Camelia, the Perl 6 bug

IRC log for #parrot, 2010-08-02

Parrot | source cross referenced

| Channels | #parrot index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:00 Austin "And my escape will be a piece of cake, 'cause nobody's gonna get in the way of a giant mass of angry dogs walking down the street."
00:00 Austin "Especially if it's carrying a urinal."
00:00 Austin "Exactly."
00:00 whiteknight :)
00:00 Austin This comic reminds me a bunch of partially clips.
00:01 Tene 'k, I'll try to get to that tonight.
00:08 Austin Who runs Bartertown?
00:09 Austin purl, who runs bartertown is <reply>Master Blaster!
00:09 purl i haven't a clue, austin
00:09 Austin :(
00:09 Austin You're no fun at all.
00:37 Coke joined #parrot
00:46 cxreg so is it a known issue that you have to have the "-dev" version of a library installed to NCI?
00:46 cxreg (that is, the .so symlink)
00:54 sorear known, yes
00:54 sorear considered an issue, no
00:55 tcurtis cxreg: if you don't have the .so, how would you load the library?
00:55 sorear the only time you can get away with not installing the symlink is when you're using a static compiler
00:55 * tcurtis probably misunderstands.
01:00 cxreg uhm
01:00 cxreg no, most linux distributions do /not/ install that symlink by default
01:01 cxreg for example:
01:01 cxreg libsqlite3-0: /usr/lib/libsqlite3.so.0.8.6
01:01 cxreg libsqlite3-dev: /usr/lib/libsqlite3.so
01:02 tcurtis cxreg: you can do that.
01:03 cxreg so you can dlopen "libsqlite3.so.0.8.6", but it requires you to know the exact version, and "libsqlite3.so" will not work without that dev package installed
01:03 * cxreg shrugs
01:09 sorear cxreg: debian policy in that regard is a fossil.  As more people start using FFIs, it'll have to be changed
01:12 cxreg it won't
01:13 cxreg well, maybe
01:13 cxreg the .a and .h packages will still be in -dev, but i guess they might move the .so symlink
01:13 * cxreg gets Gtk2.pm6 working, w00t
01:22 rurban_ joined #parrot
01:25 tcurtis sorear: hopefully LLVM will learn that lesson soon as well.
01:52 dalek parrot: r48247 | jkeenan++ | branches/tt1725_headerizer_documentation:
01:52 dalek parrot: Creating tt1725_headerizer_documentation in �https://svn.parrot.org/parrot//branches
01:52 dalek parrot: review: http://trac.parrot.org/parrot/changeset/48247/
01:52 dalek parrot: r48248 | jkeenan++ | tags/tt1725_headerizer_documentation-48246:
01:52 dalek parrot: Tagging trunk at r48246 so that the tt1725_headerizer_documentation can later be synched to it.
01:52 dalek parrot: review: http://trac.parrot.org/parrot/changeset/48248/
01:59 kid51 joined #parrot
02:09 dalek parrot: r48249 | jkeenan++ | branches/tt1725_headerizer_documentation (2 files):
02:09 dalek parrot: As a step toward writing better docs for the headerizer, write some tests for Parrot::Headerizer.
02:09 dalek parrot: review: http://trac.parrot.org/parrot/changeset/48249/
02:14 Austin msg whiteknight Of course, the *best* part of this whole namespace method crap is the part where adding methods to a class doesn't work because you also have to add the methods to the proto-class. And we're all about the encapsulation, right?
02:14 purl Message for whiteknight stored.
02:35 janus joined #parrot
03:06 TiMBuS joined #parrot
03:41 LoganLK joined #parrot
04:39 s1n joined #parrot
05:54 theory joined #parrot
05:59 uniejo joined #parrot
06:00 davidfetter joined #parrot
06:35 NotFound joined #parrot
06:35 NotFound hi
06:35 sorear Hi
06:35 purl hi, sorear.
06:59 Tene purl: msg austin please explain about adding methods to a class not working.
06:59 purl Message for austin stored.
