Parrot 2.8.0 released | parrot.org Log: irclog.perlgeek.de/parrot/today | Nopaste: nopaste.snit.ch:8001 | smoke GC-related branches and attack GC tickets
Set by moderator on 1 October 2010.
00:16 davidfetter left 00:30 whiteknight left 00:33 kid51 joined 00:41 Util joined
GeJ Good evening kid51. 00:46
kid51 GeJ Hi! 01:06
01:16 kid51 is now known as kid51_at_dinner
plobsing kid51_at_dinner: if you get the chance, I'd love to know whether the patch I posted on TT #1813 works or not 01:35
01:47 ruoso left 02:00 ruoso joined 02:07 kid51_at_dinner is now known as kid51
kid51 plobsing: I am testing now. 02:07
dukeleto 'ello
kid51 dukeleto: Hi! 02:08
dukeleto: Weather in NYC is 54F and raining -- getting me in the mood for Portland 02:09
plobsing: While I'm waiting to rebuild, let me say that for each of the 3 config steps you've recently added, I *strongly* wish you had opened TTs for them and worked in branch(es) rather than committing directly to trunk. 02:11
nopaste "kid51" at 192.168.1.3 pasted "failures persist on Darwin/PPC re inf_nan issues" (476 lines) at nopaste.snit.ch/24016 02:12
plobsing still with the inf != inf problems 02:18
hmmm
dukeleto kid51: i wrote those tests. They probably should be case-insensitive 02:21
plobsing dukeleto: kid51 is correct here. we have a special case (and lot of support code) in parrot to make inf's and nan's more or less system independant 02:22
kid51 dukeleto: Well, they were passing on this box until the past few days.
plobsing what is happening is that somehow the inf's are slipping through
dukeleto plobsing: ok, good to know my tests aren't broken :) 02:23
kid51 dukeleto: They pass for me on Linux/i386, but the things since the addition of config/auto/infnan.pm have messed up the tests on Darwin/PPC. 02:25
dukeleto kid51: i understand now.
i hadn't seen that config step fly by 02:26
plobsing yes, apparently darwin's math.h defines an INFINITY, but it doesn't play nice
my assumption: if you're system defines INFINITY and NAN, they will work
s/you're/your/
kid51 Here's the Smolder: smolder.parrot.org/app/projects/rep...etails/446 02:27
plobsing the rationale (as described in the commit message and referenced ticket), is that our method of generating these values can generate SIGFPE depending on the processor mode 02:28
and does on a couple of architectures by default
02:30 hercynium left, particle1 joined
plobsing kid51: PARROT_HAS_INF_NAN is defined, correct? 02:30
kid51 plobsing: How would I determine that, precisely? 02:32
plobsing it would be defined in include/parrot/has_header.h
'#define PARROT_HAS_INF_NAN 1' 02:33
02:33 particle left
kid51 grep -n HAS_INF_NAN include/parrot/has_header.h 02:33
128:#define PARROT_HAS_INF_NAN 1
02:35 janus left
nopaste "kid51" at 192.168.1.3 pasted "For reference: ppc/math.h" (690 lines) at nopaste.snit.ch/24019 02:35
02:35 janus joined
nopaste "kid51" at 192.168.1.3 pasted "For reference: i386/math.h" (541 lines) at nopaste.snit.ch/24020 02:35
plobsing their definitions are exactly identical 02:38
can you run gdb on a simple program that invokes a sprintf operation on an Inf? (eg: $N0 = 'Inf'; say $N0) 02:39
break at src/spf_render.c:828
actually 833 would be better 02:40
then 'p thefloat'
kid51 plobsing Can't do that now because sleep is calling me 02:42
02:43 kid51 left
plobsing ENOPPC testers 02:47
02:50 ash_ left
dukeleto plobsing: i actually have a darwin ppc box under my desk 02:54
dukeleto turns it on
plobsing oh noes. I'm secretly surrounded!
