Parrot 1.0 Released | parrot.org | 367 RTs left!
Set by moderator on 12 April 2009.
allison mikehh: aye, but that could be unrelated. Do the tests consistently pass before r37128 and fail after? 00:00
mikehh I am pretty sure it is some kind of config problem
allison mikehh: and if you revert r37128 now, do they pass again?
mikehh: yes, probably 00:01
purl Really? Probably? Are you Certain it's not certain? Are you sure it's unsure? I think you need to look harder. Because I'm an asshole.
mikehh haven't tried that - but the thing that bothers me is that the pass on Ubuntu Intrepid Amd64 00:02
it is using the same compiler and the same version of perl
allison <shrug> they are different distributions, so it doesn't surprise me that one would shake loose bugs the other doesn't 00:03
mikehh same computer
allison but different software
mikehh not that different - I am mostly using the same programs 00:04
what I think I need to do is install another copy of Kubuntu Intrepid on one of my free partitions just to test this without all the other baggage 00:06
allison I should probably set up a kubuntu chroot 00:07
I have chroots for multiple versions of debian and ubuntu
00:09 AndyA joined
mikehh unfortunately this computer is a dual core Pentium D without the vm extras and so I can't set up a proper V< on it 00:10
s /V</VM/
I have to reboot every time I switch 00:11
allison wayland76: mmm... actually, I'm using -X instead of -Y, as it sounds more like "extension" 00:13
wayland76: (combing my way through what letters are still available) 00:14
mikehh: a chroot should work without a proper VM 00:17
mikehh Actually come to think of it I have been getting some weird results on Kubuntu Intrepid i386 on the jit core - I wonder if they occur with Ubuntu i386? 00:18
pmichaud fwiw, I'm currently installing kubuntu jaunty amd64 on my notebook, I can see how things run there.
I had a lot of issues with Kubuntu Intrepid (both amd64 and i386), which is why I've stuck with Hardy. Jaunty seems to be a lot better.
(where "currently installing" === right this very moment :-) 00:19
00:20 tetragon joined
mikehh I have had major problems with their Perl 5.10 - they set up directories all over the place 00:20
I am was going to wait until next week to install Jaunty - one of the things that worries me is that is using the new version 00:25
of Grub and I hope it fits with what I have at the moment
00:26 wayland76 joined
mikehh alison: I haven't tried - maybe I should 00:27
wayland76 allison: Thanks again for the heads up :)
mikehh I was supposed to be getting a new system, but I am still waiting for some clients to come through 00:29
00:32 TonyC joined
mikehh One of my clients is setting up shop with some new Mac Pro's, but I don't think I can convince them to provide me with one, maybe :-} 00:33
a nice one with twin Xeon's and 16 or 32GB memory 00:34
allison mikehh: even just a few hours testing on one occasionally would be nice :)
mikehh: or a remote login 00:35
mikehh that I will have
bacek *incoming* 00:39
bacek runs
mikehh They will certainly be using Perl and Bio-Perl type apps on them
dalek rrot: r38088 | bacek++ | branches/packfile_revamp/t/native_pbc/annotations.pbc:
Sample annotations.pbc for testing
 00:41
rrot: r38089 | bacek++ | branches/packfile_revamp (2 files):
Handle Annotations segment during Directory unpack
rrot: r38090 | bacek++ | branches/packfile_revamp/src/pmc/packfileannotationkeys.pmc:
Add VTABLE elements to PackfileAnnotationKeys
rrot: r38091 | bacek++ | branches/packfile_revamp (2 files):
Properly unpack PackfileAnnotationKeys
rrot: r38092 | bacek++ | branches/packfile_revamp (2 files):
Implement unpacking of Annotation entities
rrot: r38093 | bacek++ | branches/packfile_revamp/docs/pdds/pdd13_bytecode.pod:
[spec] Add VTABLE elements to PackfileAnnotationKeys
bacek purl: msg jonathan I've implemented Annotations unpacking. One idea - current Annotations PMC are very low level. We can get rid of "Keys" and implement handling of it behind the scene. 00:50
purl Message for jonathan stored.
bacek purl: msg jonathan If we extend Annotation to have "STRING *key" and "PMC *value" then it will simplify usage of Annotations. And PackfileAnnotationKeys will be created during "pack" 00:51
purl Message for jonathan stored.
bacek afk & 00:52
ah!
purl: msg jonathan And there is no PMC for annotation Groups. And I don't quite understand how they should be implemented. 00:53
purl Message for jonathan stored.
