Parrot 1.0 Released | parrot.org | 367 RTs left!
Set by moderator on 12 April 2009.
00:09 AndyA joined, TiMBuS joined 00:15 s1n left, s1n joined 00:38 wayland76 joined 01:37 allison joined 01:40 Andy joined 01:47 Tene joined 02:05 kid51 joined
dalek rrot: r38097 | Util++ | trunk/docs/book (9 files):
[book] Typo corrections
 02:23
rrot: r38098 | allison++ | branches/pcc_rewiring/src/pmc/cpointer.pmc:
[pcc] Don't mark the PMC and String children of a CPointer (have made
 02:39
rrot: r38099 | allison++ | branches/pcc_rewiring/src/ops/core.ops:
[pcc] Remove unnecessary reset on unused interpreter struct member.
02:47 janus joined 03:10 eternaleye_ joined 03:31 workbench joined 03:32 ab5tract joined
dalek rrot: r38100 | allison++ | branches/pcc_rewiring (2 files):
[pcc] Implement parameter filling from C arguments, to be used by
 04:31
purl pccmethods are my friends
05:31 Theory joined
Tene any parroters around boston, ayone know? 05:37
05:53 szabgab joined 05:57 eternaleye joined 06:10 ruoso joined 06:31 uniejo joined 06:45 japhb joined 07:21 amoc joined 07:26 skv_ joined 07:27 skv__ joined 07:28 iblechbot joined 07:29 krunen joined 07:57 masak joined
cotto Cool. Now I know what "segfault" is in German. 08:07
wayland76 zegvault? 08:09
cotto Speicherzugriffsfehler 08:12
I love how you can stick together random words in German like that. 08:13
It's too bad that hasn't migrated to English.
mikehh we do have some - like busstop 08:16
wayland76 and goatherd :)
But if I understand correctly, the Germans do it in a more free-form fashion 08:17
cotto Yup.
masak as do the Swedes, and other speakers of the nordic languages. and Finnish. 08:19
wayland76 You know how sometimes in English, people say "is that even a word", and yet it's a perfectly legitimate word that just has a lot of suffixes/prefixes and stuff?
Well, the Germans do that, but with words, instead of prefixes and suffixes
08:19 riffraff joined
wayland76 (or that's the impression I get) 08:20
masak but we're suffering from English influences, and people increasingly tend to spell compond words apart.
wayland76 I didn't know about the North Germanic languages being that way
It's interesting how others influence your speech
masak totally legit Swedish word: "kyrkogårdsvaktmästare". "graveyard janitor". 08:21
wayland76 (I guess I'm just thinking about how I know lots of big words, but rarely use them in order to accomodate those around me)
masak note the extra "s" in the middle, functioning as a sort of priority marker. a primitive parenthesizing system. 08:22
wayland76 What's the vakt part mean?
masak watch. 08:23
wayland76 (I don't know Swedish, but I know enough Old English to make sense of the rest, I think)
masak so, "watchmaster"
wayland76 Ah, ok :)
masak or "guard", rather.
wayland76 kirk-yard--watch-master, yes?
masak something like that. :)
wayland76 (kirk is Scots for church) 08:24
masak it's from the Nordic languages originally, I believe.
wayland76 Well, I suppose that's possible. I guess I was under the impression that it was OE, but I couldbe wrong 08:28
Incidentally, the English have aglommerated: www.u.arizona.edu/~ctb/wordhord.html 08:29
But no longer :)
Cathode ray tube in Old English: niðerstigendescimapipe :) 08:30
English "church": O.E. cirice "church," from W.Gmc. *kirika, from Gk. kyriake (oikia) "Lord's (house)," from kyrios "ruler, lord." 08:31
The OE would probably be pronounced something like kee-ri-keh (not IPA, I know :) ), which isn't far from kirk 08:32
But this probably belongs on #perl6 -language, not perl6^Wparrot compiler :) 08:33
Anyway, food. afk& 08:34
09:42 TiMBuS joined 10:22 gaz joined 10:37 ruoso joined 10:39 bacek joined 10:48 kid51 joined 11:30 clunker3 joined 11:36 AndyA joined 12:00 iblechbot joined 12:03 TiMBuS joined
dalek rrot: r38101 | Util++ | trunk/docs/book/ch13_reference.pod:
[book] Typo correction
 12:12
12:33 flh joined
dalek rrot: r38102 | Util++ | trunk/docs/book/ch13_reference.pod:
[book] PASM Quick Reference - Added new opcodes from io.ops
 12:46
rrot: r38103 | Util++ | trunk/docs/book/ch13_reference.pod:
[book] PASM Quick Reference - Renumbered tables
 12:49
mikehh t/codingstd/copyright.t fails at r38100 - Kubuntu Intrepid Amd64 13:00
I think someone mentioned that it worked on perl 5.8.8 - I am on 5.10.0 13:01
the rest of make codetest passes just fails t/codingstd/copyright.t - Failed test: 3 13:07
13:09 gryphon joined 13:10 ruoso joined
mikehh running make fulltest_all while I go shoping - afk& 13:11
13:37 DietCoke joined 13:50 Tene_ joined
dalek rrot: r38104 | Util++ | trunk/docs/book/ch13_reference.pod:
[book] Typo correction
 14:18
Coke so I'm looking at enabling a compiler warning, unused-parameter 14:21
basically enforces the use of UNUSED or SHIM (we're already doing that most places). PMCs are a bit tricky.