07:07 wayland76 joined #parrot
07:11 eternaleye joined #parrot
08:11 robin-gvx joined #parrot
08:25 fperrad joined #parrot
08:57 AndyA joined #parrot
09:09 clinton joined #parrot
09:12 lucian joined #parrot
09:19 bacek joined #parrot
09:21 rurban_ joined #parrot
09:36 [1]Casan joined #parrot
10:01 gbacon joined #parrot
10:01 dalek rakudo: f8b73df | moritz++ | src/core/Match.pm:
10:01 dalek rakudo: Match.new() (no subcaptures yet)
10:01 dalek rakudo: review: http://github.com/rakudo/rakudo/commit/f​8b73df893066ae15b788e116baa52dbf13b463e
10:53 PacoLinux joined #parrot
11:28 bacek joined #parrot
12:03 ruoso joined #parrot
12:07 bluescreen joined #parrot
12:08 daxim joined #parrot
12:10 daxim the parrot-config from package parrot-devel-2.6.0-16.1 on openSUSE 11.3 gives me the config item revision == 0
12:10 daxim is this supposed to be this way?
12:13 kid51 joined #parrot
12:13 * kid51 reads backscroll
12:13 kid51 daxim:  I believe that situation is correct.
12:14 kid51 If you built from the monthly release, what's important is the version, in this case, 2.6.0
12:14 kid51 If you were building from Subversion HEAD, the revision number would be key
12:15 kid51 It would be the Subversion revision number
12:15 daxim since this is a RPM, it's reasonable safe to say it's from the 2.6.0 release tarball
12:15 kid51 Correct
12:15 kid51 So, nothing to worry about.
12:15 daxim now the problem is that in rakudo-star's Configure a check against this number revision is made
12:16 daxim I want to build & package rakudo-star against the external parrot, not the bundled one
12:17 daxim I'll go over there again and report it
12:17 daxim kid51++ for the prompt answer
12:23 * kid51 to $job
12:31 NotFound daxim: I think that the purpose of rakudo star is being a full package, so building it with a different parrot is pointless.
12:31 daxim not when I'm going to package this as RPM
12:32 daxim if I used the bundled parrot, it would conflict with the system parrot and a user would need to decide between rakudo-star and parrot+its ecosystem for no good reason
12:33 moritz daxim: it *should* work with an installed, non-svn parrot
12:33 moritz if the version check succeeds, the revision number check shouldn't be executed
12:34 moritz if that's not the case, file a bug report, and patch the configure script
12:34 daxim refer to http://paste.scsys.co.uk/47478 ; I'll report this when I have a working patch
12:34 dalek parrot: r48250 | NotFound++ | trunk/t/pmc/opcode.t:
12:34 dalek parrot: improve test Opcode cannot_create_directly
12:34 dalek parrot: review: http://trac.parrot.org/parrot/changeset/48250/
12:34 dalek parrot: r48251 | NotFound++ | trunk/t/pmc/oplib.t:
12:34 dalek parrot: one more OpLib test
12:34 dalek parrot: review: http://trac.parrot.org/parrot/changeset/48251/
12:41 preflex joined #parrot
13:01 bacek joined #parrot
13:02 Coke Yah, I also do not understand why people are trying to re-dist the * dist.
13:02 * Coke thinks it would be better to just bundle parrot, the compiler, and "the modules" as 3 different things.
13:07 dalek parrot: r48252 | NotFound++ | trunk/src/io/api.c:
13:07 dalek parrot: don't mention other PMC in TODO comment for FileHandle
13:07 dalek parrot: review: http://trac.parrot.org/parrot/changeset/48252/
13:12 Coke (that is, in addition to the R* release, which is for users, not distributors.)