02:57 mikehh left
dukeleto plobsing: it is going to take a little while for me to get the box into a usuable state 03:09
plobsing: if you create a TT with code that you want me to run, i can get around to it after dinner, probably
dalek rrot: r49448 | cotto++ | trunk/src/runcore/profiling.c:
[profiling] move some stat collection in front of op execution
 03:10
03:11 mikehh joined
sorear how do we feel about using emulators? parrot testing is fast enough it could work 03:14
plobsing nothing like the real thing, but an emulator could help alleviate our shortage of testers for certain platforms 03:15
if I ever manage to get my virtual machines managed acceptably, I would consider a Darwin/PPC emulator VM. 03:17
dukeleto: if you beat me to it (likely, considering the weakness of my sysadmin-fu); it would be helpful if you addressed the first avenue of attack in my last post on TT #1813 03:32
cotto Can someone familiar with the gc tell me what "src/gc/gc_ms2.c:726: failed assertion '_list == _item->owner' 03:35
" might mean?
seen bacek 03:40
aloha bacek was last seen in msg 9 hours 39 mins ago <private message>.
cotto seen bacek_at_work
aloha bacek_at_work was last seen in #parrot 4 hours 14 mins ago saying "GeJ, almost.".
cotto clock 03:41
clock?
aloha, clock
aloha, clock?
aloha cotto: cotto: LAX: Mon, 20:41 PDT / CHI: Mon, 22:41 CDT / NYC: Mon, 23:41 EDT / UTC: Tue, 03:41 UTC / LON: Tue, 04:41 BST / BER: Tue, 05:41 CEST / TOK: Tue, 12:41 JST / SYD: Tue, 14:41 EST
cotto: Sorry, I don't know.
cotto aloha, clock
aloha, clock?
aloha cotto: cotto: LAX: Mon, 20:41 PDT / CHI: Mon, 22:41 CDT / NYC: Mon, 23:41 EDT / UTC: Tue, 03:41 UTC / LON: Tue, 04:41 BST / BER: Tue, 05:41 CEST / TOK: Tue, 12:41 JST / SYD: Tue, 14:41 EST
cotto: I give up.
plobsing cotto: that assertion occurs at include/parrot/list.h:78
cotto src/gc/gc_ms2.c:726: failed assertion '_list == _item->owner' 03:42
plobsing presumably the item being removed from the linked list either wasn't added to the list, or was added to another list in the interrim
cotto: yes, it's in a macro
cotto oic
plobsing (from gdb) look at item->owner and try to guess whether it is valid 03:47
my hunch: something is overwritting that memory by mistake somewhere
dukeleto sorear: i think we are pro emulators. 03:54
cotto Sure. Why not. 04:02
04:04 davidfetter joined 04:05 theory left 05:13 davidfetter left
cotto Hmmm. libefence doesn't like 0-byte allocations. 05:17
plobsing you can set an env var to make it ignore those 05:18
EF_ALLOW_MALLOC_0 05:19
cotto this should take care of my pesky "having free memory" problem 05:20
sorear libefence? people still use that? 05:22
I thought everyone used memcheck now
plobsing IIUC, memcheck ties into valgrind and requires emulation (slow). libefence allows applications to run native while still catching some bugs. 05:28
cotto no luck with libefence 05:30
memcheck is running, but it could be a while
05:49 AzureSto_ left 06:08 AzureStone joined 06:14 plobsing left 06:21 bacek left 06:37 Drossel joined 06:39 Kulag left 06:57 betterworld left 07:35 fperrad joined
GeJ matin. 07:38
08:13 he left
cxreg libefence was replaced by duma, more or less 08:27
cotto had never heard of duma 08:31
08:56 allison left 09:30 he joined 09:43 contingencyplan left 09:50 bacek joined 10:23 ruoso left 12:07 plobsing joined 12:08 M_o_C joined 12:10 whiteknight joined
whiteknight good morning, #parrot 12:11
12:47 ruoso joined 12:57 bkuhn joined 13:00 mikehh left 13:20 Patterner left 13:30 Psyche^ joined, Psyche^ is now known as Patterner 13:37 PerlPilo1 left, PerlJam left 13:38 PerlPilot left 13:39 PerlJam joined 14:05 bkuhn left 14:15 M_o_C_ joined 14:16 M_o_C left 14:32 ash_ joined 14:34 he left 14:45 davidfetter joined, davidfetter left 14:51 M_o_C__ joined, M_o_C__ left 14:53 M_o_C_ left, M_o_C joined 15:23 M_o_C_ joined 15:25 M_o_C left 15:33 M_o_C_ left 15:36 M_o_C joined 15:49 ruoso left 15:54 fperrad left 15:56 fperrad joined 16:04 jsut joined 16:08 ruoso joined 16:09 jsut_ left 16:14 M_o_C left 16:24 davidfetter joined 16:35 pyrimidine joined
dukeleto TAP::Harness::Archive irks me 16:40
whiteknight why?
