Parrot 2.9.1 Released | parrot.org Log: irclog.perlgeek.de/parrot/today | Nopaste: nopaste.snit.ch:8001 | GC tuning | remove deprecations |
Set by moderator on 19 October 2010.
whiteknight go team 00:08
00:10 luben joined 00:12 dmalcolm left
cotto whiteknight, nice blog post. 00:12
whiteknight I wrote a blog post? 00:13
cotto on the 14th 00:16
I just now noticed
whiteknight ah, okay 00:18
I have such a backlog of things to post about, and no time to do any posting
luben about the GC related issues, there is a branch gc_ms2_tuning from nick wellnhof that dynamically adjust gc_treshold based on olready allocated space.
s/olredy/already/ 00:19
some weeks ago I have benchmarked it and it is faster than current fixed size gc_treshold (benchmarked with 32,64,128,256MB) 00:21
the benchmark was rakudo build time
chromatic Does it have an upper bound?
luben I do not think so 00:22
the times where 4 min 45s for gc_ms2(256MB), 4 min 5 sec for gc_ms2_tunung and 4 min 10 sec for the old gc_ms 00:24
this is on 4G RAM amd64
chromatic Let's get kid51 to test that branch then. 00:28
00:32 dngor_ joined
cotto couldn't you just test it with ulimit -v 262144 ? 00:34
00:34 dngor left, dngor_ is now known as dngor
chromatic That too. 00:35
00:36 theory left 00:38 dngor_ joined 00:40 dngor left 00:53 silug joined
atrodo keeps forgetting his server only has 256m 00:58
00:59 dngor_ is now known as dngor
dukeleto Perhaps Parrot has use the infrastructure that theory++ is building for PGXN to solve the distribution of module problem: blog.pgxn.org/post/1353947079/creat...tributions 01:04
s/has use/can use/ 01:05
it uses JSON, which is very similar to Plumage 01:06
01:08 bacek left
luben ok, tested the head and gc_ms2_tuning branch, both compile rakudo given 1G of memory (umilit -v), without restrictions parrot master finishes compilation in 3m 54s, gc_ms2_tuning branch finishes in 3m 16s. Given 512M memory, master crashes in parrot execution, gc_ms2_tuning crashes in gcc. 01:16
cotto is that by chance when it tries to build the fakecutable? 01:19
01:27 dngor left
luben when builing perl6.o from perl6.c 01:30
ops. sorry, mistake
nopaste "luben" at 192.168.1.3 pasted "rakudo build with 512M ram" (6 lines) at nopaste.snit.ch/24612 01:33
luben actually yes, this is the last step before linking perl6 executable
plobsing luben: can you run perl6.c through a linting tool (or a syntax highlighter even maybe) to see if it is valid C? maybe we're outputting garbage. 01:36
alternatively, do you have other C compilers?
luben its not garbage, it eats a lot of memory 01:37
plobsing it causes gcc to eat a lot of memory? 01:38
luben I could try with gcc-4.3 01:41
plobsing that could work. less -O might work as well.
why are we all so worked up about rakudo not building properly on small systems. Can't we pre-build a perl6.pbc to install on those systems? 01:43
cotto probably because we have a developer who has a small system and likes being noisy when he can't build there 01:44
plobsing that's fair 01:45
02:05 whiteknight left
dalek rrot: r49589 | bacek++ | branches/generational_gc/src/gc/gc_ms2.c:
Add some stats output when DETAILED_MEMORY_DEBUG defined.
 02:07
rrot: r49590 | bacek++ | branches/generational_gc/src/gc/gc_ms2.c:
Sweep strings as well.
rrot: r49591 | bacek++ | branches/generational_gc/src/gc/gc_ms2.c:
Remove "constant" objects from pools to avoid processing them over and
02:14 benabik joined 02:17 Tekk_ joined
Tekk_ "load_bytecode" couldn't find file 'distutils.pbc' 02:17
I keep getting that error when I try to build a language
with parrot setup.pir :P tried with shakespeare and bf 02:18
plobsing find $PATH_TO_PARROT -name distutils.pbc # might show you if it is there or not 02:19
(it should be there)
Tekk_ nothing came up 02:21
is there like a cpan for parrot? {#
:P?