allison cotto++ on removing the last traces of the UnionVal 01:06
dalek rrot: r38094 | allison++ | trunk/compilers/imcc/main.c:
[core] Add a -X option to the parrot executable for specifying an

options for -I (--include), -L (--library), and -X (--dynext).		 01:24
01:25 kid51 joined
kid51 wayland76: Re TT412: Would you be able to run a diff on the lib/Parrot/Config/Generated.pm that you get when you configure with and without --configure_trace? 01:28
wayland76 kid51: Are you sure you don't mean mikehh? 01:29
kid51 You are correct. My error in reading the scrollback. 01:30
Take 2:
wayland76 ok, fine :). I've submitted another patch for the install_files branch :)
kid51 mikehh: Re TT412: Would you be able to run a diff on the lib/Parrot/Config/Generated.pm that you get when you configure with and without --configure_trace?
wayland76: Have looked at the patch. I'm skeptical of your idea of merging the Parrot::Manifest and Parrot::Install objects. 01:36
wayland76 Why is that?
kid51 One underlies tools/dev/mk_manifest_and_skip.pl. The other underlies (or will, assuming our approach is accepted into trunk) tools/dev/install_files.pl and install_dev_files.pl. 01:37
wayland76 I'm aware of that :)
kid51 Those are different functionalities used in different parts of the configure-build-install cycle.
wayland76 Maybe so, but they both represent the same thing -- a (group of) MANIFEST file(s)
kid51 One if focused on getting the MANIFEST right. The other is focused on getting the installation right. 01:38
s/if/is/ 01:39
wayland76 Agreed, but I think if done right, they'll work fine together.
kid51 One is focused on creating the MANIFEST; the other on using it.
wayland76 At the moment, I'm playing with a class which I'm calling ManifestFile which represents one file in the MANIFEST 01:40
It's based on the $filehash in lines_to_files
kid51 I think for Parrot::Install, the sort of object we can use is one that wraps all the code in the 2 scripts that precedes the first sub we call.
mikehh kid51: will do - but I am too tired now - got to get some sleep - afk& 01:41
kid51 mikehh: Thanks. 01:42
wayland76: I think we'll have less resistance getting this branch merged into trunk if we keep it focused on accomplishing the goal you set out in the name of the TT: placing all the code in the 2 scripts in one place. 01:45
wayland76 Hmm. That may well be a good point 01:46
I'll try to do it in such a way as to facilitate future merging with Parrot::Manifest, though
And it doesn't invalidate my Parrot::ManifestFile idea, either :) 01:47
kid51 People will need to be able to put the trunk versions of the two install scripts side-by-side with your/our code and say, "Ah, I see, this part of the script was moved to this place in the module."
I was thinking that a Parrot::Install object's data structure would probably encompass a lot of the code between, say, lines 125 and 226 of what is now in branch for install_files.pl. 01:49
i.e., the code which precedes lines_to_file().
If that stuff were inside an object, then lines_to_file() would become a method on that object -- and we wouldn't have that list of 5 arguments to that function. 01:50
The more code in the scripts that we can wrap neatly up into subroutines/methods, the better position we'll be in to write tests that safely simulate the entire install process. 01:51
And if we can safely simulate the entire install process, then we've got regression tests to use to test any future modifications of the install process. 01:52
... which, IMO, is the real payoff from this sort of refactoring/testing.
01:53 eternaleye joined
kid51 The only code I would *not* try to wrap up inside an object is that which processes command-line options. 01:54
See tools/build/ops2pm.pl as an example of a program which relies on one module for its command-line processing and another for its object and methods. 01:55
And which encapsulates a tremendous amount of code in testable methods. 01:56
Similarly, tools/dev/mk_manifest_and_skip.pl: It's almost nothing but Parrot::Manifest method calls -- which means that with our tests we can very closely simulate the actual MANIFEST-creation process. 01:57
tools/build/ops2c.pl also relies on one class for its command-line processing and another for its operational calls. 01:59
wayland76 I agree with your goal for the scripts, I'm just working through the details of creating the objects 02:02
I also think the documentation patch I submitted could be applied regardless of whether we merge the objects or not.
kid51 I'll take a look at that. It's curious, from a Parrot-historical point of view, that that description of the MANIFEST format was located in the install script rather than the script that actually build the MANIFEST. 02:04
I'll have to see whether that just happened or whether someone made a conscious decision about that at some point long ago (in Parrot years). 02:05
wayland76 Based on looking at install_files and install_dev_files, I'd say that both of these "just happened" 02:09
Incidentally, are you wanting to move the metatransform stuff into the object itself? 02:16
Because the reason it's not there at the moment is because it's different between the two files
While you're investigating the history of the files, it might be useful to have someone who knows the history look at the two files in this area, and see whether they're different because they need to be, or because it "just happened" 02:17
kid51 I will. But it's not surprising to me that there are two different scripts. I know that, e.g., in Debian, there is frequently one set of packages that you install when you simply want to *use* a program -- but an additional, '-dev', package that you need if you want to be a developer of that program. 02:20
So we'll have to investigate further.