dukeleto it creates a whole bunch of dependencies for something very simple. I think we adopt the simple way Rakudo submits smoke reports, with curl (or we could use LWP::Simple) 16:41
whiteknight: tests for the dependencies of it fail on my darwin ppc box
whiteknight: so i have never been able to submit a smoke report from this box
whiteknight: it could be something peculiar to this box, but TAP::Harness::Archive is a huge dependency chain to do a POST request. It is a bit ridiculous. 16:42
whiteknight: and it requires any smoke tester to be savvy enough to install it, which makes the people who are willing to smoke parrot a much smaller group, i would imagine
specifically, Compress::Zlib does not like my darwin ppc box, and most of it's test suite fails 16:43
16:43 M_o_C joined
whiteknight haha, I wasn't expecting such a laundry list of complaints! 16:44
but I agree with you, Compress::Zlib routinely fails tests when I install it. I always have to use force 16:45
here's the million dollar question though: If you don't like TAP::Harness::Archive, what do we replace it with? 16:46
dukeleto whiteknight: curl or LWP::Simple 16:47
whiteknight: rakudo submits smoke reports with curl from their Makefile 16:48
whiteknight can we rely on curl, even on windows platforms?
dukeleto whiteknight: that is why I suggest LWP
whiteknight: which has been in core perl for a long time
whiteknight okay. I'm not a huge fan of keeping our build/test infrastructure written in Perl, but you can't argue with the portability
Coke (smoke tester savvy) that's what Task::Parrot is for.
"cpanm Task::Parrot" is a pretty low hurdle. 16:49
(you know, assuming Task::Parrot actually contained all the modules we cared about.) 16:50
whiteknight that's still an extra thing to install, and doesn't get around the problems with Compress::Zlib failing tests 16:52
16:53 mikehh joined
dukeleto it seems that LWP is not core like I thought 16:55
Coke: adding more deps doesn't help
whiteknight LWP isn't core? 16:56
Coke dukeleto: no, but making those deps easy to install does.
so then we can say "just install <foo>" instead of "foo, bar, baz," 16:57
dukeleto Coke: the deps fail tests => more fail
whiteknight: corelist says it isn't
Coke dukeleto: I cannot parse that send. 16:58
dukeleto Coke: the dependencies for TAP::Harness::Archive fail tests. Therefore, no matter how easy we make it to install TAP::Harness::Archive, it doesn't fix this problem. 16:59
Coke Ah. 17:00
must have missed that in the scroll. 17:01
rakudo is using TAP::Harness::Archive, btw. 17:03
unless it changed very recently. 17:04
whiteknight ha. I'm reading over the list of core modules now and I find it a little bit...inconsistent 17:07
some of these modules hardly strike me as having central importance
17:09 theory joined
Coke is this the corelist for perl5? 17:11
dukeleto Coke: grep curl in rakudo's Makefile 17:12
i just force installed TAP::Harness::Archive and now Parrot::Harness::Smoke fails because LWP::UserAgent doesn't exist 17:13
sigh.