we have some methods that are unimplemented, those are easy enough to generate UNUSED() for. we have a few cases, though, where the body of a VTABLE or METHOD doesn't refer to the interp that was passed in. It seems a bit onerous to the PMC designer to force them to add UNUSED(INTERP) in the body decl. in the PMC, no? 14:23
particle- by doesn't refer, you mean interp is unused? 14:24
Coke yes.
and interp doesn't even appear in the VTABLE decl.
dalek rrot: r38105 | Util++ | trunk/docs/book/ch13_reference.pod:
[book] PASM Quick Reference - Removed old dotgnu.ops opcodes
particle- yes, that's tricky. 14:25
i don't want to force folks to add that there, it doesn't make sense.
so how do you tell the warning to shut up, skip pmc files?
Coke ok. but if an explicitly mentioned arg is unused, that's OK, neh?
particle-: I'd rather not skip PMC files.
particle- yes
i'd rather not skip, too. 14:26
but how do you stop the warnings?
Coke I'd rather have the code generation DTRT.
particle- ah, yeah, static analysis during codegen
Coke easy, generate UNUSED(INTERP) properly. =-)
particle- yep, that's the way to do it.
Coke just wanted to verify that forcing the PMC to add it manually was wrong.
Util You both out-typed me, w.r.t code generation. idea++ 14:27
14:30 pyrimidine joined
Coke hurm. default.pmc and null.pmc share a bit of codegen. 14:32
er, /should/, but it's cut and pasted.
particle- refactoring will probably cause failing tests 14:33
14:34 iblechbot joined
Coke nothing functional. 14:35
particle- si, es verdad. 14:36
14:36 Theory joined 15:00 bsdz joined 15:04 davidfetter joined 15:09 buildbot joined 15:20 rg joined
particle- rakudobug? 15:39
purl i heard rakudobug was mailto:rakudobug@perl.org
dalek kudo: 8062a1b | (Moritz Lenz)++ | t/spectest.data:
three more passing tests for t/spectest.data
 15:46
shorten dalek's url is at xrl.us/beoxpr
dalek kudo: 48c2042 | (Moritz Lenz)++ | Test.pm:
fix logic error in unary nok() function in Test.pm
shorten dalek's url is at xrl.us/beoxpt
dalek kudo: 13c2c7c | (Moritz Lenz)++ | (2 files):
The sub form of split() expects the delimiter as second parameter
shorten dalek's url is at xrl.us/beoxpv
dalek kudo: f989187 | (Moritz Lenz)++ | t/spectest.data:
return2.t was merged into return.t; remove it from spectest.data
shorten dalek's url is at xrl.us/beoxpx
dalek kudo: 9ee6477 | (Moritz Lenz)++ | t/spectest.data:
we pass angle-brackets.t
shorten dalek's url is at xrl.us/beoxpz
dalek kudo: d208e1c | (Moritz Lenz)++ | (2 files):
Hash.reverse is now called .invert
 16:00
shorten dalek's url is at xrl.us/beoxti
16:13 Tene joined 16:15 contingencyplan joined 16:16 iblechbot joined 16:53 davidfetter joined
Coke_afk is thawfinish a vtable? 17:08
Coke (it is) - why is not declared VTABLE in most PMCs? 17:09
17:19 braceta joined
rg ok, i believe i've figured out why t/op/trans.t is failing on jit for freebsd aswell as some linux platforms 17:21
what's interesting is that my workaround makes gives me a segfault on other tests 17:22
makes me wonder how useable this jit implementation really is :( 17:24
17:27 kj joined, barney joined
dalek rrot: r38106 | coke++ | trunk/src/pmc (6 files):
[cage] mark these VTABLE implementations.