One of the benefits of this sort of process is that it increases the number of contributors to the project who are forced to become intimately familiar with particular swaths of code. 02:21
wayland76 Fedora has the same packaging split, except they call it "devel" 02:22
(ie. -devel)
With the Parrot RPM, it's not possible to build Rakudo on it without installing a separate copy of Parrot from source 02:24
While that's an automated process in Rakudo, it's still bad
That's one thing I'm trying to fix :)
kid51 Yes, I see that Allison just applied something of yours in TT 442.
kid51 has yet to actually install a copy of Parrot! 02:25
... because until very recently we knew the whole process was broken.
02:35 janus joined
kid51 grep the repository for DEVELOPING and you'll see that we have a lot of configuration code that works one way with DEVELOPING and the other way without. 02:37
kid51 must sleep 02:43
purl $kid51->sleep(8 * 3600);
cotto allison, ping 03:31
allison cotto: pont 03:33
g
cotto Was the intent of RT #48014 that all uses of the unionval be deprecated, or only in PMCs? 03:34
allison cotto: ultimately, that the unionval be removed from the PMC struct entirely 03:35
cotto: (though maybe not that ticket, I'll look at it)
cotto It'll be tricky to get it out of of PObjs. 03:36
allison cotto: what's it being used for in PObjs? 03:37
cotto I tried switching the order of the UnionVal and the flags in all PObjs (by editing pobj.h) and Parrot exploded when compiling config_lib
beats me
allison cotto: ah, that may have to do with structure alignment 03:38
cotto: but you might have just missed one
cotto I'd expect gcc to deal correctly with struct alignment. 03:39
allison (or missed someplace that was poking directly into the struct)
do you still have the error message from the explosion?
cotto: we still have trouble with struct alignment on various platforms 03:40
cotto: C (gcc or otherwise) can be very finicky about such low-level details. My guess is that it's making some optimizations. 03:41
cotto eew.
allison cotto: you can also try padding the struct (insert a random pointer variable or something similar) to see if that makes any difference 03:42
cotto But we don't want to eliminate all uses of the UnionVal from Parrot, right?
allison and, also try just removing the cache member entirely, instead of reordering it
cotto: well, we want to remove it from PObjs, PMCs, and STRINGs 03:43
that doesn't mean there's an absolute prohibition against using C unionvals, just that the Parrot defined UnionVal goes away
cotto It doesn't even come close to compiling if I remove it. Looks like there's still work to be done.
allison cotto: STRINGs will have to have the buffer elements added directly to the string structure 03:44
cotto Ok. C unions are ok (occasionally), but the UnionVal needs to go away eventually.
allison cotto: that is parrot_string_t
cotto right
allison cotto: aye 03:45
03:45 Andy joined
allison cotto: but it's a continuation of the issue, not part of that original ticket 03:46
cotto I'm glad that one ticket could be closed. 03:48
04:14 iblechbot joined 04:21 Tene joined
cotto irclog? 04:53
purl i think irclog is irclog.perlgeek.de/parrot/today or see also: infrared clogs
05:36 dukeleto joined 07:14 particle joined 07:50 masak joined 08:08 barney joined 08:12 samlh joined 08:56 Counterspell joined 09:02 szabgab joined 09:27 iblechbot joined 10:22 Woody4286 joined 10:25 Ademan joined 10:33 uniejo joined 11:03 uniejo joined 11:40 uniejo joined 12:23 Woody4286 joined 13:14 gryphon joined 13:23 masak joined 13:31 gryphon_ joined 13:40 bsdz joined 13:41 barney joined 13:51 Tene_ joined 14:02 wayland76 joined
dalek rrot: r38095 | coke++ | trunk (3 files):
[cage] fix some codingstd test failures. (TT #543 still an issue)
 14:22
14:31 amoc joined
Coke so getting rid of perl for our config may be more effort than it's worth. 14:37
(esp since autoconf itself requires perl)
still might be nice to rely on something more standard for config processing, though. (it just won't remove perl from our dep list, just push it further down the stack) 14:38
14:39 PacoLinux joined, davidfetter joined
barney I still think that cmake could be a good choice, but I don't really know it 14:40
Coke ... then on what do you base your recommendation? 14:45
bsdz hi, anyone know a little about pir roles? 14:47
Coke not really, but can probably help point you in the right direction.