we actually have a pure-PIR solution to submit smoke reports
examples/io/post.pir <-- we should just use that to submit to smolder => NO DEPS! 17:15
17:16 ruoso left
atrodo dukeleto> Unless parrot doesn't build 17:21
whiteknight or Parrot fails tests in areas necessary to post the reports 17:23
I like the idea of cutting TAP::Harness::Archive out of our test system
dukeleto atrodo, whiteknight: true. checking for curl/wget or LWP may be a more robust solution 17:26
i just submitted a passing smoke report on darwin ppc
msg plobsing passing smoke test on darwin ppc smolder.parrot.org/app/projects/rep...etails/454 17:27
aloha OK. I'll deliver the message.
dukeleto it seems that Rakudo just uses curl without even checking if it installed 17:29
whiteknight tsk, tsk
17:31 Tene joined 17:32 ash_ left
Coke dukeleto: grep "::Archive" in rakudo's t/harness. 17:36
TAP::Harness::Archive generates the .tar.gz - curl is used to upload it. 17:38
17:41 contingencyplan joined
dukeleto Coke: i see that now 17:48
dalek rrot: r49449 | dukeleto++ | trunk/t/tools/create_language.t:
[t] Add a TODO test for create_language.pl
 18:02
18:03 tadzik joined
whiteknight that smolder graphs bug is weird 18:34
it's also weird to me that 0x0D 0x0A is the start sequence for a PNG image 18:35
especially when parts of the html header are delimited by whitespace like that
I wonder what server software smolder.parrot.org is running? 18:37
18:48 whiteknight left 18:54 fperrad left 18:58 ash_ joined
dalek rrot: r49450 | cotto++ | trunk/src/runcore/profiling.c:
[profiling] move stat collection back, after op execution

I thought it would.		 19:04
rrot: r49451 | cotto++ | trunk/src/runcore/profiling.c:
[profiling] names of functions that call runcore->output_fn now start with record_
cotto dukeleto, how big of a project is it to get github commit messages to parrot-commits with a full diff? 19:18
I can feel the profiling runcore getting less confusing by the day. This is great. 19:21
19:28 chromatic joined
Coke cotto: ... didn't you write that code? 19:29
cotto yes
chromatic msg masak Lambdas are branches at the processor level, but the syntax level is important for conceptual understanding, abstraction, encapsulation, and correctness. 19:30
aloha OK. I'll deliver the message.
cotto That doesn't mean it can't be confusing 6 months later.
cotto still has many lessons to learn about managing complexity. 19:31
chromatic Step one: don't. 19:33
19:34 M_o_C left
atrodo Don't manage the complexity, or don't have the complexity? 19:43
19:43 theory left 19:45 theory joined 19:47 cotto left, cotto joined
chromatic Manage simplicity instead! 19:48
GeJ Bonjour everyone.
cotto wonders how to simplify timing calculations for nested runloops. 19:49
ash_ like clock ticks? 19:51
cotto whatever's the Parrot_hires_get_time uses 19:53
s/'s the//
It shoud pick the best available method. 19:55
cotto discovers that he can put "-Rprofiling" into nqp-rx's makefile 19:56
Actually, I can do that for the Parrot build too. 19:57
20:02 luben left
dalek rrot: r49452 | cotto++ | trunk/src/runcore/profiling.c:
[profiling] improve inline pod docs
 20:06
20:08 tcurtis joined 20:12 Andy joined 20:16 tadzik left, cognominal left
bacek ~~ 20:17
aloha, humans
cotto bacek, aloha 20:19
mikehh how do bacek
bacek hi cotto mikehh
mikehh #ps in 10 20:20
cotto right you are
tcurtis Hi, bacek, cotto, mikehh.