 17:30
17:31 masak joined 17:34 Theory joined
Coke someone needs to take up a marketing position for parrot and for perl6. 18:01
Util In .ops files, what is the difference between `invar` and `inout` parameters? 18:02
18:07 mikehh_ joined
allison Util: 'invar' just means it's a variable rather than a constant. 'inout' means it's a result (a "destination" register) 18:10
Util: 'in' can be either a register or a constant 18:11
Util thanks, allison 18:19
18:20 chromatic joined
bsdz hi, anyone have any ideas how i can trace a "set_pmc_keyed() not implemented in class 'Integer'" error in a 'parrot;PCT;HLLCompiler;eval' sub? it says (EVAL_7:168) for the location; i wonder if that's the same as i woudl see in --target=pir? 18:20
chromatic Should be. 18:21
bsdz oh cool. i'll take a closer look. thanks
Coke ah, great, parrot and perl6 abuse. 18:27
(other server)
to get the EVAL files to stick around, use -D60, IIRC. 18:28
bsdz very cook 18:29
i meant cool :)
rg ok, the segfault is just another jit bug
allison Util: I have a pile of uncommitted changes to the book chapters. Is it easier for me to go ahead and commit them, or to integrate your changes into my changes? (How far along are you?) 18:30
Util: and, thanks for the typo fixes, and work on the book!
chromatic #ps time 18:31
moritz I don't have time for #ps right now; but I have nothing to report either
have the appropriate amount of fun ;-)
18:44 eggyknap joined
eggyknap Question: per #parrotsketch, calls to PIR subs/methods from C work. How stable can I assume the API involved to be? (in other words, I wanna play - should I expect to have to rewrite a bunch of stuff b/c of API changes) 18:45
chromatic That depends on which API you use.
eggyknap Fair 'nuff. I admit to not having done my homework recently enough to know what my choices are. 18:46
allison If you're using the Parrot_pcc_invoke set of routines, you're fine
If you're using any of the other variants, they're on the way out
eggyknap Sweet... many thx. 18:47
Coke so, the boston.pm mailing list is suggesting that the perl6 community has dumped parrot and has moved on the CLR. 19:06
www.mail-archive.com/boston-pm@mail...05520.html
News to me. 19:07
chromatic s/News/Crack/ 19:08
gravity That makes me feel better that I've not gone to boston.pm
particle- no, no, that's just mitchell charity, who is somewhat wacky. 19:09
bsdz They'd have more luck if they targeted the DLR not the CLR. Perhaps they should forget Perl altogether and use Powershell - dog slow mind you.
chromatic "Elf is based on the hypothesis that developers could be found if the implementation language was itself Perl 6." 19:11
I question that hypothesis, based on the history of Elf.
PerlJam chromatic: you noticed that too, eh? :)
chromatic I've been saying that for a few years.
particle- one lone developer, who never sent a mailing list message about it until now. 19:12
masak maybe it just Isn't Ready for developers yet?
chromatic If you want to run Perl 6, you have to invent most of Parrot.
19:12 altious joined, uniejo joined
PerlJam masak: release early, release often. 19:13
masak PerlJam: true.
chromatic "Perl 5 suffers from [its] lack of sufficient testing." 19:14
Perl 5 suffers from many things. That is not on the first page of my list.
masak frankly, I'm impressed by Elf simply because it's a fairly large piece of runnable Perl 6 code.
particle- but by definition, it's throwaway. 19:15
'Any large project written in p6 now is almost by definition throw away code. ... "I'd have architected this differently if I had full p6"... '
masak I'm not sure I agree fully to that. 19:17
Half a year ago, yes.
particle- i don't agree at all.
masak Today, I'd guess the code in Druid is pretty close to Perl 6.0.0 code.
particle- except laziness, and some other small things :)
chromatic "Raise enough funds for TPF to hire a full time, dedicated project manager (someone from the Perl community willing to work on minimal income)."
As if lack of project management were a problem. 19:18
particle- yeah, it's the lack of contributors that's closer to the problem
chromatic "Break down the features into self-contained projects, and put them out to bid. Something a contract programmer could work on for a month or two between projects. "
particle- what's a paid pm going to manage? volunteers with limited, unpredictable tuits? 19:19
chromatic As if TPF can afford to pay contract rates.
Coke if we really did need a paid PM, would we want the cheapest one we could find? =-)
how do we fix a lack of contributors? I know some very sharp people who have told me flat out they'd never work on parrot.
chromatic You show visible, frequent progress. 19:20
Coke because of issues with the architecture, or crappy docs, or...
chromatic: our progress isn't as visible as you might like.
chromatic I agree, but imagine if we had one release a year.