13:16 macroron joined #parrot
13:24 dalek parrot: r48253 | NotFound++ | trunk/src (2 files):
13:24 dalek parrot: avoid special case for StringHandle in Parrot_io_close and Parrot_io_is_closed
13:24 dalek parrot: review: http://trac.parrot.org/parrot/changeset/48253/
13:24 whiteknight joined #parrot
13:41 dalek rakudo: 65eb876 | moritz++ | src/core/Any-list.pm:
13:41 dalek rakudo: attempt to produce more awesme error message when you do map { hash => 1}
13:41 dalek rakudo: review: http://github.com/rakudo/rakudo/commit/6​5eb876cb4284a3d96c5910e21a2153ce55c5535
13:41 dalek rakudo: 8f8c519 | rurban++ | build/Makefile.in:
13:41 dalek rakudo: Fix installation on cygwin
13:41 dalek rakudo: Signed-off-by: Moritz Lenz <moritz@faui2k3.org>
13:41 dalek rakudo: review: http://github.com/rakudo/rakudo/commit/8​f8c5195b1fc04fab17799825bcea54e860a8c1b
13:41 dalek rakudo: 1a5d4a3 | (Christopher J. Madsen)++ | build/gen_core_pm.pl:
13:41 dalek rakudo: Explicitly request :utf8 layer in gen_core_pm.pl to fix RT #76856
13:42 dalek rakudo: Signed-off-by: Moritz Lenz <moritz@faui2k3.org>
13:42 dalek rakudo: review: http://github.com/rakudo/rakudo/commit/1​a5d4a38470a07dbaaaf2d9a5042a8f97637d6a1
13:51 zostay joined #parrot
13:57 whiteknight good morning, #parrot
13:58 plobsing joined #parrot
13:59 whiteknight purl msg Austin: I hadn't even thought about the metaclass, that might explain some of the errors I am still seeing. Encapsulation is just a crutch good programmers use, but if you're bad enough you don't need it
13:59 purl Message for austin stored.
14:51 Andy_ joined #parrot
14:59 jdv79 joined #parrot
15:09 jsut joined #parrot
15:19 bluescreen joined #parrot
15:25 gbacon joined #parrot
15:25 jsut_ joined #parrot
15:36 bluescreen joined #parrot
15:36 tcurtis joined #parrot
15:37 pyrimidine joined #parrot
15:40 davidfetter joined #parrot
15:45 mbp joined #parrot
15:45 mbp whois mbp
15:45 purl you are becoming the ian+emily post-bowling snack or MacBook Pro
15:58 * Coke doesn't recognize any of the names responding in the current parrot-dev thread.
16:08 Casan was rakudo* based on the same version as atlanta #31, or did r* also include additional commits after #31 was released ?
16:08 tcurtis "This is Rakudo Perl 6, version 2010.07-47-g9fd5eaa built on parrot 2.6.0"
16:08 tcurtis So, the latter, if I understand that correctly.
16:10 Casan let's assume that. thnx.
16:12 Coke yes, the latter.
16:12 tcurtis I think it was initially planned to just use Atlanta, but I guess something came up that necessitated not doing so.
16:12 Coke as I did the release for atlanta, and was sad to hear of bugs. =-)
16:12 tcurtis Hopefully(for convenience of redistributors), that won't be the case next month.
16:14 Casan ok good. I was just installing strawbery perl 5.12.1 and padre, and now I wanted to add parrot/rakudo, and wanted to get the latest release.. so it was parrot2.6.0+atlanta vs. r* .. now I'll try adding the perl6 plugin and r* and see if everything goes smoothly.
16:15 Casan problems are good, they are opportunities. problems we know about are even better, can pave the way for solutions.
16:15 Coke casan - I think Alias is working on a super-bundle with everything.
16:16 Coke (strawberry, parrot, rakudo, R* extras...)
16:16 Casan Coke: yes if not outsourced to csjewel.. Until then, I like to get a bit more familiar with the toolchain myself.
16:17 Austin joined #parrot
16:17 jdv79 anyone remember where mpeters hangs out on irc?
16:17 Casan a superbundle with everything, including an improving blizkost, and embedded update mechanism.. that'll be yummie for starters.
16:22 Austin seen tene?
16:22 purl tene was last seen on #parrot 9 hours, 23 minutes and 11 seconds ago, saying: purl: msg austin please explain about adding methods to a class not working.
16:23 Coke seen mpeters?
16:23 purl I haven't seen 'mpeters', Coke
16:23 Austin msg Tene To add a method to a class pmc, you need to use the pmc's .remove_method and .add_method methods. But to add a method to a p6object class, you have to add the method to the "parrotclass' and also add it to the "protoclass" so the protoobject will respond to it.
16:23 purl Message for tene stored.
16:24 Austin msg Tene (especially if the method in question is 'new')
16:24 purl Message for tene stored.
16:25 Tene 'k, thanks.