20:24 ash_ left
cotto has a heisenbug 20:26
atrodo cotto> my favorite 20:27
cotto gives it to atrodo
atrodo *my favorite to ignore 20:28
sorry, enter key too close to the e
cotto nice save 20:29
20:46 pyrimidine left 20:56 plobsing_ joined
Coke also, that bug I opened. 20:56
chromatic: your fix to integer to use more init_int broke partcl-old 20:57
(and partcl-new, I think.)
chromatic Which vtable is the culprit?
Coke divide is where I saw it. 21:10
(I override division)(
int <div> float should result in a float, not an int.
(for tcl)
chromatic That should be easy to fix.
Coke here's hoping. 21:11
-> offline 21:13
21:36 nwellnhof joined
mikehh moritz: ping 21:42
aloha msg moritz it seems like a good idea to bring purl back here for a bit, at least until aloha is ready for prime time 21:44
aloha mikehh: OK. I'll deliver the message.
21:45 allison joined 21:57 davidfetter left 22:01 tcurtis left
Coke AIIIIIIIIRRRRRRRGH 22:05
-1 on bringing purl back.
cotto She had some useful features and state that aloha hasn't inherited yet.
mostly state 22:06
mikehh Coke: aloha is NOT ready for prime time yety and purl has a lot of other functionality
Coke still, -1 from me. 22:07
though my ARGH is directed at chromatic's request for PIR.
mikehh even if she can be annoying :-}
22:08 Coke left
cotto We should start a wiki page for required bot features. 22:08
mikehh karma, messaging, answer questions, not be too annoying, time 22:09
as in clock
cotto clock?
aloha, clock?
aloha cotto: cotto: LAX: Tue, 14:36 PDT / CHI: Tue, 16:36 CDT / NYC: Tue, 17:36 EDT / UTC: Tue, 21:36 UTC / LON: Tue, 22:36 BST / BER: Tue, 23:36 CEST / TOK: Wed, 06:36 JST / SYD: Wed, 08:36 EST
cotto: I give up.
22:10 kid51 joined, nopaste left, TonyC left
mikehh better than last time I looked, but needs better messaging as in message erase and stuff, and also ...? queries, like lorito? or something 22:12
cotto points mikehh at the wiki
mikehh and the time is off 22:13
22:14 plobsing_ left
kid51 on aloha, I have not been able to get "messages erase" ... but messages eventually seem to "age out" 22:15
22:20 PacoLinux left 22:22 kid51 left
chromatic Coke, alternately can you show me the Tcl PMC with the HLL map to Integer? 22:36
22:38 nwellnhof left
bacek_at_work ~~ 22:40
cotto ~~ 22:41
plobsing /\\//\\/ 22:42
22:43 TonyC joined
bacek_at_work plobsing, it's other way around! /\\\\/\\// 22:44
22:44 luben joined
plobsing ENOTAREGEX # it was an attempt at a sawtooth, but it's clipped beyond all recognition 22:45
TonyC nopaste.snit.ch will be down for 4-5 hours due to a power outage 22:46
bacek_at_work cplusplus-soup.com/2010/10/04/clang-gc/ 22:48
Yes folks, if you’re on UNIX-like platforms (Mac, BSD, Linux) then you’re in luck: the clang project in its latest release (version 2.8) is now considered a production-ready compiler for C++
22:51 dngor_ joined
bacek_at_work mikehh, aloha does have "Infobot" plugin. 22:52
aloha, lorito?
aloha bacek_at_work: lorito is "little parrot" in spanish or trac.parrot.org/parrot/wiki/Lorito
mikehh bacek_at_work: gettin' better n better 22:53
how do you tell aloha to remember somethin'
bacek_at_work mikehh, aloha just not so annoying as purl and will not respond without asking.
mikehh, just tell her
aloha, foo? 22:54
aloha bacek_at_work: I have no idea.
bacek_at_work aloha, foo is bar
aloha bacek_at_work: Okay.
bacek_at_work aloha, foo?
