Parrot 3.0.0 Released | parrot.org | Log: irclog.perlgeek.de/parrot/today | Goals: Fix ipv6-related failures
Set by moderator on 10 February 2011.
00:00 protorom joined 00:04 protorom left
dalek TT #2009 closed by jkeenan++: Remove dead svn-related code from Perl modules 00:09
TT #2009: trac.parrot.org/parrot/ticket/2009
whiteknight in Perl6, what is the =:= operator? 00:14
dalek rrot-linear-algebra: 6ece133 | Whiteknight++ | t/ (9 files):
more fixes
rrot-linear-algebra: 0a3cf06 | Whiteknight++ | t/methods/complexmatrix2d/ (5 files):
final fixes. All tests pass
sorear whiteknight: checks if two lvalue expressions are aliases
00:26 theory joined
mikehh Hey folkes - How about some more NEWS updates 00:28
whiteknight I don't think I've committed anything to master this whole cycle 00:35
mikehh whiteknight: you can comment on parrot-linear-algebra in Languages or something :-} 00:36
whiteknight I haven't done anything to it, except unbreak it 00:37
plobsing why do we feel we need to stuff the kitchen sink into the news associated with our releases?
mikehh dunno - maybe to make it look more impressive :-} 00:38
hey I am just following protocol here 00:39
whiteknight pla is much better than the kitchen sink
plobsing whiteknight: I'm not objecting to PLA being added to NEWS. I'm just stating my opinion that we tend to pad news out a little bit too much, leading to poor signal-to-noise ratio. 00:41
whiteknight I agree. I'm not going to write about PLA in NEWS
plobsing which is why most of my endeavours which make it into the news are not put there by me
mikehh plobsing: how about some super impressive thing you have done recently? 00:42
like some of your IMCC work etc. 00:43
plobsing I tend to view most of my contributions as internal, and (hopefully) not user visible. They are not news-worthy.
whiteknight NEWS is precisely to shine light on internal improvements that users might not know about otherwise 00:44
mikehh course they are - to devs anyway
and how many persons, other than devs actuall read NEWS anyway? 00:45
actually
kid51 linux/i386: make fulltest PASS at cccb01e71961 (gcc all around, non-optimize) 00:46
plobsing news is completely redundant if it is written for an audience of core or core-familiar devs. they already know this stuff.
00:46 bacek left
plobsing I always viewed news as a kind of status update for people sitting on the sidelines. the kind of people we'd like developping HLLs and libraries. 00:47
to them, it doesn't matter that, taking an old example, dynops work from PBC now (even though our documentation suggested this was always true). I read that one as "hey, we were lying to you. but now we're not. awesome!" 00:48
mikehh that too of course, however I think a lot of devs concentrate on a specific area and other areas are relevant to them
I certainly find the workings of IMCC pretty opaque, any improvements there are news to me 00:50
plobsing mikehh: but do they *affect* you? 00:51
bacek_at_work blog.urth.org/2011/01/changes-file-...ot-to.html
jfyi 00:52
00:55 protorom_ joined 00:56 protorom_ left
mikehh bacek_at_work: very appropriate 00:58
plobsing my reading of this article reinforces my opinion that news files should be streamlined to only show changes from within the project that affect external users 01:02
we do get the organization thing right. and regularly scheduled releases make the release date-stamp less of an issue.
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#9470) fulltest) at 3_0_0-937-gcccb01e - Ubuntu 10.10 i386 (gcc-4.5) 01:18
01:26 lucian left 01:27 luben left 01:37 Maurice left 01:38 cosimo joined
whiteknight gitorious is slow 02:08
slow..ow...ow...ow
plobsing what's hosted on gitorious? 02:17
02:24 lateau_ joined 02:27 kid51 is now known as kid51_at_dinner
bacek_at_work msj jnthn jfyi, we have statement_mod_loop:sym<for> in nqp. "for @!mro { @tc.push($_); }" can be written as "@tc.push($_) for @!mro" :) 02:27
dalek rrot-test: eab52e0 | Whiteknight++ | mockobject/ (6 files):
a second path over the mockobject classes
 02:29
rrot-test: 9b30b2a | Whiteknight++ | / (2 files):
fixes to MockFS. Add the mockobject library to setup.nqp. It builds. I haven't tried running it
rrot-test: 9baa193 | Whiteknight++ | / (2 files):
add a missing file, and update .gitignore
rrot-test: bc8a6ae | Whiteknight++ | tap_harness/ (4 files):
Add in stubs of a library for a TAP consumer harness library. Borrowed from PLA
plobsing bacek_at_work: you might want to s/msj/msg/
bacek_at_work msg jnthn jfyi, we have statement_mod_loop:sym<for> in nqp. "for @!mro { @tc.push($_); }" can be written as "@tc.push($_) for @!mro" :)
aloha OK. I'll deliver the message.