16:28 jdv79 Coke: thanks
16:36 theory joined #parrot
16:36 Casan when installing the R* on windows using the .msi I would expect it to automatically get added to path, so it is possible to open a prompt and type perl6... but I have to run c:\Rakudo\bin\perl6.exe - Is it deliberate that the users shall add it to the path manually?
16:39 Coke Casan: please ask in #perl6 on freenode.
16:40 Coke You'll probably get a faster, saner, more targetted response to R* questions there.
16:49 tommyd joined #parrot
16:52 Casan Coke: thanks.
16:54 cxreg joined #parrot
16:58 whiteknight Austin: I would make the suggestion that Parrot's object model should have built-in support for pluggable meta-class objects, and certain calls to the class should get mirrored to the metaclass automatically, if it exists
16:59 Austin Heh
16:59 Austin Sure.
16:59 Austin But it's just a simple matter of incredibly painful programming.
17:00 Austin I'd much rather everyone rewrite pmcs as derived from object.
17:00 Austin But it's the protoclass, not the metaclass, that's causing problems.
17:01 Austin There's this thing called the proto-object.
17:01 Austin The proto-object is the object that lives in the namespace: Foo::Bar is a global variable that holds a proto-object pmc.
17:01 Coke ... that's not used by parrot core, but by NQP & rakudo?
17:01 Coke or do you not mean the that proto-object?
17:02 Austin Coke: This is the "p6object.pir" way of doing things. I can't speak to Rakudo, although I believe it started this way and has drifted since.
17:02 Austin Anyway:
17:02 tcurtis Austin: Does Foo.HOW.add_method(Foo, "name", &methodref) add the method just to the instances?
17:03 whiteknight oh, I misunderstood. I guess I didn't realize P6object used a protoobject
17:03 whiteknight what's the purpose of it?
17:03 purl the purpose of it is probably to make it easier to maintain.
17:03 Austin During the class declaration process, a new class (the "parrotclass") is created. It gets methods and attributes. Then a second class is created, with both metaclass and $parrotclass as parents. This is the protoclass.
17:04 whiteknight ok. That's the "how", but not the "why"
17:05 Austin The protoclass has different semantics because it gets .new() where the parrotclass might not have it, etc. This protoclass is used as the class of the protoobject. The protoobject is then installed as Foo::Bar, so when you say Foo::Bar.new it works, and returns an instance of the $parrotclass due to how the .new method is written.
17:05 Austin tcurtis: Don't know. Bear with me for a moment.
17:05 whiteknight Ah, I see
17:05 Austin Except ...
17:06 Austin In order to try to get PMC types to interoperate with this, the P6metaclass.register method accepts their names as well. But PMC Proxies (pmcproxy.pmc) don't do the whole attribute thing, so the code that links the parrotclass, the metaclass instance, the protoclass, and the protoobject together can't do it for proxies (no attributes).
17:09 Austin So trying to install methods on pmc types is a bit of a challenge now, because the namespace .get_class returns the pmcproxy, but there's no way to connect from the proxy to the protoclass. I have to restart with the global symbol approach, which of course won't work for anoynmous classes (no global symbols).
17:09 Austin Hence, "simple matter of incredibly painful programming"
17:10 whiteknight I think I missed a point: Why don't pmcproxies support attributes?
17:10 whiteknight or, properties, or whatever? Why doesn't (can't?) the pmcproxy contain a link to the protoobject?
17:11 Austin tcurtis: The Foo.HOW method returns the metaclass object. The metaclass.add_method method fetches the parrotclass and adds the method to it. It ignores the protoclass. (This isn't bad unless there's a conflicting method, which in the case of 'new' there definitely is.)
17:12 cotto_work2 joined #parrot
17:12 cotto_work2 ~~
17:13 Austin whiteknight: It's a pmc. Clearly pmc types have no need for attributes, because <cough><mumble><harrumph>
17:13 smash joined #parrot
17:13 smash hello everyone
17:13 tcurtis Austin: why do staticly-defined "new" methods not suffer the same problem?
17:13 Austin whiteknight: Or, to put it another way: get_attr_str() not implemented in class 'PMCProxy'
17:14 Austin tcurtis: Magic.
17:14 radu_ joined #parrot
17:14 whiteknight Austin: ah, okay. I thought the metaclass would be stored as a property though, not an attribute?