aloha bacek_at_work: foo is bar
bacek_at_work aloha, foo is also baz
aloha bacek_at_work: Okay.
bacek_at_work aloha, foo?
aloha bacek_at_work: foo is bar or baz
bacek_at_work aloha, forget foo
aloha bacek_at_work: I forgot about foo.
mikehh excell;ent!
bacek_at_work aloha, foo?
aloha bacek_at_work: Dunno.
bacek_at_work foo is bar
aloha, foo?
aloha bacek_at_work: foo is bar
22:55 dngor left
bacek_at_work hmm. I did switch passive learning by cotto++ request few weeks ago. 22:55
mikehh now all we need is to tell her stuff
bacek_at_work karma cotto
aloha cotto has karma of 12.
22:55 dngor_ is now known as dngor
bacek_at_work explain cotto 22:55
aloha positive: | trunk/src/runcore/profiling.c: (dalek), | trunk/src/runcore/profiling.c: (dalek), | trunk/src/runcore/profiling.c: (dalek); negative: nothing; overall: 12.
bacek_at_work mikehh, yes :)
mikehh aloha clock? 22:56
aloha mikehh: mikehh: LAX: Tue, 15:22 PDT / CHI: Tue, 17:22 CDT / NYC: Tue, 18:22 EDT / UTC: Tue, 22:22 UTC / LON: Tue, 23:22 BST / BER: Wed, 00:22 CEST / TOK: Wed, 07:22 JST / SYD: Wed, 09:22 EST
mikehh: Sorry, I don't know.
bacek_at_work mikehh, known bug...
mikehh why don't she know and the clock is 33 minutes slow
22:56 aloha left 22:57 aloha joined
mikehh 43 minutes now 22:57
ARRGGHHH 34
bacek_at_work mikehh, aloha running on VM. Which has some problem with ntpd.
aloha, clock?
aloha bacek_at_work: bacek_at_work: LAX: Tue, 15:57 PDT / CHI: Tue, 17:57 CDT / NYC: Tue, 18:57 EDT / UTC: Tue, 22:57 UTC / LON: Tue, 23:57 BST / BER: Wed, 00:57 CEST / TOK: Wed, 07:57 JST / SYD: Wed, 09:57 EST
bacek_at_work: Sorry, I don't know.
mikehh better
23:16 Andy left
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#456) fulltest) at r49452 - Kubuntu 10.10 RC amd64 (g++-4.5) 23:31
23:32 kid51 joined
kid51 ~~ 23:33
mikehh kid51: hi there 23:35
kid51 mikehh: Yo! 23:36
GeJ heya kid51. 23:40
dalek rrot: r49453 | cotto++ | trunk/src/runcore/profiling.c:
[profiling] disambiguate variable names
 23:41
rrot: r49454 | plobsing++ | trunk (6 files):
revert commits r49425, r49427, r49428, r49430, and r49431 as discussed on #ps
rrot: r49455 | plobsing++ | branches/infnan:
creating branch to resolve inf/nan constant generation issues
kid51 GeJ: hello
plobsing: thanks!
plobsing kid51: can you confirm that this resolves TTs #1813-1815? 23:50
23:50 ash_ joined
kid51 I'm testing on Darwin/PPC (with my customary config options, i.e., g==) now. 23:54
Once that's done, I'll checkout branch and see if the problem lies with g++
s/g==/g++/
plobsing g gets too much karma 23:55
kid51 This will take a while as this box is, by today's standards, small.
dalek rrot: r49456 | jkeenan++ | branches/tt1810_missing_step_tests:
Creating branch to work on TT #1810: reduce noise from missing step tests.
 23:56
rrot: r49457 | plobsing++ | branches/infnan (4 files):
import commits r49425, r49427, r49428, r49430, and r49431 for further review and testing
rrot: r49458 | plobsing++ | branches/infnan/config/auto (3 files):
forgot to add actual configure step

Powered by Cro and the Raku® Programming Language.

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

Thank you!