Parrot 4.5.0 "Buff-faced Pygmy Parrot" | parrot.org/ | Log: irclog.perlgeek.de/parrot | #parrotsketch meeting Tuesday 19:30 UTC
Set by moderator on 8 July 2012.
00:00 kid51 joined 00:21 particle joined 00:24 particle1 joined
dalek kudo/nom: 95dfc00 | pmichaud++ | src/core/Version.pm:
Revise Version.Str / .gist / .perl .

After thinking about it a bit more, I'm of the opinion that
  $version.Str should produce a string that doesn't include
a leading 'v'. The .gist form does have a leading 'v', however, so C< say $version > still does what one expects.
When putting a version in the middle of a string we'll often want to exclude any leading 'v' (e.g., when generating filenames), it's much easier to add a 'v' when it's needed than it is to eliminate one we don't want. Also, the form without the leading 'v' is the only one that can (so far) reliably have much chance of round-tripping back into Version.new(...), since alphabetics in strings arguments to Version.new are meaningful.
At the moment the spec is somewhat silent on this topic, so I'm hoping this will make more sense. If not, a spec clarification or some official tests will let us know how to improve it. :-)		 00:38
00:46 benabik_ joined
whiteknight msg dukeleto I just sent out an email with you CC'd. Let me know if you have anything to add. I don't want to pile on more than necessary 00:56
aloha OK. I'll deliver the message.
01:15 seiferteric joined
dalek kudo/nom: 499cc68 | pmichaud++ | src/core/Version.pm:
Cleanup Version .parts handling a bit.
 01:42
01:44 benabik_ joined 01:52 preflex joined 03:43 seiferteric left 04:34 fperrad joined 07:11 brrt joined
dalek kudo/nom: c758c1b | (Felix Herrmann)++ | src/core/ (2 files):
add X::Numeric::Real

Signed-off-by: Moritz Lenz <moritz@faui2k3.org>		 07:36
08:14 preflex_ joined 08:16 lucian joined
arnsholt So, if I just invoke pthread_create inside parrot code without any additional protections, fireworks and explosions will ensue, right? 08:37
sorear It should be fine 08:50
(as long as you don't try to access the interpreter from the child thread) 08:51
moritz which means, you only run C code :-)
and no parrot-related C code either
arnsholt Yeah, that's what I thought 08:54
No pthread_create(...) is native("pthread") { * } =)
sorear parrot is designed to operate with one pthread per interpreter 09:11
there was a project to implement a full-featured erlangish message-passing/actors concurrency system; I don't know how far it got
moritz sorear: stalled in a branch, iirc 09:12
the fastest way to get threads into parrot would be to fix the few remaining failures in the threads branch
brrt .. maybe readifying could start a green thread? if necessary that is, i use 'Parrot_api_ready_bytecode' to pre-load libraries 09:19
09:31 Khisanth joined
dalek kudo/nom: c2c02ab | (Gabor Szabo)++ | / (2 files):
change most of the Test.pm functions to return their truth value and test this in the sanity test
 09:34
09:50 Khisanth joined 10:20 Psyche^ joined 10:34 zby_home joined
NotFound ~~ 11:11
brrt \\o NotFound 11:49
I must say I'm a little confused w/regards to the starting of execution 11:55
moritz you speak up, we execute you :-) 12:00
erm, probably not what you wanted to know :-)
NotFound brrt: Yes, Is a little confusing. 12:02
brrt ... oh moritz, if only you'd known 12:21
i have a hand-crafted call stack from parrot_api_load_bytecode to parrot_pcc_invoke_from_sig_object 12:22
and another from Parrot_api_pmc_invoke 12:23
they differ.. markedly
but they both end in Parrot_pcc_invoke_from_sig_object 12:24
12:24 particle joined
brrt ... would it be a good idea to add task creation to interpreter instantiation? 12:34
or, do we want to kill green threads?
dalek kudo/nom: 74e1838 | moritz++ | src/core/Exception.pm:
[Exception] add methods resumable and resume
 12:46
12:58 PacoAir joined 13:32 bluescreen joined 13:40 JimmyZ joined 14:25 JimmyZ joined 14:29 dmalcolm joined
dalek p: 62b9fc9 | (Gerhard R)++ | 3rdparty/dyncall/buildsys/gmake/tool/gcc.gmake:
[dyncall] workaround for build issues on case-insensitive file systems