17:15 Austin The process of creating the protoclass looks for a method? named PROTOOVERRIDES and that returns a list, which by default is [ new ] of methods that should be special cased.
17:15 Austin whiteknight: That may be the case.
17:15 Austin My eyes are so blurry at this point that anything is possible. Lemme check.
17:18 tommyd joined #parrot
17:19 Austin Okay, whiteknight, you're right. There's a property (not attribute) set on parrotclass -- 'metaclass' --> how, then $how has an attribute 'protoobject', then $protoobject.get_class should do it. Let me see if that works.
17:20 Austin Why is it "getattribute" but "getprop"
17:20 Austin Why is it "getattribute" but "get_class"
17:20 Austin ??
17:21 particle because the api has been ad-hoc designed
17:22 rurban_ joined #parrot
17:22 particle iirc _ are preferred, but that came later, so they're not implemented everywhere
17:23 Austin Bah. We've got several thousand lines of testcase code to ensure that every c function has a useless header attached, but the part that people have to type has no consistency.
17:23 * Austin wants some cheese to go with his whine...
17:24 Austin Woo-hoo. Null PMC access.
17:24 Austin Now we're making progress...
17:24 * cotto_work points Austin at either trac or DEPRECATED.pod
17:24 Austin cotto_work?
17:24 purl rumour has it cotto_work is always at work
17:24 Austin botsnack
17:24 purl thanks Austin :)
17:25 cotto_work cotto_work is also cotto's employed alter ego
17:25 purl okay, cotto_work.
17:28 Austin And why is it getattribute $pmc, 'name', but it's getprop 'name', $pmc?
17:28 atrodo to make the php people feel at home?
17:29 Austin atrodo++
17:29 Austin Best answer yet.
17:29 cotto_work atrodo++
17:30 Austin Frailty, thy name is parrot.
17:34 cotto_work The question now is whether it's worth the churn to get those names consistent.
17:38 Austin TT#1727
17:38 ambs joined #parrot
17:38 Austin cotto_work++
17:39 whiteknight Strong +1 on #1727
17:41 cotto_work The deprecation notice will be simple.
17:41 Austin The problem is changing imcc, PAST, etc.
17:44 Austin Oh, and what's the difference between $P0 = get_class $P1 and $P0 = class $P1
17:44 Austin ??
17:44 cotto_work Parrot_oo_get_class_str vs VTABLE_get_class
17:44 cotto_work (no idea)
17:44 Austin Of course.
17:44 purl Indubitably.
17:45 Austin Well, opcodes are like handbags. It's good to have choices..
17:45 whiteknight get_class looks up a class by key. class calls the VTABLE_get_class to get the class object of the given PMC
17:45 Austin Working on parrot is making me metrosexual...
17:46 Austin Ah
17:46 Austin So the $P1 in the get_class is really [ Foo ; Bar ] or something?
17:46 whiteknight class "Foo" == String. get_class "Foo" == Foo
17:46 whiteknight right
17:46 Austin whiteknight++
17:46 whiteknight naming obviously leaves something to be desired
17:47 Austin Hell, that one almost makes sense.
17:47 Austin I'm just coming at it from the wrong angle.
17:48 dalek TT #1727 created by Austin++: Rationalize get/set opcodes
17:48 dalek TT #1727: http://trac.parrot.org/parrot/ticket/1727
17:48 Austin Stupid english language.. every operation winds up being named "get"
17:48 Austin Woot! No method named 'can' to remove in class 'NameSpace'.
17:49 Austin less $_PMC/class.pmc
17:49 Austin oops
17:49 Austin ww
17:49 tcurtis Austin: in this case, get_class's naming does kinda make sense. It's similar to the get_*global family of ops.
17:50 Austin tcurtis: Agreed. See above.
17:51 Austin It's the fact that english maps "look up by name" to "get" and "look up attribute of this object" to "get".
17:52 tcurtis True. Maybe class should be classof to parallel typeof. Or even better, class_of.
17:53 Austin That's a simple, workable change.
17:54 Austin (IMO, "get" should have been reserved to pmc/objects, and "find" or something used for the name lookups. But that would be a horrific change at this point.)
17:54 tcurtis Austin: well, if we are to change things, we might as well change them drastically all at once.
17:55 Austin Umm, no.