Coke (I'm not saying we gear ourselves towards the vocal idiots I deal with.)
true. even worwse.
chromatic What would have happened if we'd set the 1.0 release date and failed to meet it?
Coke but I can't answer questions like "are anything of the languages that run on parrot feature complete? fast?"
particle- coke, do you really hear "i'd never contribute to that project. there aren't enough docs."
chromatic When we set the release date, a lot of people said "You'll never hit it." 19:21
particle- coke: lua is "complete"
Coke not not enough docs. quality thereof.
particle-: ... really?
particle- yes.
Coke interesting how someone on the board had no idea. =-)
chromatic Lua's been complete for at least a year.
Lua needs profiling and optimization (and shows how our JIT just isn't cutting it), but it's been complete. 19:22
barney put that on reddit
bsdz parrot would certainly have a lot more uptake if it was bundled as a vm in ie, firefox and chrome. also if it had the same systray applet that java has. probably need to be locked down security-wise first. 19:23
chromatic "One could certainly imagine doing a javascript implementation of the p5 vm for instance."
I can't imagine anyone familiar with Perl 5's internals doing that. 19:24
particle- bsdz: yes, if parrot 1.0 were parrot 2.0, it would be more useful for all users
we targeted high-level language developers with the 1.0 release, not business or casual users. 19:25
that's targetted for 2.0 and beyond, so we'll be working on it this year.
bsdz small steps first then take larger ones. i for one can see parrot doing what sage math does but more efficiently. a unifying vm that brings together lisp, python, m-script and perl. 19:26
PerlJam bsdz: while I like the sentiment, that list makes me go ugh.
bsdz ha ha. it has that effect on some :) 19:27
i must admit i don't like lisp but that's what maxima is written in (perhaps even emacs?) 19:28
19:36 PerlJam joined
bsdz on a different topic. are there any known issues using a keyed PAST::Var and vivibase on a multidimensional pmc when extending the dimensions. pseudo code like this x[1] = 5; /* x= ( undef, 5 ) */; x[1][2] = 3; /* x= ( undef, (undef, undef, 3) ) */. ignoring the fact i'm also clobbering one of my elements. 19:37
Coke particle-: I'm not sure that 1.0 was really a stable release for HLL developers. 19:39
all we did was move our deprecation window from 1 month to six months there.
everything is still subject to change. 19:40
PerlJam Coke: that sounds slightly more stable :)
Coke for large values of everything. (if not actually everything)
I just want to be careful about overselling as well as underselling.
particle- it's a stable platform for hll developers. 19:41
last year saw a significant number of segfaults and memory leaks eliminated
many apis were standardized
deprecation pace was slowed 19:42
a release policy was implemented.
a support policy, too
Coke particle-: I call bullshit. =-)
the 1.0 release basically said that all the api's were up for shuffle before the 1.4 release. 19:43
chromatic stable != complete
19:43 riffraff joined
Coke s/before/in/ 19:43
particle- coke: java deprecates apis, so does .net. how is this different than those production products?
PerlJam particle, Coke: If I'm interested in implementing a language on parrot, what advantages do I get? (That's the only marketing question that needs to be answered based on the claims of 1.0 IMHO) 19:44
Coke PerlJam: right now? none?
all of the good bits are in the future, no?
PerlJam Coke: I think "none" is a little severe.
bsdz software is never complete - it's like Aristotle's Arrow
Coke PerlJam: I'm a little jaded, and have just spent an hour listening to someone rant about such things, apologies. 19:45
19:45 uniejo joined
particle- coke: obviously, you listen too much. 19:45
tell raye i said that.
Coke lua? 19:46
purl lua is at www.lua.org/
chromatic I've always see 1.0 as a shift in focus. "Here is the minimum supported feature set we believe will allow an HLL developer without commit access to the project to begin to build a usable, real-world language."
s/(?<incomplete>see)/$+{incomplete}n/ 19:47
particle- unlike perl 6, which is a perfect-world language :)
Coke it would be interesting to see current benchmarks of lua on parrot vs. other lua.s 19:48
msg fperrad it would be interesting to see current benchmarks of lua on parrot vs. other lua.s
purl Message for fperrad stored.
cotto Coke, I think fperrad's Lua does too much reinventing of the wheel to be a meaningful benchmark of a Parrot-hosted language implementation. 19:49
PerlJam cotto: which wheels? 19:50
chromatic Dispatch, for one.
cotto PerlJam, verifying so I don't feel stupid later. 19:52
Coke cotto: well, partcl was also godawful slow.
PerlJam "was"?
purl i think "was" is "what"
Coke PerlJam: yes. now it's just broken.