bacek_at_work plobsing, ouch. Thanks :)
whiteknight bacek is so good, it doesn't matter what he types. The correct message will always get to the correct recient 02:30
dalek is being slow tonight, methinks
bacek_at_work whiteknight, what is parrot-test? 02:31
02:32 luben joined
whiteknight unitteseting library 02:32
unit testing
bacek_at_work whiteknight, how it's different from existing Test::More in parrot? 02:33
whiteknight At the moment it uses xUnit style tests instead of Test::More test 02:34
also, I want to include other helper libraries, harness libraries, etc
I'm going to add harness libraries first, then mock object libraries 02:35
whiteknight is going to bed now. Goodnight 02:36
02:37 whiteknight left 02:42 lateau_ left 02:55 dduncan joined 02:57 dduncan left 03:02 lateau joined 03:15 kid51_at_dinner left 03:44 particle1 joined 03:49 particle left
cotto ~~ 04:20
mikehh hey cotto, how you doin' 04:39
cotto mikehh, not bad. Did what you pinged me about get addressed? 04:40
mikehh yeah mostly, wanted to check on $Id$ usage in MANIFEST/MANIFEST.SKIP and the generation thereof, but it is fixed now 04:42
cotto ok
mikehh just another removal of SVN leftovers :-} 04:43
anyway need to do some 64 bit testin', some backups and get some sleep maybe 04:46
cotto Doing a release in a sleep-deprived state is highly discouraged.
It's a good recipe for a 3.1.1 release. 04:47
;)
mikehh yeah, but I had access to a server in your part of the world for a client and had to get that done (that's finished now) 04:48
it was only available to me over yje weekend
the
testing on parrot seems fine at the moment 04:49
actually better get some sleep before I really do somthin' stoopid 04:50
cotto 'night
mikehh cu
05:25 cognominal left 05:40 theory left 05:43 theory joined 06:09 rurban_ joined 06:12 rurban left, rurban_ is now known as rurban 06:49 jan left 07:29 clunker3_ left, clunker3 joined 08:17 fperrad joined 08:20 theory left
dukeleto ~~ 08:27
08:36 cotto left 09:02 dip joined 09:09 dngor_ joined 09:13 dngor left
dalek rrot/m0-spec: a2d3648 | dukeleto++ | docs/pdds/draft/pdd32_m0.pod:
Add more M0 ops
 09:20
09:46 lucian joined 09:57 lucian_ joined 10:01 lucian left
dalek rrot/lorito: 1bfda66 | dukeleto++ | / (2 files):
Add a basic test for Lorito dynops, that fails
 10:01
p: 5c5b494 | moritz++ | src/HLL/Compiler.pm:
rewrite parts of command_line in nqp
 10:16
p: 2255560 | bacek++ | src/ops/nqp.ops:
Fix write barrier in updating method_cache
p: f6b7f76 | moritz++ | t/p6regex/01-regex.t:
make p6regex tests runnable again
moritz sorry for the dupe report, that was the "Test Hook" button on github 10:17
10:26 lucian_ left 10:33 nwellnhof joined 10:57 contingencyplan left 11:03 bacek joined 11:34 mikehh left 11:45 mikehh joined
dalek rrot-linear-algebra: 31c445b | Whiteknight++ | README:
remove mention of Kakapo from the README. Replace with Parrot-Test
 12:03
rrot-linear-algebra: a3d770b | Whiteknight++ | t/pmc/pmcmatrix2d.t:
remove debugging statement
 12:08
bacek ~~ 12:16
seen moritz 12:17
clunker3 moritz was last seen on #parrot 1 hour, 59 minutes and 19 seconds ago, saying: sorry for the dupe report, that was the "Test Hook" button on github
aloha moritz was last seen in #parrot 1 hours 59 mins ago saying "sorry for the dupe report, that was the "Test Hook" button on github".