*
cant type tonight ><
plobsing plumage is supposed to be that (I think). but it is mostly built on top of distutils.pbc and setup.pir 02:22
chromatic Not quite yet.
plobsing so that won't really help
Tekk_ nope
plobsing distutils.pbc is supposed to be part of core
Tekk_ apparently it isn't for the debian package :P
plobsing oh then you probably need to also install libparrot-annoyingly-broken-out-package-of-some-kind 02:23
chromatic Is it in a parrot-dev package?
Tekk_ nope
just installed and che....
benabik Is there a parrot-dev package?
Tekk_ every time I install something more parrot packages show up O.o
benabik (Guh, slow typing.)
Tekk_ okay, maybe parrot devel
way to break the naming convention guys, I thought this was debian not fedora >> 02:24
not in parrot-devel either
chromatic libparrot-dev? 02:26
Tekk_ nope 02:29
I have everything parrot related installed except for the docs
in main at least
on the plus side it meant I got to submit my first debian bug report :P
ew O.o 02:31
plobsing doesn't look like distutils.pbc falls under any of the rules in ports/debian/*.install.in
Tekk_ I thought debian testing was going to be reasonably up to date, it's using 2.0.0
plobsing which is probably the problem
Tekk_ oh right, it was frozen
benabik distutils.pbc doesnt appear to be in any debian package, according to the search on debian.org 02:32
plobsing steps to remedy: 1) figure out which debian package it should be under 2) add it to the appropriate .install.in 3) rejoice 02:33
Tekk_ plobsing: parrot-devel I would guess 02:34
benabik Is github.com/parrot/parrot official? It's four days behind (notably missing 2.9.0 and 2.9.1). 02:35
02:35 janus left
plobsing benabik: it is the eventual location of the repo. we have yet to complete our transition to git 02:36
until then, the official location is svn.parrot.org/parrot/trunk
benabik plobsing: So it may be updated from time to time, but I can't rely on it. Check. 02:37
plobsing Tekk_: I don't know much about debian packaging rules, so I'll take your word for it (someone can always fix it later) 02:41
Tekk_ plobsing: yeah, I filed the bug report so it should be taken care of
plobsing Tekk_: it's a problem on our end.
dukeleto benabik: official-ish 02:42
02:42 janus joined
dukeleto benabik: i am updating it now 02:42
Tekk_ plobsing: how so?
plobsing we provide the debian control files used in those packages. debian will most likely just roll over and ask us to fix it.
Tekk_ ah, okay
benabik dukeleto: awesome. I can stop trying to get git-svn from syncing. 02:44
dukeleto benabik: it is updated now, with the new tags 02:48
benabik: git-svn-ing the whole history takes many hours 02:49
plobsing Tekk_: control file has been updated appropriately in r49592. you can build the deb package yourself by following docs/project/debian_packaging_guide.pod from an svn checkout, or wait a couple months until debian updates.
benabik dukeleto: I know that. You can use a git clone to jump-start the process. Just have to make sure the config is just right. 02:50
dukeleto: I'd still rather not bother. Thanks for updating it. 02:51
02:52 Tekk_ left, kid51 joined
dalek rrot: r49592 | plobsing++ | trunk/ports/debian/parrot-devel.install.in:
add distutils.pbc to parrot-devel debian package
 02:53
dukeleto benabik: the sha1's of parrot.git will probably change again soon, to unify our authors. Just so ya know.
kid51 ~~
dukeleto kid51: howdy
kid51 hello
benabik dukeleto: Fair enough. I just prefer the git clone for bisecting and the like. 02:54
dukeleto benabik: indeed. it makes life a bunch nicer 02:55
benabik: the branches with /'s in them on github are git-only branches. All other branches are svn branches 02:56
03:05 dngor joined 03:16 benabik left 03:25 kid51 left 03:47 brianwisti joined 04:15 theory joined
GeJ aloha, clock? 04:35
aloha GeJ: GeJ: LAX: Tue, 21:35 PDT / CHI: Tue, 23:35 CDT / NYC: Wed, 00:35 EDT / UTC: Wed, 04:35 UTC / LON: Wed, 05:35 BST / BER: Wed, 06:35 CEST / TOK: Wed, 13:35 JST / SYD: Wed, 15:35 EST
GeJ: I give up.
bacek_at_work aloha, clock? 04:43
aloha bacek_at_work: bacek_at_work: LAX: Tue, 21:43 PDT / CHI: Tue, 23:43 CDT / NYC: Wed, 00:43 EDT / UTC: Wed, 04:43 UTC / LON: Wed, 05:43 BST / BER: Wed, 06:43 CEST / TOK: Wed, 13:43 JST / SYD: Wed, 15:43 EST
bacek_at_work: I have no idea.