fixes issue #47		 15:18
15:57 fperrad joined 16:02 jashwanth joined
dalek p/toqast: f014bdc | jnthn++ | src/QAST/Compiler.nqp:
Handle an optimization escape hatch for the way $_, $!, $/ and %_ are handled.
 16:38
kudo/toqast: 8721264 | jnthn++ | src/QPerl6/Actions.pm:
Updates to use new exception handling approach in QAST. Gets 12-try.t passing again.
16:47 preflex_ joined 17:39 rich joined 18:06 mtk joined
nine Good evening, #parrot 18:10
dalek p/toqast: 3dc36d1 | jnthn++ | src/QAST/Compiler.nqp:
Implement falling back to symbol if there are QAST::Var nodes with no scope declared.
 18:15
kudo/toqast: 869ffc9 | jnthn++ | src/QPerl6/Actions.pm:
Toss :lvalue marker that isn't needed.
 18:19
kudo/toqast: e99f5b5 | jnthn++ | t/harness:
Run spectest with qperl6.
kudo/toqast: 3589ee7 | jnthn++ | src/QPerl6/Actions.pm:
Fix self, booleans.
kudo/toqast: 4b2dc33 | jnthn++ | src/QPerl6/Actions.pm:
Fix up Whatever * hanlding a bit.
18:54 tuxit joined 18:55 lucian joined
dalek kudo/toqast: fd5eecc | jnthn++ | src/QPerl6/Actions.pm:
Fix &foo.
 18:56
kudo/toqast: 5768ec4 | jnthn++ | src/QPerl6/Actions.pm:
Fix return.
19:02 whiteknight joined
whiteknight good afternoon, #parrot 19:02
nine Good evening, whiteknight
whiteknight hello nine!
nine: I'm trying to get those threads issues dealt with, but I haven't been able to find a PPC machine to work on 19:03
I'm leaning towards the "we don't support that platform anymore" side of things, considering nobody who can fixes these kinds of issues has access to that platform 19:04
but that's not a decision I can make by myself
nine That would have been pretty much the same with Windows considering how much time went by until we could sort that out... 19:06
whiteknight yeah
I suspect what rurban++ says is correct: It is probably specific to the test 19:07
but without looking at it, I can't say anything conclusively
nine The test was problematic before. Might still have some issues left 19:08
whiteknight msg dukeleto either nine or I really need to get shell access to a PPC machine to handle a test failure on that platform in the threads branch. Do you have any ideas for such a thing?
aloha OK. I'll deliver the message.
whiteknight cross your fingers... 19:09
nine Speaking of tests. I had my final bachelor's exam on Monday and passed so I got more time now :)
whiteknight nine++
Congratulations!
nine thank you very much!
niner.name/Hybrid_Threads_for_the_Parrot_VM.pdf 19:10
This also means that my paper is finished. I even added a couple of diagrams ;)
dalek kudo/toqast: c37bf13 | jnthn++ | src/QPerl6/Actions.pm:
Fix :foo and :!foo colonpairs.
 19:15
kudo/toqast: ad2340a | jnthn++ | src/QPerl6/Actions.pm:
Fix named arg passing.
nine The failing test 37 in nci.t is nci_cb_C1 which is also the first test to use sleep
19:20 rich left
whiteknight so do you think sleep is problematic? 19:24
nine It sleeps for 10ms in a loop with at most 10 iterations waiting for the callback to happen. That's not exactly much time. Looks similar to the problem I fixed in the same file where a sleep was just too little on a loaded system.
benabik whiteknight: I know sleep is problematic... oh, wait, you meant in Parrot, nvm. 19:25
whiteknight benabik: :)
benabik I think the gcc farm has PCC machines...
whiteknight nine: So we can increase the timeout there or we can scrap the test as being fundamentally flawed 19:26
do other sleep-related tests pass? (do we have other sleep-related tests?)
benabik gcc.gnu.org/wiki/CompileFarm
whiteknight benabik: I figured that was the answer. Now all we need is for nine or I to get access to it
dangerous, scarey SSH access
scary
benabik whiteknight: Looks like they're pretty liberal with access. Instructions are on the wikis 19:27
whiteknight I've been hunting through craigslist for an affordable PPC-based macbook
nine t/op/time.t tests sleep
whiteknight so far, the ones I can find are not really affordable
Coke whiteknight: you want, I'll ship you my dead one. ;)
whiteknight Coke: how dead are we talking?
Coke I think that macbook has a dead logic board. 19:28
whiteknight ouch
that's....like the one thing we need, right?
Coke I have one with a dead board, one with a dead drive.
benabik I have an old G4 Powerbook around somewhere... I think it might have just been display problems that made me put it away.