bacek seen moritz_
clunker3 moritz was last seen on #parrot 1 hour, 59 minutes and 24 seconds ago, saying: sorry for the dupe report, that was the "Test Hook" button on github
aloha moritz_ was last seen in #perl6 45 mins 28 seconds ago saying "huh.".
nopaste "bacek" at 192.168.1.3 pasted "More write barriers for rakudo" (122 lines) at nopaste.snit.ch/31882 12:18
bacek msg moritz Can you commit nopaste.snit.ch/31882 please? Preferably after check with jnthn about binder changes.
aloha OK. I'll deliver the message.
bacek !~~
12:24 bluescreen left
bacek msg moritz_ Can you commit nopaste.snit.ch/31882 please? Preferably after check with jnthn about binder changes. 12:29
aloha OK. I'll deliver the message.
12:35 bluescreen joined 12:41 lucian joined 12:45 dngor_ is now known as dngor
dalek TT #2011 created by bacek++: Replace MMD_cache with Hash PMC. 12:46
TT #2011: trac.parrot.org/parrot/ticket/2011
jnthn bacek: fwiw, after the nom branch lands, Rakudo won't use that MMD_cache any more. Also, that MMD_cache could be re-worked to perform rather better. 12:51
The use of a hash is suboptimal
And the typedef was so it could be swapped out for another data structure in the future. 12:52
moritz \\o 12:57
13:11 mtk joined 13:13 whiteknight joined
whiteknight good morning, #parrot 13:21
dalek rrot/whiteknight/imcc_compreg_pmc: 082b020 | Whiteknight++ | compilers/imcc/ (2 files):
several small fixes so we build again
 13:23
rrot/whiteknight/imcc_compreg_pmc: 19f0540 | Whiteknight++ | / (3 files):
add in a new function that gives us a standard way to extract an Interp* from a ParrotInterpreter PMC without having an interp available already. Also, fix one obvious bug I saw in passing, where STRING* != char*
rrot/whiteknight/imcc_compreg_pmc: 0b177f4 | Whiteknight++ | src/pmc/imccompiler.pmc:
IMCCompiler.invoke is basically the same as NCI.invoke. Handle tailcalls here the same way. This fixes test failures that explicitly test for tailcall into IMCC, but there are many others
rrot/whiteknight/imcc_compreg_pmc: a4626d5 | Whiteknight++ | / (5 files):
load_bytecode and load_language now properly handle .pasm file again
rrot/whiteknight/imcc_compreg_pmc: 01d3186 | Whiteknight++ | / (3 files):
use Parrot_ex_throw_* for handling errors in IMCC instead of IMCC_THROW. This fixes some tests for exceptions thrown on syntax errors, but breaks other tests expecting fixed-format error messages. Line numbers aren't being included in the exception messages now, that needs to be fixed
moritz if that branch also cleans up tests that expect fixed error messages, it's a huge win :-) 13:24
whiteknight it may
hard part is figuring out what behavior we like and want to keep
for instance, some of the tests for fixed error messages are actually testing for accurate reporting of line numbers
and we want accurate reporting of line numbers
I *think* I may be able to fix that "syntax error ... somewhere" message 13:25
moritz yes, we want accurate line numbers 13:26
but that should be more like
lineno = ex['line']
13:26 bluescreen left
moritz then test explicitly for line number 13:26
lucian is envious of smalltalk's tracebacks
whiteknight lucian: Parrot doesn't have those yet only because nobody has taken the time to implement them 13:27
our exceptions system really needs some love, but it's not high priority enough right now for anybody to be focusing on it 13:28
lucian i mean the tracebacks in the classical smalltalk IDEs, where it just points you to the code
whiteknight that's very possible. the input code is all Parrot STRINGs that we could keep around and point into if we wanted 13:29
lucian i know, it's just a rare feature nowadays
probably because so many IDEs suck (at least for dynamic languages), and no one uses an IDE anymore 13:30
Coke You must not program in java. ;) 13:31
lucian oh, i do. and java ides are a bit better indeed
Coke wot? checkdepend.t is failing? 13:32
whiteknight for once, I'm pretty confident that I had nothing to do with that failure 13:33
Coke it's src/platform.c 13:38
ah. that file isn't part of the repo anymore.
must be a leftover from my git bisect fest.