17:55 Austin All at once, yes. Drastically is not something I'm really happy with.
17:56 whiteknight I like "classof"
17:57 Austin Me, too.
17:57 whiteknight but if we're standardizing on using underscores as Austin suggests, it would be "class_of"
17:57 whiteknight (and "type_of")
17:57 Austin Also ok.
17:57 whiteknight but actually, doesn't typeof_p_p do the same thing?
17:58 Austin Heh
17:59 Austin typeof_pp is $1 = vtable-get-class $2
17:59 Austin whereas class_pp is $1 = vtable-get-class $2
17:59 Austin Totally different.
18:00 tcurtis Austin: I should clarify. I don't think the negative effects of renaming any given opcode to make word-separation and abbreviation(or lack thereof) more consistent is significantly less than changing the name more drastically if we provide a good way of migrating(especially since we could probably provide a script to do it automatically).
18:00 tcurtis Changing setattribute to set_attribute isn't much different in terms of migration difficulty than get_hll_global to find_hll_global.
18:01 Austin whiteknight: After all, 'class' has the :object_classes modifier, while typeof does not. See?
18:03 Austin tcurtis: I understand. But I don't agree. I suspect that proposing the latter change (get* -> find*) will generate significant pushback, and I don't want the other bits (tt#1727) to be blocked because of that.
18:04 Austin Rather, *suspect = fear
18:06 tcurtis Austin: I agree with that. I just think that if we can decide on a semantically, as well as syntactically, consistent naming convention, it wouldn't be significantly harder for people to migrate to. That's a significant if, though.
18:06 Austin +1
18:06 purl 1
18:07 ambs purl: shut up
18:07 purl ;-(
18:07 Austin whiteknight: What are the chances of getting some kind of parrot-interp level automatic pbc load?
18:10 whiteknight Austin: what do you mean by that?
18:11 whiteknight shut up purl
18:11 purl ;-(
18:11 Austin Like, the interp pretends that the first opcode it sees is load_bytecode "parrot.pbc"
18:12 Austin Totally ignoring Kakapo, I've had this idea for a while that a big percentage of the vtables and methods for the existing pmcs could be written in pir.
18:13 Austin There's a few that obviously shouldn't - like the pcc stuff. But so much of the code is "check this, throw exception, else call that"
18:14 Austin So if a boatload of the C could be replaced by pir, I think there would be a lot more information available for the performance guys, vis a vis what needed to be fast. Likewise, the lorito project.
18:15 ambs_ joined #parrot
18:15 atrodo karma lorito
18:15 purl lorito has neutral karma
18:15 Austin lorito--
18:15 Austin karma lorit0
18:15 purl lorit0 has neutral karma
18:15 Austin karma lorito
18:15 purl lorito has karma of -1
18:16 atrodo awww, poor little parrot
18:17 Austin Anyway, if such a thing was possible, it would need to be loaded pretty much before any command line processing took place. Hence my question.
18:19 bubaflub joined #parrot
18:21 tommyd joined #parrot
18:43 jevin joined #parrot
19:09 Coke Are there any command line irc clients that will interleave discussions? (show sends from multiple channels in the same window)?
19:12 szbalint bitchx
19:13 szbalint http://www.bitchx.com/
19:13 dalek rakudo: 9667975 | moritz++ | src/core/Any-list.pm:
19:13 dalek rakudo: fix typo, noticed by madsen++
19:13 dalek rakudo: review: http://github.com/rakudo/rakudo/commit/9​667975fceefcb601fe55d52baf4bd562f178b36
19:19 hercynium joined #parrot
19:20 gbacon joined #parrot
19:29 Coke szbalint: danke. I find the lack of docs a bit scary, but I'll check it out.
19:29 tommyd any Mac OS X related person around? It would be cool to get some help to fix http://trac.parrot.org/parrot/ticket/344
19:30 tommyd I'd love to fix that issue in the parrot port and finally submit my rakudo port...
19:32 Coke tommyd: the fix uses a post-install hack to fix the paths, that's been part of the macport, not 'make install'.
19:32 Coke I suppose it could be moved to a custom OS X install task.
19:33 Coke I suspect I'm the only person with a mac and appropriately shaped tuits atm. I'll take a look later this week.