bacek_at_work aloha, clock
aloha, clock? 04:59
aloha bacek_at_work: bacek_at_work: LAX: Tue, 21:59 PDT / CHI: Tue, 23:59 CDT / NYC: Wed, 00:59 EDT / UTC: Wed, 04:59 UTC / LON: Wed, 05:59 BST / BER: Wed, 06:59 CEST / TOK: Wed, 13:59 JST / SYD: Wed, 15:59 EST
bacek_at_work ok. I fixed clash between Infobot and Clock modules in aloha.
cotto I need to clarify that I value kid51's input about his issues with the Rakudo build. I apologize if anyone (especially him) felt like I was putting him down for making that problem known. 05:00
05:07 brianwisti left 05:20 tcurtis left 05:37 theory left
dalek rrot: r49593 | bacek++ | branches/generational_gc/src/gc/gc_ms2.c:
More statistic output
 05:42
06:17 theory joined
dalek rrot: r49594 | bacek++ | branches/generational_gc/src/gc (2 files):
Add calculation of Pool_Allocator allocated memory
 06:28
rrot: r49595 | bacek++ | branches/generational_gc/src/gc/gc_ms2.c:
Use pool_allocated_memory function
rrot: r49596 | bacek++ | branches/generational_gc/src/gc (2 files):
Add function to calculate allocated memory in Fixed_Size_Allocator
rrot: r49597 | bacek++ | branches/generational_gc/src/gc/gc_ms2.c:
Output amount of allocated attributes
rrot: r49598 | bacek++ | branches/generational_gc/src/gc/gc_ms2.c:
Fix sweep_pmc_cb to actually deallocate PMC.
rrot: r49599 | bacek++ | branches/generational_gc/src/gc/gc_ms2.c:
Increase amount of allocated memory between collects to 16MB.
06:32 theory left 06:43 chromatic left 07:21 contingencyplan left 07:37 fperrad_ joined 07:40 fperrad_ is now known as fperrad 08:18 bacek joined
bacek aloha, humans 08:28
cotto aloha, bacek 08:30
bacek cotto, hi 08:31
cotto, do you win32 box around? 08:34
cotto nope
something about physical memory detection? 08:35
bacek yes
cotto bacek++ 08:38
I should buy a cheap used windows laptop for testing. 08:42
bacek aloha, taptinder? 08:49
aloha bacek: taptinder is continues integration tool - taptinder.org . For Parrot project running on tt.taptinder.org/ and reporting build failures to #parrot channel as ttbot.
bacek ok. We have win32 box in taptinder. 08:50
here we go 08:51
ttbot Parrot trunk/ r49602 MSWin32-x86-multi-thread make error tt.taptinder.org/file/cmdout/411137.txt ( tt.taptinder.org//buildstatus/pr-Pa.../rp-trunk/ ) 08:58
cotto bacek, the man page for sysconf on my system indicates that _SC_AVPHYS_PAGES and _SC_PHYS_PAGES may not be standard. 09:01
bacek yes 09:02
but it works on many POSIX compliant systems
dalek rrot: r49600 | bacek++ | trunk (4 files):
Add Parrot_sysmem_amount to get available memory.
rrot: r49601 | bacek++ | trunk/config/gen/platform/win32/sysmem.c:
Implement Win32 version of Parrot_sysmem_amount based on MSDN example.
rrot: r49602 | bacek++ | trunk/src/gc/gc_ms2.c:
Use 1/8 of available memory for gc_threshold
cotto We should have a fallback for the systems where it doesn't.
sorear should just clamp sysmem_amount to a sane range 09:04
64MB-8GB should be good
bacek oookey... I broke win32 build. Somehow MEMORYSTATUSEX is not defined. 09:05
cotto that's what you get for reading the documentation
I guess
bacek msdn.microsoft.com/en-us/library/aa...S.85).aspx 09:08
I _did_ include <windows.h>...
cotto yeah
bacek www.perlmonks.org/?node_id=749964 09:15
dalek rrot: r49603 | bacek++ | trunk/config/gen/platform/win32/sysmem.c:
Blind commit to fix win32 build.