13:39 bluescreen joined
Coke whiteknight: did the financials from previous GSOCodes ever get resolved? 13:47
whiteknight I wasn't aware of unresolved financial issues from them
Coke Talk to dukeleto. Pretty sure TPF owes PaFo money (that TPF itself hasn't gotten from google for one reason or other.) 13:48
that would inform my decision whether to recommend we go solo this year or not.
moritz if PaFo goes separately for gsoc, how would handle Perl 6 projects? 13:50
or more specifically, Rakudo projects?
jnthn TPF I'd guess
Coke I imagine that rakudo projects could go either way, but anything else would have to be TPF.
jnthn Unless it was a project with a large Parrot component.
dalek rrot-linear-algebra: d6be1ab | Whiteknight++ | src/pmc/complexmatrix2d.pmc:
remove whitespace
 14:00
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#9539) fulltest) at 3_0_0-937-gcccb01e - Kubuntu 10.10 amd64 (g++-4.5) 14:04
14:09 rurban_ joined 14:10 cognominal joined 14:12 rurban left, rurban_ is now known as rurban 14:14 cognominal left 14:18 cognominal joined
whiteknight I get a "Malformed UTF-8 string" error message in distutils when I try to upload a smolder report 14:19
14:20 cognominal left
whiteknight ah, I am getting a nice backtrace from it after the rethrow 14:20
called from Sub 'slurp' pc 2981 (ext/nqp-rx/src/stage0/nqp-setting.pir:1278) 14:21
Coke "my $count" in t/dynpmc/os.t appears to be unused. 14:22
whiteknight ...which is clearly the wrong line number
nwellnhof whiteknight: nqp seems to default to utf8 encoding for files (like Rakudo). the file encoding should be set to binary somewhere. 14:23
whiteknight for reading in files?
nwellnhof yes
14:25 plobsing left
whiteknight The code for NQP-RX slurp is: 14:28
my $handle := open($filename, :r);
my $contents := $handle.readall;
$handle.close();
$contents;
so somewhere in there we have to set the encoding to binary? 14:29
nwellnhof slurp should accept an encoding argument
whiteknight I suspect it's a problem in distutils. It's calling the wrong "slurp" method 14:30
there's one in the nqp setting lib, and one in distutils, and of course nobody is using proper namespacing 14:31
and, of course, the version in distutils properly handles the encoding 14:32
awesome
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#9545) fulltest) at 3_0_0-937-gcccb01e - Kubuntu 10.10 amd64 (g++-4.5 with --optimize) 14:41
14:44 particle1 is now known as particle
whiteknight Yay! I've submitted the first passing smolder report for PLA since september 14:48
hopefully I can get into a more regular rhythm with it now, now that I'm not relying on Kakapo 14:50
14:53 PacoLinux joined
mikehh whiteknight: still think that is NEWS ;-} 14:54
dalek TT #2012 created by whiteknight++: osutils does not play nice with nqp-setting
TT #2012: trac.parrot.org/parrot/ticket/2012
mikehh which reminds me 14:55
all devs - please update NEWS - I intend to Release in less than 24 hours 14:56
14:57 lucian left 15:03 jsut_ joined 15:05 lucian joined, cotto joined
Coke mikehh: I'm going to add #2013, as it should regain us some platforms for this release at little risk. 15:19
15:20 pyrimidine joined
dalek rrot: a488ccc | Coke++ | src/platform/generic/sysmem.c:
include <sys/param.h> if present.
