: 580f68f | jkeenan++ | t/dynpmc/foo-07.t:
[codingstd] No trailing whitespace.
 23:31
rrot: bc3b59a | jkeenan++ | t/dynpmc/foo- (10 files):
[codingstd] Need coda for PIR, not for Perl 5.
dukeleto kid51: which dates will you be in NC for YAPC::NA ? 23:32
kid51 Sun afternoon - Thu afternoon 23:33
dukeleto kid51: and the hackathon is wednesday? 23:34
cotto_work www.yapc2011.us/yn2011/wiki?node=Parrot/Perl6 Hackathon
Thursday all day and either Tuesday or Wednesday evening 23:35
23:35 kid51 is now known as kid51_at_dinner 23:41 whiteknight joined
whiteknight good evening, #parrot 23:41
it feels very good to be able to say that again 23:46
dalek Heuristic branch merge: pushed 26 commits to Rosella by Whiteknight 23:47
cotto_work wb, whiteknight 23:48
whiteknight hello cotto_work 23:49
cotto_work nice to have you back online
whiteknight it's nice to be back :)
cotto_work I guess lawnmowers are the new bus. 23:52

Powered by Cro and the Raku® Programming Language.

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

Thank you!