 09:18
cotto You could also try the solution in the last comment if ttbot keeps complaining. 09:19
bacek cotto, I will. But I prefer someone with win32 box to actually test it 09:23
cotto good find
09:24 silug left
ttbot Parrot trunk/ r49603 MSWin32-x86-multi-thread make error tt.taptinder.org/file/cmdout/411212.txt ( tt.taptinder.org//buildstatus/pr-Pa.../rp-trunk/ ) 09:30
cotto sad ttbot is sad 09:32
bacek, on the bright side trying that fix from perlmonks won't make the windows build more broken 09:34
bacek cotto, r49604 09:35
cotto whee 09:37
09:44 aloha left
ttbot Parrot trunk/ r49604 MSWin32-x86-multi-thread make error tt.taptinder.org/file/cmdout/411238.txt ( tt.taptinder.org//buildstatus/pr-Pa.../rp-trunk/ ) 09:44
09:45 bacek_at_work left
bacek still broken... 09:45
09:46 mikehh left 09:47 aloha joined
cotto aloha, aloha 09:47
dalek rrot: r49604 | bacek++ | trunk/config/gen/platform/win32/sysmem.c:
Second blind commit to fix win32 build based on �www.perlmonks.org/?node_id=749964
 09:49
rrot: r49605 | bacek++ | trunk/config/gen/platform/win32/sysmem.c:
Last blind commit to fix win32.
rrot: r49606 | bacek++ | trunk/config/gen/platform/win32/sysmem.c:
Close #ifdef statement.
bacek interesting. What happened to my devbox at $work... 09:52
seen fperrad 09:53
aloha, seed fperrad
aloha, clock?
sigh...
ttbot Parrot trunk/ r49605 MSWin32-x86-multi-thread make error tt.taptinder.org/file/cmdout/411263.txt ( tt.taptinder.org//buildstatus/pr-Pa.../rp-trunk/ )
cotto aloha, seen bacek 09:54
Hmm: config/gen/platform/win32/sysmem.c:24:1: warning: "_WIN32_WINNT" redefined
bacek we do need someone on win32 to fix it properly.. 09:55
cotto It's hard to escape that conclusion. 09:56
09:56 bacek_at_work joined
bacek hey! I'm not _at_work now! 09:58
09:59 aloha left
cotto everything is not coming up roses 09:59
10:03 bacek_at_work left 10:04 bacek_at_work joined 10:05 lucas joined
bacek ookey 10:05
win32 builds now 10:06
cotto bacek++
I bet that was more fun than you were hoping for. 10:07
bacek we need someone on "low-memory-box" to validate r49602
cotto will that patch be sensitive to ulimit?
bacek I don't like "blind commits"...
Unlikely
(it's about ulimit) 10:08
cotto it's sleeping time 10:10
night
bacek cotto, night 10:12
10:13 mikehh joined 10:23 aloha joined
dalek rrot: r49607 | mikehh++ | trunk/MANIFEST:
re-generate MANIFEST
 10:35
10:47 arnsholt joined
dalek rrot: r49608 | mikehh++ | trunk (3 files):
add svn properties
 10:50
10:54 dngor_ joined 10:55 dngor left
dalek rrot: r49609 | mikehh++ | trunk/config/gen/platform/win32/sysmem.c:
fix codetest failures - there should be at least one space between a C

fix documentation item header		 11:06
rrot: r49610 | mikehh++ | trunk/config/gen/platform/generic/sysmem.c:
correct c function docs item header
11:16 dngor joined, dngor_ left
dalek rrot: r49611 | mikehh++ | trunk/include/parrot/sysmem.h:
fix codetest failure - add Id keyword line
 11:21
11:52 bacek_at_work left 11:53 aloha left 11:58 lucas left 12:01 bacek_at_work joined 12:02 simcop2387 left 12:03 simcop2387 joined 12:04 aloha joined 12:13 bacek_at_work left, bacek_at_work joined
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#716) fulltest) at r49611 - Kubuntu 10.10 amd64 (gcc-4.5) 12:14
12:19 fperrad left 12:22 tcurtis joined, tcurtis left 12:44 contingencyplan joined 13:07 fperrad joined 13:56 bacek_at_work left 13:59 aloha left 14:06 bacek_at_work joined 14:21 whiteknight joined
whiteknight good morning, #parrot 14:39
14:42 brianwisti joined 14:43 brianwisti left 14:48 brianwisti joined 14:51 aloha joined 15:04 he left, theory joined 15:05 dmalcolm joined
dukeleto whiteknight: howdy 15:07
whiteknight hello duke. How are you doing?