 15:22
mikehh Coke: was just looking at it
Coke: ok done - testing now
actually waiting for rakudo test to finish 15:23
Coke already hit "push".
mikehh Coke: saw that :-} 15:24
dalek TT #2013 created by doughera++: Build failure with parrot 3.0 on OpenBSD 15:26
TT #2013: trac.parrot.org/parrot/ticket/2013
TT #2013 closed by coke++: Build failure with parrot 3.0 on OpenBSD
TT #2013: trac.parrot.org/parrot/ticket/2013
mikehh Coke: btw I am still getting warning messages in t/src/checkdepend.t 1 in make corevm/make coretest, 3 in make test 15:27
Coke: # couldn't find /usr/include/malloc.h, included from src/gc/malloc.c and two others in make test
Coke: it is there 15:28
Coke those warnings were introduced with the refactor andy just noted in #2013. 15:29
lemme give a quick check.
mikehh: first one is due to using "" quotes on the include (signifying that it's a parrot include) instead of <> (signifying a system include) 15:30
15:30 particle left 15:31 particle joined
Coke not sure on the second one. 15:31
mikehh rakudo (f4ec358) - builds on parrot (3_0_0-937-gcccb01e) - make test, make spectest_smolder[(#9553), roast (051af06)] - Kubuntu 10.10 amd64 (g++-4.5 with --optimize) 15:34
27,634 ok, 0 failed, 610 todo, 1,841 skipped and 0 unexpectedly succeeded
that should have PASS after the ]
15:36 plobsing joined 15:40 particle1 joined 15:45 particle left
cotto_work ~~ 15:45
mikehh Coke: looks like that line in src/gc/malloc.c was included/changed by particle in 2006 sometime 15:47
looking at github and using the blame 15:49
15:49 contingencyplan joined 16:03 plobsing left 16:10 plobsing joined 16:17 Patterner left, Psyche^ joined, Psyche^ is now known as Patterner 16:22 jsut joined 16:23 Andy joined 16:26 lucian left 16:27 jsut_ left
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#9556) fulltest) at 3_0_0-938-ga488ccc - Kubuntu 10.10 amd64 (g++-4.5) 16:27
whiteknight atrodo: ping 16:34
16:37 dipthegeezer joined, dipthegeezer left, dip left, dip joined
atrodo whiteknight> pong 16:39
16:39 theory joined 16:51 dmalcolm joined 17:00 nwellnhof left
whiteknight atrodo: We got your CLA 17:14
I haven't looked it over yet, but we did receive it 17:15
atrodo whiteknight> Excellent
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#9561) fulltest) at 3_0_0-938-ga488ccc - Kubuntu 10.10 amd64 (gcc-4.5 with --optimize)
winxed (806) - make all ok, meke test/test1/test2 PASS - examples/fly/opengl.winxed ok - Kubuntu 10.10 amd64 (gcc-4.5 with --optimize) 17:27
mikehh heading back to Ubuntu 10.10 i386 to do some more tests there - bbiab 17:30
whiteknight atrodo: We'll bring your name up tomorrow at #ps 17:35
atrodo whiteknight++ Thanks
whiteknight no, thank you 17:37
17:42 mikehh left 17:49 mikehh joined 17:54 mtk left 17:56 plobsing left, lucian joined
whiteknight I hate when you click "Undo" in a text editor and it only backs out the last character you typed 17:59
then, to actually undo any kind of substantial change, you need to replay your entire keyboard history
18:01 mtk joined
cotto_work dukeleto: ping 18:06
18:08 plobsing joined
Coke tt.taptinder.org/buildstatus/parrot/master/ -- mj41++ #machine #10 is my desktop. 18:09
ttbot Parrot d4309a5d i386-linux-thread-multi make error tt.taptinder.org/cmdinfo/11149 18:15
18:15 bacek left
Coke ignore that one. 18:15
(I hit ^C)
cotto_work phew 18:16
Coke I'll ask mj41 to delete that one so I rerun it when I startup again. 18:17
dukeleto ~~ 18:54
~~ 18:55
cotto_work: pong
cotto_work dukeleto: see privmsg
dukeleto our NEWS really needs some love, peeps 18:58
dalek rrot: 498cfe8 | dukeleto++ | / (2 files):
[doc] Update NEWS a bit
 19:01
19:03 rdesfo joined 19:13 vmspb joined 19:20 plobsing left 19:21 darbelo joined 19:23 plobsing joined 19:27 bacek joined 19:49 nwellnhof joined
dukeleto GCI finalists announced: google-opensource.blogspot.com/2011...nners.html 20:17
Five of the 14 top students did lots of Parrot tasks. We should be very proud.
cotto_work indeed
dukeleto cotto_work: do you think that is worthy of putting in NEWS? 20:18
cotto_work: it is looking pretty sad now
cotto_work dukeleto: yeah. I'm planning on trolling #ps for more items tonight
Putting gci in there is a stretch though.