PerlJam oh, bummer
Coke the patch allison used to test building against an installed parrot is broken in 1.0. I haven't had the tuits to fix it, and am basically waiting for rakudo to do it for me. 19:53
chromatic We still don't have good profiling tools to figure out why, though.
(Random guess: PMCs.)
Coke partcl has to use PMCs for pretty much everything anyway, at least the last I checked. 19:54
if parrot is using too many of them internally, well...
but yes. need profiling.
chromatic,particle- : so, for partcl at least, the promise of 1.0 hasn't been realized, and probably won't be until at least 1.1 19:55
I'm not sure it's nicely fixable without fixing parrot_config 19:56
(that is, I'm not sure it'll ever work against a pristine 1.0)
chromatic Now we know.
Coke yup.
chromatic When would we have managed to get to that point without trying to put out a 1.0?
Coke chromatic: I'm not saying there's a better path. I'm just saying, let's clearly label the sign posts. 19:57
chromatic Sure, that's fine.
masak which would be the least work, building an XML parser in PGE, or hooking up an existing C library?
chromatic Were we wrong about some of our goals? Yep. Were we right about some of them? Yep.
masak, depends on the library. If the shared library does SAXy callbacks, PGE is easier. 19:58
Coke masak: I would say that depends on if you're a perl6 programmer or a C programmer.
chromatic Do we know more about what we need to do to support HLL developers? Yep.
masak definitely a Perl 6 programmer.
Coke then I'd probably push you towards PGE. I can't imagine it's going to be fast, though.
PerlJam masak: didn't moritz or someone write a simplish XML parser a while back?
masak PerlJam: aye.
PerlJam masak: then you could reuse the grammar with PGE ;) 19:59
masak PerlJam: indeed. just wondering what is the best way to do it.
PerlJam masak: I don't know about "best" but I know I'd go with PGE before C 20:00
masak understood.
purl understood. are you on schedule?
masak purl: no. :)
purl masak: excuse me?
bsdz any one have any expertise on vivibase capability in pct? 20:01
particle- vivibase is when you try to create a hash key, and the key doesn't exist, right? 20:02
bsdz yes, and same for arrays too
particle- what expertise do you need?
PerlJam (I know what it is too now ... am I an expert? ;) 20:03
bsdz well. i need to extend my array by another dimension and i'm getting "et_pmc_keyed() not implemented in class 'Integer'" errors. i see the same thing in squaak. i'm wondering if it's a limitation of the implementation. 20:04
my squaak code looks like this: var y; y[1] = 5; y[1][2] = 3; 20:05
Coke i don't think arrays are implicity multidimensional. 20:10
I imagine that's (under the covers) trying to set 5[2] = 3;
particle- we don't have multidim arrays
you have to implement your own pmc for that
20:11 Theory joined
Coke wierd, given all the theoretically optimized pmc variants we have. 20:11
(tell me again why we have those?)
chromatic Because C is fast! 20:12
Coke liar. 20:13
rg actually multidimensional arrays work in squaak (or did some time in the past). however, i guess there is no auto vivification
bsdz well i am creating my own multidimensional array using my own :inline past::op in my vivibase. thing is it stumbles before it gets there.
chromatic Perceptively, C is fast. If Ruby is too slow, write parts of your app in C. If Python is too slow, write parts of your app in C. If Perl is too slow, write parts of your app in C. 20:14
If JavaScript is too slow, write more of your JavaScript compiler in JavaScript.
20:14 eggyknap left
bsdz chromatic: i couldn't agree more. C is fast but assembly can be even faster as i proved to some colleagues some time back 20:15
chromatic It's not that C isn't fast, but that memory is slow. 20:16
PerlJam chromatic: In your perl testing book do you talk about when one should create test data versus mocking a database for tests that require a database? 20:18
PerlJam is looking for a good rule of thumb.
Coke rule of thumb?
wow, thought purl would have something. =-)
chromatic Some, yes.
PerlJam One of my coworkers asked about it and I couldn't give a good answer as to when one approach would be favored over the other (except when the size of data needed for the test is large) 20:20
chromatic My rule of thumb re: mocking is to mock only those things that are horribly difficult to set up or for which exceptional conditions are unlikely to occur.
Disk full! Abort! LP1 on fire! Socket connection broken! IPv6 implemented everywhere!
particle- porn! 20:21
PerlJam heh
chromatic Mocking is appropriate in about 5% of the cases, I believe. 20:22
The BDD folks (I hate that initialism) pull out the mock card at the start.