dukeleto whiteknight: pretty good. I am in California for the GSoC mentor summit and the GIT Together 15:09
whiteknight: we are going to be writing a "student manual" in the next few days
whiteknight: what are you hacking on lately? 15:11
brianwisti dukeleto: You're just all over the place, aren't you? 15:14
whiteknight dukeleto: I haven't been doing much hacking lately at all, unfortunately. I've been doing some codereview on the generational_gc branch for bacek, but I can't hardly keep up with reading all the code he writes
my output and availability should increase pretty dramatically starting next weekend I hope 15:15
15:16 krunen left
dukeleto brianwisti: yeah, my position field is a bit fuzzy 15:19
whiteknight: glad to hear someone is code reviewing all that code!
whiteknight: can you summarize what he is doing in a few sentences?
whiteknight "taking all the old shit out and replacing it with something better" 15:26
actually, the new GC algorithm that he's using is very cool. It's like a normal generational GC algorithm but including chromatic's idea of limiting GC sweep to only a subset of objects that are known to have been alive 15:27
where the old system swept over all of raw memory, the new system only sweeps objects which had been on the "alive" list
objects that pass two consecutive sweeps bump up to an older generation. Older generations are swept even less frequently than average 15:28
plus, I think he doesn't need to mark the older generations either, but I may be wrong about that idea
dukeleto whiteknight: cool stuff 15:29
15:30 muixirt joined
muixirt hi there 15:31
whiteknight it is very interesting. I need to dig through more of it because there is some significant complexity added in some areas
hello muixirt
muixirt I'm a little confused about nqp, pct and pge, what depends on what?
hi whiteknight
whiteknight muixirt: Good question. I think NQP is stand-alone now 15:32
muixirt nqp replace compilers/pge, right? 15:33
PerlJam yes
whiteknight the new NQP, yes
muixirt but nqp depends on compilers/pct (?) 15:34
whiteknight I don't think it does, no
maybe some of the libraries from it
compilers/pct does contain definiitions for the PAST and POST node objects 15:35
pmichaud nqp depends on pct, yes. 15:39
pge is basically obsolete 15:40
whiteknight okay, that's what I thought
pmichaud we're expecting to rewrite much of the pct stuff in nqp
muixirt thanks pmichaud and whiteknight 15:41
muixirt thought about dependency on pir and was confused (as most of the time) 15:42
so stuff written in pir will go extinct due to lorito? 15:49
dukeleto muixirt: kinda 15:54
muixirt: PIR will be used in Parrot internals, but people outside parrot core should have better tools to use than PIR
whiteknight muixirt: Initially, I think that we will be considering PIR to be an "HLL" that runs on top of Lorito 15:57
in the same way that an HLL like Rakudo runs on top of PIR/PASM now
muixirt whiteknight: for what purpose (other than supporting legacy code for some time)? 15:59
15:59 aloha left
brianwisti Would that shift to PIR as HLL for Lorito eventually phase out PIR itself? 16:00
you kinda answered. Let me rethink my question.
muixirt dukeleto: why would parrot core devs use an inferior tool than "people outside parrot core"?
16:02 bacek_at_work left
brianwisti Here's the real question: should I start thinking about using something other than PIR for the Parrot Babysteps? 16:02
dukeleto brianwisti: good question. when something better exists? 16:03
brianwisti muixirt: The opcodes (PIR / Lorito) are perfect for somebody doing low-level work in Parrot. Most developers who use Parrot won't be working on such a low level, so it won't be a good tool for them. 16:04
dukeleto muixirt: it isn't so much inferior as too low-level for most people
muixirt: we are hoping that people won't need to write C or PIR to hack on stuff. We are hoping they will be able to use their HLL of choice
16:08 he joined
atrodo whiteknight> re your pcc blog posting. Do you know why pcc is so complicated? Is there a significant speed advantage over doing something simpler, like simply having a resizable array of arguments? 16:10
16:14 bacek_at_work joined
dukeleto has been notified that my blog post about GSoC this summer is wanted for the Google Open Source Blog. They are asking me if I can add any pictures to snazz it up. Halp? 16:16
16:32 baest left 16:38 bacek_at_work left 16:39 aloha joined 16:42 bacek_at_work joined
cotto aloha, clock? 16:43
aloha cotto: cotto: LAX: Wed, 09:43 PDT / CHI: Wed, 11:43 CDT / NYC: Wed, 12:43 EDT / UTC: Wed, 16:43 UTC / LON: Wed, 17:43 BST / BER: Wed, 18:43 CEST / TOK: Thu, 01:43 JST / SYD: Thu, 03:43 EST 16:45
cotto aloha, clock?