whiteknight Yeah, that's the problem in a nutshell. Either nobody has them, or they are all old and broken by now 19:29
Coke ppc is an EOL product from apple. 19:30
How old is the oldest OS that supports it?
whiteknight right, which is why I question whether we continue to support it in the future
We should set some kind of a deadline, if that deadline has not already passed
benabik 10.6 was the first OS not to support it, released 6/8/09 19:31
Sorry, 8/28
whiteknight that's....a while ago
nine Let's not forget that nci.t also gives problems on Linux/i386
whiteknight Next up from Captain Obvious: The sky is up! The ground is down!
nine: I had forgotten about those. 19:32
I need to fire up a VM
benabik The only other place I can think of that uses PPC is some old Sun boxes.
whiteknight benabik: access to one of those is even less likely
nine any ideas how I could run this test really slowed down? 19:33
benabik whiteknight: I might still have access to a couple at school, but the sysadmin was pretty set on trying to remove student access to them (or remove them completely)
whiteknight sounds like we may need to go dumpster diving
nine: What do you mean?
nine if the problems are timing related, my two months old i7 is not gonna reproduce them... I need it to slow down 19:35
whiteknight Create a VM. 1 core. limited ram. 19:36
Compile GCC in the background
in each of three or four separate shells 19:37
benabik O.o
nine Last time I used like 30 cat /dev/random > /dev/null on my quad core laptop. qemu might give me a head start :) 19:38
moritz just compile rakudo a few times in parallel 19:39
until your machine swaps 19:40
whiteknight The last two times I tried Rakudo spectest, my computer crashed. But that's because I've got various hardware problems and need to get it cleaned 19:44
dalek p/toqast: 200e52f | jnthn++ | src/QAST/Operations.nqp:
Implement a vivify op.
p/toqast: 4f2cb27 | jnthn++ | src/QRegex/P6Regex/Actions.nqp:
Add (not really done yet) QAST variants of various things in QRegex::P6Regex::Actions.
kudo/toqast: 5a7d74b | jnthn++ | src/QPerl6/Actions.pm:
Unbustify smartmatching.
 19:45
kudo/toqast: 815a09e | jnthn++ | src/QPerl6/Actions.pm:
Fix a couple of .type => .returns.
 19:46
kudo/toqast: 197f1ea | jnthn++ | src/QPerl6/ (2 files):
Add perl6_assert_bind_ok => p6bindassert.
kudo/toqast: fe865e9 | jnthn++ | src/QPerl6/Actions.pm:
Update compilation of attribute access.
kudo/toqast: 83c62e8 | jnthn++ | src/QPerl6/Actions.pm:
A couple more small qastifications.
whiteknight wowsers. I love watching big streams of nqp and rakudo commits 19:58
rurban whiteknight: kid51 has such a PPC machine available. ask him for ssh access 19:59
benabik I think it's a G4 laptop, not a server. :-/ 20:00
benabik could be wrong though.
rurban yes, a tiny and slow machine
benabik More to the point "not something that's online regularly"
rurban He had it at the YAPC US in madison. I saw it
Still faster than a qemu software-simulated instance 20:02
whiteknight I put out a call for help on Facebook. See if anybody I know has an old machine to donate to the cause 20:04
rurban I am trying this out now: people.debian.org/~aurel32/qemu/powerpc/ 20:06
whiteknight I think I need to install qemu
rurban sure :) 20:07
Mu HW accelerated kvm images are faster then real servers 20:08
Whow, this is fast. Looks like a workable image. qemu-system-ppc -hda debian_lenny_powerpc_standard.qcow2 20:11
I'll try out a sparc64 with solaris also then. 20:12
whiteknight heh, nice
Coke nine: use sleep opcode? 20:28
rurban Setting up a new bridge between the qemu machines is not so easy.
benabik Hm. Winxed heredocs can't have anything after their terminator. 20:30
Is there an easy way to get a Packfile PMC from a PackfileView? 21:11
Hm. I could get/set_pointer, but I don't think those are accessible from inside the VM... 21:15
I could serialize/unpack but that's kinda wasteful. 21:17
msg whiteknight Is there an easy way to get a Packfile PMC from a PackfileView? "Best" way I see now is to call PackfileView.serialize and Packfile.unpack. But that, IIUC, results in two copies of the Packfile being loaded. 21:18
aloha OK. I'll deliver the message.
nine Good night, #parrot 21:19
22:35 bluescreen_ joined 23:32 alvis joined 23:33 rurban left

Powered by Cro and the Raku® Programming Language.

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

Thank you!