To their credit, some of them replace the mocks as they write real code.
particle- bdd?
chromatic It's like test-driven development, but with more awesome.
particle- mock turtles. 20:23
chromatic Instead of programming, YOU'RE A ROCK STAR.
PerlJam the ruby folks have taken BDD to great extremes.
chromatic At the end of a successful day of BDD, they smash their keyboards.
allison purl: bdd is Behavior Driven Development
purl OK, allison.
chromatic Never mind the Sex Pistols, here come the BDDers! 20:24
allison purl: bdd is also en.wikipedia.org/wiki/Behavior_Driv...evelopment
purl okay, allison.
PerlJam oddly, some of the arguments for BDD tools sound very familiar ... "*managers* can write tests. *clients* can write tests" etc. 20:25
chromatic That's because they're written in English!
Well, a domain-specific language.
Okay, Ruby.
Well, Ruby with a lot of stupid helper methods like .a .the .an .and
Just like English. 20:26
PerlJam chromatic: wiki.github.com/aslakhellesoy/cucumber
chromatic If you're a three year old.
PerlJam (that's the latest thing my ruby friend is crazy about)
chromatic Those who do not know Prolog are condemned to reinvent it, poorly. 20:27
20:29 Tene_ joined 20:32 gryphon joined 20:40 jan joined
bsdz msg pmichaud i think there's an issue in PAST::Compiler::vivify. An example in Squaak "var y; y[1] = 5; y[1][2] = 3;" gives "get_pmc_keyed() not implemented in class 'Integer'. This seems to be because storeop does not check name/basepost is an aggregate before setting. Also seems to be no way PCT way around it. 21:02
purl Message for pmichaud stored.
bsdz purl? 21:03
purl bsdz?
rg bsdz: what would you expect y[1] to be after the second assignment? 21:11
particle- bsdz: parrot has an op for checking what type of interface a pmc provides
is it the 'provides' op? i forget whether that was renamed from 'does'. 21:12
bsdz the best thing would be y_before = (, 5) and y_after = ((, 5, ), (,,3)) 21:13
particle- hopefully it was, and 'does' properly checks roles now instead of interfaces.
rg and you couldn't set y[0][1] = 5 instead? 21:14
bsdz i imagine same problem would occur. 21:15
yes same problem 21:16
particle: the op is "does" i think
"provides" is mentioned in .pmc files 21:17
rg ok, so it does not auto-create an array. but even perl (5) refuses to coerce a scalar into an array
bsdz rg: it would be useful if there was some way that PCT allowed you to specify what happened. like a morphbase parameter. 21:18
in M-script (matlab/octave/freemat) it does exactly this. 21:19
particle- bsdz: a custom pmc can implement this behavior 21:20
21:20 Whiteknight joined
particle- much better than a compiler 21:20
bsdz particle: i'm not sure how. the pmc has no control on how/when PCT calls vivibase. 21:21
particle- no, the emitted vtable call on the pmc, get_int_keyed...
bsdz ah i see. override the vtable method
particle- is implemented in the custom pmc to recognize, oh, he wants a new dimension, let's create one and upgrade 21:22
bsdz i mean the vtable function
particle- yep.
bsdz i was hoping to stick to pure parrot types mind you
particle- you'll need a RPA of RPAs 21:25
ResizablePMCArray that is
bsdz yes. i've got a routine to create an n-dim rpa already
21:28 braceta left
Coke OOC, in your example, do you still expect to have a [1] when you've set [1][2] ? 21:29
bsdz Coke: yes. (what does ooc mean?) 21:30
particle- out of curiousity
Coke ooc? 21:33
purl rumour has it ooc is Out of curiosity. or out of commission or out of control or out of cocaine or out of context or Out of Character
bsdz i dig purl! :)
Coke hurm. so not only does your array have multiple dimensions it both 1D /and/ 2D /and/ ... ND ? huh.
I am more used to fixed-dimensional arrays, but then, I've been staring at CFMX for months. 21:34
bsdz yep. manage to do the n dimensional stuff with a recursive tree algorithm. tbh that's all working
particle- i'm still getting over that fortnight i spent in the eight-and-a-halfth dimension 21:35
bsdz sounds trippy! i think partial dimensions have something to do with fractals
21:36 davidfetter joined
Coke ayup. 21:40
not that math has anything to do with reality, mind you. 21:41
bsdz plato might disagree
particle- i beat plato in a verbal debate last night. 21:42
bsdz it's not a fair challenge when the other party speaks ancient greek 21:43
particle- i can sing better than elvis, too.