aloha cotto: cotto: LAX: Wed, 09:45 PDT / CHI: Wed, 11:45 CDT / NYC: Wed, 12:45 EDT / UTC: Wed, 16:45 UTC / LON: Wed, 17:45 BST / BER: Wed, 18:45 CEST / TOK: Thu, 01:45 JST / SYD: Thu, 03:45 EST
cotto ...ok
seen kid51 16:50
aloha kid51 was last seen in #parrot 13 hours 56 mins ago saying "hello".
17:19 silug joined
dukeleto this is cool: pgbuildfarm.org/cgi-bin/show_status.pl 17:25
ttbot Parrot trunk/ r49612 MSWin32-x86-multi-thread make error tt.taptinder.org/file/cmdout/411476.txt ( tt.taptinder.org//buildstatus/pr-Pa.../rp-trunk/ ) 17:27
brianwisti dukeleto: You want to see a parrot icon in there someday, don't you?
17:30 muixirt left
dukeleto brianwisti: indeed I do :) 17:30
brianwisti: i want to see a similar parrot build farm even more
brianwisti That would be lovely.
dukeleto brianwisti: trac.parrot.org/parrot/ticket/1825 tracks the progress on that 17:31
cotto fperrad, the code in 49612 was necessary gcc on windows: tt.taptinder.org//buildstatus/pr-Pa.../rp-trunk/
dalek rrot: r49612 | fperrad++ | trunk/config/gen/platform/win32/sysmem.c:
[win32] fix build with mingw
rrot: r49613 | fperrad++ | trunk/config/gen/platform/generic/stat.c:
fix warning on Windows
ttbot Parrot trunk/ r49613 MSWin32-x86-multi-thread make error tt.taptinder.org/file/cmdout/411480.txt ( tt.taptinder.org//buildstatus/pr-Pa.../rp-trunk/ ) 17:32
17:32 fperrad left
whiteknight cotto: ping 17:45
cotto whiteknight, pong
whiteknight cotto: I would like to be a member of the architecture team, please 17:46
:)
cotto ok
I guess that should be documented somewhere.
whiteknight I read through the #ps notes from yesterday. I don't know how I would go about accepting my nomination for the product management team 17:48
or if that constituted a real "nomination" anyway
cotto I thought you'd already accepted. 17:49
whiteknight ...sure, we'll just say that already happened
cotto At this point, accepting means saying "ok" where people will know about it.
whiteknight is this teams thing like a done-deal? Do we just assume we are moving forward with it? 17:50
cotto We will be using the structure, but we can change it as needed. 17:51
trac.parrot.org/parrot/wiki/ParrotTeamMembers 17:56
whiteknight cotto++
First step is to figure out what the hell the Product Management team is supposed to be doing. Second step is to do that thing better than any other teams! 17:57
cotto For now, PirateTodo may be a good place to start. 17:58
dukeleto goes and reads the logs for the last #ps
cotto dukeleto, I didn't mention your question about a weekly video chat meeting. I have it on my list for the next #ps, which hopefully will be less chaotic. 18:00
18:02 baest joined 18:05 aloha left 18:06 lucian joined 18:12 PerlJam left 18:15 PerlJam joined 18:16 bacek_at_work left
dukeleto cotto: yeah, no worries. Getting the bugfix release out was more important 18:18
dukeleto just backlogged the #ps. It was quite exciting.
18:19 bacek_at_work joined
cotto looks forward to a more boring #ps where not as many things are on fire 18:20
also, one where both kid51++'s laptop and the Rakudo guys are happy 18:24
atrodo speaking of which, I need to let make test finish 18:30
it was up to 30 minutes w/2.9.1 earlier today 18:31
cotto do you have a low-resource machine or is something just broken?
atrodo I have a 256m server, apparently 18:32
i tried a few branches last night, but couldn't get anything to run slower than 4 minutes. so today i'm trying a download of 2.9.1
18:35 rbuels left
mikehh the change by fperrad in r49612 seems to have broken the win32 build on TapTinder again 18:38
looks like we need something different for mingw vs win32 18:39
cotto I mentioned it to him but didn't get a response. 18:40
18:40 bacek_at_work left 18:41 bacek_at_work joined 18:44 fperrad joined 18:46 mikehh left, mikehh joined 18:55 aloha joined 19:01 Kulag left, Kulag joined
dukeleto so are the new parrot teams "go for launch" ? 19:16
cotto yes 19:20
We'll find out exactly what that means as we progress.