bsdz hope you can. been trawling everything. i need to subclass a parrot type and assign a existing parrot role to my new class. any pointers? 14:48
Coke barney: a quick skim makes me think that cmake wouldn't replace Configure.pl
by type, do you mean "core pmc" ?
bsdz yes
14:49 ingy joined
Coke ok, and you can do that part? 14:49
bsdz want to subclass an pmcarray type and add_role 'array' to it
yes can subclass easily
barney Coke: It works on the major platforms, not UNIX-centric. Projects like KDE probably neen anything that Parrot needs
bsdz can't work out how to assign a role from an existing role. also don't know if roles are inherited
ps: notepad++ uses cmake i think 14:50
Coke barney: but what does it /do/ ?
anything other than what we already do with make?
bsdz: where do you see the 'array' role defined? 14:52
bsdz: I am unsure that there are any core Roles. 14:53
You'd have to create the role before you passed it around.
bsdz it's in a .pmc file. fixedpmcarray provides array. also in dumper i noticed it checking for does 'array'
barney Coke: It's a popular build system, that's about all I know 14:54
Coke those aren't roles.
at least, those aren't Role PMCs.
bsdz ah. so "does" is not for roles?
Coke that just sets a text string somewhere to indicate that something 'does' array, SFAIK, it's used only by 'does' at the PIR level. 14:55
no. Does is pretty much for decoration, SFAIK.
there's no real code behind it.
bsdz oh right. you know anyway i can set my derived array class so it "does" array too>
?
Coke sure. 14:56
oh. Hurm. I can do that from a PMC subclass...
bsdz i'm subclassing in pir. that should be okay right?
Coke yes. I think does will 'inherit' there. 14:57
bsdz hmm. doesn't appear to. actually trying to use dumper on my inherited type and it "does" not array :( 14:58
nopaste "Coke" at 65.91.151.195 pasted "does on a subclass" (8 lines) at nopaste.snit.ch/16258
bsdz hold on. might be something in dumper then.
Coke or I could be checking does on the class, not the instance. =-) 14:59
nopaste "Coke" at 65.91.151.195 pasted "does on a subclass (both class and instance)" (14 lines) at nopaste.snit.ch/16259 15:01
Coke looks like both the class and the instance both report does == true.
bsdz hmm 15:02
Coke so dumper might be checking something else before it checks the array-ness.
bsdz thanks Coke. I'll take a closer look
15:03 cognominal joined
Coke bsdz: if it's an object, it checks that first. 15:03
subclasses in pir are objects, so it calls pmcDefault for dumping instead of genericArray 15:04
bsdz yes, looks like it gets caught there
Coke you can create your own __dump method to override the default behavior of Dumper 15:05
lame, but involves changing no core code.
bsdz yes was looking at that
in fact, first thing i started doing but found out there was no *array.__dump method i could steal :( 15:06
though actually i could jump into dumper's namespace and borrow genericarray 15:07
Coke yup.
15:08 rafl joined
bsdz thanks for your help Coke. it was dead simple in a __dump method. then i wonder why i waste so much time! :) 15:10
Coke I would tend to blame parrot, not you. =-) 15:14
15:21 jan joined 15:45 riffraff joined 15:58 lucs joined 16:02 darbelo joined
darbelo � 16:03
16:05 rdice joined 16:09 pmichaud joined 16:10 pmichaud joined 16:24 amoc joined
dalek kudo: 2c13d6c | pmichaud++ | docs/spectest-progress.csv:
spectest-progress.csv update: 359 files, 10298 passing, 5 failing

   S32-hash/reverse.t aborted 5 test(s)		 16:49
shorten dalek's url is at xrl.us/beotpg
17:35 particle1 joined 17:48 Tene joined 18:05 wayland76 joined 18:21 bsdz joined 18:28 luca joined
dalek a: 7d5e669 | (Francois Perrad)++ | src/pmc/luanil.pmc:
fix PMC build, PMC_union was removed from Parrot
 18:34
shorten dalek's url is at xrl.us/beot5n
18:40 dduncan joined, dduncan left 18:47 japhb joined 19:13 luca joined
dalek rrot: r38096 | coke++ | trunk:
[cage] ignore files generated by 'make codetest'
 19:20
19:24 darbelo left 19:43 gryphon joined, Andy joined 19:51 luca left 19:58 cosimo joined
Coke svn st 20:14
ww
particle- alias ww issri 20:15
er, typing fail 20:16
cotto seems to be a theme around here 20:22
particle- sure, break the trend.