cotto and is dead
Coke poo EE-ne ee too-ah-LEH-tah? 21:45
21:46 Limbic_Region joined
cotto Are there any gotchas about branching/merging that aren't covered in docs/project/branching_guide.pod? 21:49
21:51 Whiteknight joined 21:56 contingencyplan joined
dalek tracwiki: v6 | cotto++ | BranchDescriptions 21:58
tracwiki: updating list of branches, descriptions still needed
tracwiki: trac.parrot.org/parrot/wiki/Branch...ction=diff
shorten dalek's url is at xrl.us/beozhh
mj41 Hi. Seems like TapTinder is a little bit more useful. tt.ro.vutbr.cz/buildstatus/pr-Parrot/rp-trunk 22:10
rg is about to give up on the current jit. is there anyone around who could help me? 22:11
nopaste "bsdz" at 78.105.78.157 pasted "pass by ref on derived type seems to drop data" (29 lines) at nopaste.snit.ch/16270
bsdz a little puzzle in my nopaste. seems that a derived parrot class loses data when you assign to it in a pass by ref. any one see me doing something wrong? 22:12
particle- bsdz: what is setref p, p supposed to do? 22:15
bsdz it sets p to p as past by reference
basically points one to the other. i use it in my pir subs to make a var pass by ref usually. 22:16
particle- um. 22:18
src/pmc/fixedpmcarray.pmc, line 386
VTABLE void set_pmc(PMC *value)
if (SELF == value) return;
...which is exactly what i would expect. it's a noop 22:19
particle- heads home & 22:20
bsdz i can see that if p[0] == x and you try setting p[0] = x then should be no op. but in the example i pasted. seems to be different behavoir for base class to derived class?
good night particle 22:21
pmichaud bsdz: PCT isn't intended to directly support every operation one might imagine. 22:27
instead, it provides common primitives to make whatever operation might be wanted
so, in the case of attempting to convert a scalar to an aggregate, either the PMC itself should do it, or the thing generating the PAST should do the checks for an aggregate prior to attempting the assignment 22:28
It's also possible to create custom PAST node behaviors (derive from PAST::* and override the behavior)
one could also set an exception handler that traps the "get_pmc_keyed in class Integer" and decides how to handle it. 22:29
bsdz okay thanks pmichaud. i think i'll try a custom pmc for now. it would be nice if PCT had a hook already. actually the EH idea sounds interesting. it would need to be sure it was getting thrown from the right place. would that be done in PCT? 22:30
pmichaud I'm not sure... we haven't explored that part of PCT yet 22:31
at any rate, in general PCT (and parrot) doesn't like the "check if an operation is valid prior to attempting it"
usually we go with an exception-based model.
not only that, there's not a reliable way to determine if a given PMC is an aggregate or understands the set_pmc_keyed, short of actually doing it. 22:32
(which is why PCT leaves it up to the HLL compiler to decide when a check is needed and how to handle it)
22:32 Xrblackwe joined
bsdz thanks. that's very useful. 22:32
pmichaud decides to wander around the airport a bit to find a power outlet. 22:33
bsdz i seem to recall seeing exceptions being used for returning values.
pmichaud yes.
that's pretty standard.
bsdz standard as in dynamic languages i guess? i recall c/c++ shoving everything back on the stack 22:34
Tene_ pmichaud: did you see my issues with HLL line numbers recently? 22:37
pmichaud tene: no
Tene_ pmichaud: I created a branch for it, but changing that message would break many tests.
pmichaud: I'd like confirmation that Parrot wants it changed before cleaning up many tests. 22:38
maybe I should implamant it at the PCT or Rakudo level for now?
argh spelling. 22:39
EKEYBOARDWAYTOOSMALL
pmichaud perhaps have it change in Parrot in response to a switch?
so that Parrot by default uses its traditional error message, but if some global switch is set it uses the new error message? 22:40
we could potentially do it in PCT as well... but having Parrot do the basic reporting seems cleaner.
Tene_ hmm 22:41
pmichaud bsdz: well, doing ".return" in a Parrot Sub doesn't result in an exception. The exceptions are currently used to handle the case where a HLL subroutine has nested blocks (i.e., nested Parrot Subs)
Tene_ and PCT turns on that switch?
22:41 bacek joined
pmichaud Tene_: yes, or the HLL could. 22:41
(it would probably make sense for PCT to do it by default)
bsdz: that said, I'd really like us to extend Parrot a bit so that we could easily invoke a Sub's return continuation (returning values) without having to do exceptions. 22:42
Tene any idea where to store that flag?
bsdz ah yes. that makes perfect sense now. quite good too since we need to implement nested functions at some point.
pmichaud Tene: no, it was just a thought to avoid having to develop additional PCT code to do it.