whiteknight dukeleto: I would like to hereby solicit your participation in the "Product Management" team, if you're interested 19:22
the team has at least some focus on extensions and embedding, and PLParrot is an important project in that regard
PerlJam Where are these teams described? 19:24
cotto PerlJam, trac.parrot.org/parrot/wiki/ParrotTeams
PerlJam thanks 19:26
cotto you're welcome
PerlJam Does the Product Management team also do "marketing"? 19:28
whiteknight I'm not sure yet 19:29
I'm still formulating a comprehensive vision of the team. I'm currently thinking that it's going to be a little bit more technical, working with existing users to improve parrots interfaces
adding a marketing component, or even creating a new separate marketing team might be a good idea though 19:30
I wonder also if that might overlap with the community team
cotto It's somewhere between community and product, perhaps more on the product side.
I don't want to create any more teams until we have some idea how the current ones will function. 19:31
whiteknight marketing is a very important role though. We do need to make sure it fits in somewhere. Reading the definitions of the teams, it does sound to me like more of the Community team's job 19:32
19:32 x3nU left
whiteknight I'm sure we'll all take a share of the pie, however 19:32
cotto agreed 19:33
It needs to get done.
19:34 x3nU joined
PerlJam That's not how it reads to me. Community Management seems to be about proselytizing new developers where as (to me) marketing is about converting new users. 19:34
whiteknight ah, that is a slightly different reading 19:35
Maybe I'm too jaded into thinking that there is little divide between our current users and developers
Okay, Product Management team has dibs on marketing. No callbacks
PerlJam heh
parrot users are developers just of a different sort . They're building things on top of parrot, rather than building parrot. 19:36
whiteknight right
cotto that's why it can be confusing to talk about our users 19:37
NotFound Can we ask access to the GCC compile farm individually, or are we going to use it only in some automated way?
whiteknight NotFound: automated, I hope
smoke reports on a cron job would be the best for now 19:38
NotFound whiteknight: note the "only"
whiteknight if we need to do development on an exotic system, I'm sure we could also use it for that
cotto unless it's windows 19:39
PerlJam Is there a place that lists what people are using parrot for today? 19:40
cotto sounds like a good marketing idea 19:42
whiteknight very good idea 19:44
PerlJam: you want to hop abord the S.S. Product Management Ship?
PerlJam whiteknight: Assuming I haven't already exhausted my ability to contribute, sure :) 19:46
whiteknight PerlJam: Well, don't strain yourself! We don't want our contributors burning out, or the Community team might hunt me down and hit me with a stick 19:48
ttbot Parrot trunk/ r49614 MSWin32-x86-multi-thread make error tt.taptinder.org/file/cmdout/411706.txt ( tt.taptinder.org//buildstatus/pr-Pa.../rp-trunk/ )
dalek rrot: r49614 | fperrad++ | trunk (4 files):
fix and use parrot/sysmem.h
 19:50
cotto fsvo "fix" 19:54
20:08 aloha left 20:10 aloha joined 20:16 whiteknight left 20:23 aloha left 20:24 aloha joined 20:38 lucian left 20:55 aloha left 21:12 aloha joined 21:17 cognominal left 21:20 cognominal joined
dalek rrot: r49615 | fperrad++ | trunk/config/gen/platform/win32/sysmem.c:
partial revert of r49612, but with a better condition guard,
 21:23
mikehh fperrad++ 21:37
dalek rrot: r49616 | fperrad++ | trunk/include/parrot/thr_windows.h:
fix warning with MinGW gcc 3.4.5
 21:38
mikehh fperrad: you also managed to fix the g++ build (in r49614), I got the changes in gc_ms2.c, but couldn't figgure out the other additions of sysmem.h 21:40
fperrad generic Parrot_sysmem_amount() doesn't compile on cygwin, see TapTinder (machine #12) 21:53
21:56 fperrad left 22:05 dmalcolm left 22:08 aloha left 22:09 brianwisti left 22:11 bacek_at_work left
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#721) fulltest) at r49616 - Kubuntu 10.10 amd64 (g++-4.5 with --optimize) 22:11
22:11 bacek_at_work joined
ttbot Parrot trunk/ r49616 cygwin-thread-multi-64int make error tt.taptinder.org/file/cmdout/411800.txt ( tt.taptinder.org//buildstatus/pr-Pa.../rp-trunk/ ) 22:16
22:17 aloha joined 22:32 dngor_ joined 22:36 dngor left, dngor_ is now known as dngor 22:37 kid51 joined
mikehh bacek_at_work: ping 22:39
dalek rrot: r49617 | mikehh++ | trunk/config/gen/platform/generic/sysmem.c:
an attempt to get parrot to build on cygwin (on TapTinder)
 23:08
23:18 nwellnhof joined
kid51 parrot built -- albeit *very* slowly on Darwin at r49591 -- but not since 23:19
nopaste "kid51" at 192.168.1.3 pasted "Darwin: Parrot build failure somewhere after r49591" (6 lines) at nopaste.snit.ch/24641 23:21
kid51 Hmm, lemme see if I can get a more complete 'make' log 23:23
nwellnhof bacek: ping 23:27
i had a look at the gen_gc branch. one thing i noticed: 23:29
in gc_ms2_mark_and_sweep the root_objects list is always cleared. 23:30
so, after a gen0 GC, we discard all gen2 -> gen1 references.