19:33 tommyd Coke: unfortunately I don't think this is the proper fix for that
19:33 tommyd the problem is that parrot_config still echoes the wrong paths
19:34 tommyd and from what i can see parrot_config is a binary which spits out these paths hardcoded
19:34 Coke tommyd: ok. I don't see anything about parrot_config in the ticket.
19:34 Coke perhaps you could update it with your latest intel?
19:34 tommyd intel?
19:34 purl intel is making a killing on that chip
19:35 PerlJam intelligence, knowledge
19:35 Coke intelligence. information about the current situation.
19:35 tommyd ah - sorry, I'm not a native speaker :)
19:35 PerlJam stuff you know that no one else does
19:35 Coke sokay. that's more of a military connotation.
19:35 Coke but yah, if you can update the ticket, that'll help when I finally do have tuits.
19:36 estrabd joined #parrot
19:40 tommyd done
19:41 theory joined #parrot
19:43 Coke I would be interested to compare the dump of those values on linux.
19:43 Coke or some other non-OS X platform.
19:44 Coke szbalint: bitchx segfaults for me on os x. =-)
19:44 Coke ah well.
19:44 tcurtis Coke: sounds like the perfect client for a Parrot dev, then. :)
19:45 Casan joined #parrot
19:46 szbalint =)
19:49 tommyd Coke: I looked at the other files in the hints/ directory and nowhere else is build_dir related path mungling done
19:49 tommyd then, on the other hand, I don't know of many other distros who make use of installing to DESTDIR and linking to prefix
19:51 Coke as I recall, linux avoids this with rpath, but that didn't JustWork on osx.
19:52 tommyd but using `$lib_dir = $conf->data->get('build_dir') . "/blib/lib";` as path prefix for the install name sounds definitely wrong
19:52 tommyd PREFIX/blib/lib isn't even installed
19:52 Coke may just need to rip out code that was improperly added for osx 8 years ago.
19:52 particle rpath-- runpath++
19:53 tommyd so my original thought of replacing build_dir by prefix did not work either
19:53 tommyd Coke: you can expect most people on 10.5 and 10.6
19:53 Coke tommyd: I'm sorry, what?
19:54 Coke was that a response or just a helpful note?
19:54 Coke (yes, I'm not going to support 10.4 =-)
19:54 tommyd merely a note
19:54 tommyd right ;)
19:54 Coke I can't even support 10.5 at this point.
19:54 Coke but I can duplicate your problem on 10.6, so...
19:54 tommyd whats bad with 10.5?
19:54 tommyd just that you can't test it?
19:55 tommyd I'm still on 10.5.8
19:57 Coke tommyd: I have no 10.5 to test on.
19:57 Coke so I'll make this work on 10.6 and cross my fingers for you. =-)
19:57 tommyd thanks an awful lot for this ;)
19:58 Coke particle: aren't we using rpath for linux?
19:59 particle you're asking the wrong guy.  all i know is, you can't override it, whereas with runpath, you can.
20:13 cotto_work seen khairul
20:13 purl khairul was last seen on #parrot 6 days, 13 hours, 28 minutes and 48 seconds ago, saying: cotto: sorry bout that. done.  [Jul 27 06:44:32 2010]
20:35 cotto_work Coke: ping
20:37 Coke pong
20:37 Coke cotto_work, pong, even.
20:37 Coke in case you're lazy.
20:38 Coke and cotto.
20:38 cotto_work I'm trying to figure out the best way to gauge the opinions of parrot hackers in order to decide about where we'll host our most official git repo and how source code browsing will work.
20:38 smash joined #parrot
20:38 cotto_work Can you think of something better than a message to parrot-dev with a request to not reply-all?
20:39 cotto_work (er, not reply all when voting for an option, mainly to keep the noise down)
20:39 Coke any reason NOT to reply all? we only have about, what, 30 core committers atm?
20:39 cotto_work noise is all
20:39 Coke you could make a survey on parrot.org
20:40 Coke but I don't think an open solicitation of feedback is in your best interest. =-)
20:40 Coke is the gittrac bridge in place anywhere?
20:40 cotto_work you may be correct
20:40 cotto_work http://trac.parrot.org/parrottest/browser
20:41 cotto_work I'm not a fan, mainly because it's so slow.