Tene 'kay 22:43
pmichaud Tene: do lots of tests check for the specific text of the error message?
Tene Yes. 22:44
pmichaud ick.
oh well.
another possibility might be to change _only_ if annotations are detected.
i.e., if we have an annotation, then use the new report, otherwise fall back to the old style.
Tene That was behaving oddly... I was getting a 0 in some cases instead of pmcnull, it looked like. 22:45
bsdz pmichaud,tene: on a different topic. i'm thinking of filing a trac on setref and derived aggregate classes. i posted a nopaste a little earlier. can either of you see if there's something i;ve done fundamentally wrong before i file it?
Tene Was almost sleephacking at that point, though.
pmichaud bsdz: I've never seen definitive documentation for what setref is supposed to do.
Tene need to re-check.
pmichaud I think setref is PMC specific.
(i.e., it ends up calling the set_pointer vtable) 22:46
bsdz yes. usually works to pass by ref. seems to break if i use a derived aggregate pmc type. i better file a trac.
pmichaud there are a number of problems with derived pmc types in Parrot still, I think.
bsdz okay thanks again. 22:47
Tene pmichaud: I'll try keeping it the same if no annotations are found. Think you could get filename and HLL annotaions in PCT soon? 22:50
pmichaud filename, yes.
hll, yes. Not exactly sure what that should be -- I guess a special PAST::Node for that
there's always :inline(...) :-P 22:51
PAST::Op.new(:inline('.annotate "foo", "bar"'))
Tene :pirop('annotate') 22:52
pmichaud I try to avoid too many special pirops
especially since annotate isn't really an op.
Tene ah
right
pmichaud it kind of feels like a PAST::Val or PAST::Var somehow 22:53
since it has name+value attributes
I'll think about it on the plane.
Tene I'm not sure we need HLL 22:54
pmichaud if there's an easy way to insert annotation nodes into the PAST, then having an HLL compiler insert HLL and filename information is pretty straightforward.
Tene pmichaud: should fatal exceptions print a full stack trace, and resumed warnings just source? That's my current plan. 22:55
pmichaud I wonder if fatals should print the full trace.
right now the major complaint I hear is that the traces are too long. 22:56
Tene ah
pmichaud TimToady says that the error message should contain a url to the trace :-)
Tene file://
is doable
pmichaud if the traces can be made shorter (and aren't obscure) then I think it's okay to include them. 22:57
Tene one big issue is printing a line for evessue is printing a line for every nested block. 22:58
pmichaud perhaps we should annotate whether blocks get an entry in the stack trace 22:59
Tene we could add a 'sub' annotation. 23:00
pmichaud yes, something like that.
purl something like that. is, like, that true
Tene this reminds me of the 'inline' attribute of contexts mentioned in the P6 specs. 23:01
pmichaud yes, I suspect tehre are some similarities.
okay, I think I need to move closer to my gate -- they're starting to call people to board. 23:02
I'll catch you all later/tomorrow/whenever.
Tene seeya 23:03
23:07 tetragon joined
nopaste "coke" at 72.228.52.192 pasted "why is this a c90 error?" (11 lines) at nopaste.snit.ch/16271 23:09
chromatic Coke, no trailing semicolon? 23:11
Coke_afk I thought UNUSED didn't require one. checking...
Coke nope. semicolon doesn't help. 23:12
rg can't hurt and looks less wrong
Coke rg; I'm following the style elsewhere in the codebase.
rg ok, i guess it wouldn't hurt elsewhere either ;)
so it complains about the UNUSED line? 23:13
Coke it's complaining about the following line.
rg is the following line a declaration? 23:14
nopaste "coke" at 72.228.52.192 pasted "patch that result in error" (14 lines) at nopaste.snit.ch/16272
rg ok, so it's correct 23:15
you need to put the UNUSED last (behind the declarations)
it expands to 'if (0) (void)(a);' i.e. code.
Coke bah. you're right.
rg++
rg (and you're right it doesn't need the semicolon, but imho it still looks better with one) 23:16
bsdz not too keen on languages that insist declarations are always at the top
Coke welcome to C. =-)
bsdz tell me about it
purl it is, like, a hack
rg is giving up the fight with jit. a workaround is all i will offer (if it's going to pass all tests) 23:17
bsdz is there the equivalent of SUPER in pir?
Coke prepares to commit a bunch of non-functional changes. :| 23:21
23:27 TiMBuS joined
bsdz good night all 23:29
dalek rrot: r38107 | coke++ | trunk (25 files):
[cage] reduce -unused-parameter warnings (*)
 23:44
23:56 bacek joined 23:58 megazooch joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!