that doesn't seem right to me. 23:31
bacek_at_work mikehh, pong 23:32
nwellnhof, no. lines 723-731. I move root_objects into proper generations
nwellnhof but we lose track of the inter-generational reference. 23:33
nopaste "kid51" at 192.168.1.3 pasted "complete log: Darwin: Parrot build failure somewhere after r49591" (6 lines) at nopaste.snit.ch/24642
mikehh bacek_at_work: I was gonna ask about r49617 but looking at nopaste.snit.ch/24641, probably need something similar for Darwin and _SC_AVPHYS_PAGES 23:34
kid51 What is wrong with nopaste.pl?
bacek_at_work nwellnhof, why?
nwellnhof afaics, the inter-generational references are tracked in the root_objects list, right? 23:35
nopaste "kid51" at 192.168.1.3 pasted "complete log (I hope): Darwin: Parrot build failure somewhere after r49591" (359 lines) at nopaste.snit.ch/24643 23:36
bacek_at_work mikehh, kid51 look at www.opensource.apple.com/source/gcc.../physmem.c There is a lot of examples to get physical memory on different platform
nwellnhof, no. root_objects are filled by write_barrier. After m&s I move all x-referenced objects into same generation. 23:37
nwellnhof, gc_ms2_bring_them_together
nwellnhof looking at gc_ms2_bring_them_together 23:39
bacek: what about the FIXME in line 856? 23:47
mikehh kid51: if you look at TapTinder (error report for machine 12 (cygwin), that is what I attempted to fix in r49617 (no test yet) 23:49
kid51: if you find what the equivalent of _SC_AVPHYS_PAGES we can try something similar for that 23:50
23:50 s1n joined
bacek_at_work nwellnhof, basically I shouldn't iterate whole generation for bringing them together. 23:51
mikehh mind you I still don't know if r49617 worked for cyhwin 23:52
cygwin
bacek_at_work _SC_AVPHYS_PAGES should work on cygwin. 23:53
nwellnhof bacek: but gc_ms2_bring_them_together doesn't add any objects to the root_objects list. so root_objects is always empty after mark_and_sweep, right?
bacek_at_work nwellnhof, correct 23:54
root_objects is populated by trace_root_objects or write_barrier
mikehh bacek_at_work: yes, but kid51 is looking at his Darwin maching which failed on that
s/maching/machine/ 23:55
23:56 davidfetter joined
bacek_at_work mikehh, we need different approach on darwin and *bsd. 23:56
{ /* This works on *bsd and darwin. */
unsigned int usermem;
size_t len = sizeof usermem;
static int mib[2] = { CTL_HW, HW_USERMEM };
if (sysctl (mib, ARRAY_SIZE (mib), &usermem, &len, NULL, 0) == 0
&& len == sizeof (usermem))
return (double) usermem;
kid51 Okay, I just posted where things stand for me here: trac.parrot.org/parrot/ticket/1829#comment:3
bacek_at_work }
23:56 kid51 is now known as kid51_at_dinner 23:57 cotto left
nwellnhof bacek: let me give you an example... 23:59
23:59 cotto 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!