20:39 acajou joined 20:42 iblechbot joined 20:51 cosimo left, raiph joined 21:10 japhb joined 21:15 Casan joined
raiph hi all; trying to build rakudo; on feather; ulimit -v is 524288; during pbc_to_exe perl6.pbc I get "virtual memory exhausted"; if I interpret things at face value then I need to ask juerd if he'll up vmem; what would y'all advise I ask for -- or is this in fact really about something else, not vmem? tia 21:19
21:30 Whiteknight joined 21:32 contingencyplan joined 21:35 acajou left
pmichaud raiph: how large is perl6.pbc ? 21:40
Andy hey, pmichaud, that oslo hackathon thing would be a godo post on rakudo.org
pmichaud posting.
purl posting is probably not nice
Coke no, posting is <reply>
purl okay, Coke.
pmichaud can we get the blog posts to appear on rakudo.org's front page somehow? 21:41
otherwise people don't really find them.
and the rss link is never up-to-date :-| 21:42
Andy Yeah, I oughta poke at that
but won't be before Friday
I'm driving to Omaha tomorrow. 21:43
pmichaud is there anyone else who can poke at it? Or is it an Andy-only sort of thing?
Andy Anyone else could, if they knew it.
lemme try something
raiph pmichaud: 3159712
pmichaud raiph: hmmm... seems odd that a 3.1MB file would expand to eat up over 524MB of virtual memory. 21:44
I suspect a bug or memory leak somewhere.
Andy ok now look
pmichaud just discovered that his flight to oslo has power plugs in every seat :-) 21:45
well, I didn't want the front page to be _only_ blog posts -- I wanted intro information and then blog posts.
similar to what is done on parrot.org
Andy well, I don't know what is done on parrot.org
or how Allison set that up. 21:46
Coke are you using drupal on rakudo.org?
Andy yes
allison Andy: parrot.org doesn't use blog posts, we just use Stories and Pages and flag which ones we want to appear on the front page
Andy We don't use blog posts either 21:47
we just have Stories and Pages
allison Andy: we also archive older posts, to keep the front page uncluttered
(They appear in "News".)
Andy I don't know what you are doing to do that.
pmichaud allison: how do you get the box of information that is at the top of parrot.org and remains there?
Andy If someone can tell me what you want, I will make it happen.
I can't go figure it out right now.
Coke looks like "mission" under "site information" in admin. 21:48
pmichaud I'd be very happy to fix this up but I don't think I have admin privileges to either site :-|
Andy pmichaud: Are you saying you want admin privs? 21:49
allison Coke: aye, mission
(I was looking it up too)
pmichaud Andy: yes, that would be nice :-)
I would bug you less :-)
allison Andy: some themes will show the Mission, others don't
pmichaud: and some themes show the mission in an odd location
pmichaud allison: okay, very helpful. 21:50
allison but you can edit the templates if you don't like the default location
bacek good morning
Andy pmichaud: ping me in AIM 21:51
purl I can't find me in the DNS.
pmichaud aim?
purl aim is their free version, that only allows instant messages, the AOL equivelent of a MSG. or the prog that WON'T UNINSTALL or kickass or AOL Instant Messenger or the deeply moronic and utterly misnamed Accuracy In Media organization (www.aim.org) or exploitable in several ways or adding free.aol.com/ to IE trusted sites zone
Andy don't you have an AIM account? 21:54
I'm sure I've talkedto you on there.
pmichaud I did a long long time ago. I don't even remember my screen name.
(getting a new one now)
Andy ok, i have to deal with this some othertime
I'm sorry
21:55 skv_ joined 21:56 skv__ joined
raiph pmichaud: thanks. sounds like it might make sense for me to wait a few days, git pull, and try make again. right? 21:58
pmichaud I'll try on my local machine with a ulimit in place and see if I get the same results. 21:59
raiph pmichaud: sweet. thanks.
23:07 ruoso joined
GeJ Good morning all 23:13
cotto Hi GeJ 23:24
raiph evenin' GeJ 23:27
23:38 tetragon joined 23:58 bacek_ 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!