20:41 Coke http://trac.parrot.org/parrottest/search?q=r20000
20:41 Coke it's also broken?
20:42 Coke (compare with: http://trac.parrot.org/parrot/changeset/20000)
20:42 Coke or, more directly:
20:42 Coke http://trac.parrot.org/parrottest/changeset/20000
20:42 cotto_work http://trac.parrot.org/parrottest/changeset​/0b316e9821516153aee2eca6dab5ea432166511a/
20:42 particle cotto_work: do a poll on the website
20:42 particle coke++
20:42 cotto_work but I see what you mean
20:43 senf_statt_oel joined #parrot
20:43 Coke cotto_work: the WHOLE point of this was to keep the old revision numbers working. no?
20:43 cotto_work Yes
20:43 cotto_work I'll get that working before setting up a poll.
20:44 Coke ah, thought the osuosl guys were doing that. ok.
20:44 cotto_work I think it'll have to be a custom extension.
20:45 Coke so, we only need the source code browser on trac.parrot.org so that we can say "what was r20000 in the old system?", as I understand it. for anything else, folks will have the entire git repository on their desktop. If they want a nice web based version, I vote github.
20:45 Coke (but only because it's treated me well for perl6 & partcl. I'm not married to it.)
20:46 cotto_work Sure.  http://trac.parrot.org/parrottest/changeset/2000 would just redirect to the right revision on github or something
20:47 cotto_work s/test//
20:47 particle i like gitorious on our own parrot vm, rather than an external host
20:47 particle gitorious+cgit
20:48 Coke cgit-- at first look.
20:49 Coke I don't mind selfhosting. doesn't really matter as far as git goes, as long as we have a decent admin interface.
20:49 particle git.openefs.org <== cgit
20:50 Coke yah. looks very ugly to me.
20:50 particle the css isn't perfect. that's my fault, not the product's.
20:50 particle well, it's not shiny, that's for sure
20:55 cotto_work I'll get the plugin working and post a poll.  I can see the wisdom of coke's recommendation.
20:56 cotto_work and happily, there's nothing that'll prevent any alternate hosts of people prefer them
20:59 tewk_ gitolite is really nice, it allows creation of new repos on the fly, and user specific repos too.
21:03 ambs joined #parrot
21:21 lucian joined #parrot
21:39 AndyA joined #parrot
21:48 tommyd joined #parrot
21:53 cotto_work http://www.scribd.com/doc/35240506/Maki​ng-Python-Fast-PyPy-and-Unladen-Swallow - slide 24 sounds familar
22:02 pyrimidine left #parrot
22:02 senf_statt_oel left #parrot
22:12 gbacon joined #parrot
22:39 bacek joined #parrot
22:44 ruoso joined #parrot
23:07 cotto_work dukeleto: ping
23:08 Chandon joined #parrot
23:11 Chandon I'm getting an error when I try to run examples/sdl/tetris. Is there something simple I need to do to get it to work?
23:22 dngor_ joined #parrot
23:22 dalek plparrot: 849e6a8 | (Jonathan "Duke" Leto)++ |  (2 files):
23:22 dalek plparrot: Attempt to support nested pointy blocks
23:22 dalek plparrot: review: http://github.com/leto/plparrot/commit/8​49e6a82a8aec97cfb60bf0640b8fbf2dfdb1d5a
23:22 dalek plparrot: 21df97c | (Jonathan "Duke" Leto)++ | t/sql/plperl6.sql:
23:22 dalek plparrot: Fix the pointy block test
23:22 dalek plparrot: review: http://github.com/leto/plparrot/commit/2​1df97c64e5767af0dea4e990047f411fa21e011
23:22 dalek plparrot: 6a74227 | (Jonathan "Duke" Leto)++ | plparrot.c:
23:22 dalek plparrot: Clean up Nil/Parcel handling a bit
23:22 dalek plparrot: review: http://github.com/leto/plparrot/commit/6​a742270d53dac7984ef50adede0085b5c89844b
23:23 whiteknight joined #parrot
23:34 workbench joined #parrot
23:41 kthakore NotFound: ping
23:58 Psyche^ joined #parrot

| Channels | #parrot index | Today | | Search | Google Search | Plain-Text | summary

Parrot | source cross referenced