If little happened, little happened. 20:19
dukeleto hokey dokey
20:19 bacek left
whiteknight I haven't committed to master at all, with the imcc branch and other work on my plate 20:19
20:22 lateau left
dukeleto lots of development has happened outside of master in the last month 20:27
so that is why NEWS looks skimpy
cotto_work I hate to say it, but we should remove the gc entries. 20:33
whiteknight yeah, if it's not merged it shouldn't be in there
20:39 ambs joined
dalek rrot: 7db4a1f | cotto++ | NEWS:
remove gc NEWS entries
 20:43
Coke I would put GCI in the main body of the release announcement. not NEWS. 20:44
I would also like to get rolling on the "this code contains this much change over this many files by these authors" portion of the message.
(to mimic p5's recent, very nice announcements.) 20:45
20:47 lucian_ joined
dukeleto cotto_work: the git summary command from git-extras can help with that 20:48
Coke: ^^^
github.com/visionmedia/git-extras
cotto_work That was handy for the last release announcement. 20:49
20:51 ttbot left, lucian left, ttbot joined
Coke can you add it to the release prep docs? 20:51
20:51 plobsing left
cotto_work I'm surprised that only 17 people are listed. 20:52
nwellnhof what's up with dalek? 20:53
dalek rrot/nwellnhof/unicode_dynpmcs: f8adb9b | nwellnhof++ | src/ (2 files):
Fix Unicode stat
rrot/nwellnhof/unicode_dynpmcs: ab8b519 | nwellnhof++ | / (2 files):
Add first tests for Unicode filenames
nwellnhof he's slower than usual
bacek_at_work ~~ 20:57
cotto, put GC entries back into NEWS. It was about dynamic threshold :) 20:58
cotto_work wfm 20:59
dalek rrot: 8601021 | cotto++ | NEWS:
Revert "remove gc NEWS entries"

This reverts commit 7db4a1f04ea21c24c3a9882e50125cc4d989fb64.
cotto_work I love how easy that is.
bacek_at_work cotto, why do you use pointer_hash in profiling runcore? Can it be just int_hash? 21:02
21:05 benabik joined 21:06 benabik left
mj41 New jobs and new machine (Coke++) added to tt.taptinder.org/ 21:15
bacek_at_work mj41++ # Hooray!
mj41, can you switch generational_gc branch to gen_gc2? 21:19
21:20 whiteknight left 21:21 contingencyplan left
mj41 bacek_at_work: yes, will look at it tomorrow ... it should be easy as creating new row to tt.taptinder.org/cwm/wconf_rref but there is probably a bug somewhere in github.com/mj41/TapTinder/blob/mas...nt.pm#L452 21:25
... so developers are welcome, ping me on irc://irc.freenode.net/#taptinder ... good night 21:28
bacek_at_work night, mj41 21:29
dalek rrot/gen_gc2: 5e0c8a7 | bacek++ | src/pmc/namespace.pmc:
Add write barrier to Namespace PMC when we poke inside hash
 21:30
rrot/gen_gc2: d1b5f09 | bacek++ | src/pmc/class.pmc:
Add write barrier to Class PMC when we poke inside cache
rrot/gen_gc2: df5d0a5 | bacek++ | include/parrot/context.h:
Fix preprocessor condition
rrot/gen_gc2: 3e7678b | bacek++ | lib/Parrot/Pmc2c/PMC.pm:
Simplify checks for creating WB override for VTABLE
rrot/gen_gc2: e4adcb5 | bacek++ | src/gc/gc_gms.c:
Do GMS sanity check only when DETAILED_GC_DEBUG set.

This speed-up debug builds a lot.
21:36 plobsing joined 21:37 bluescreen left 21:57 ambs left 22:00 contempt joined 22:02 darbelo left
cotto_work bacek_at_work: I was thinking of the pc as a pointer. If there are performance implications, that could could just as easily treat it as an INTVAL. 22:05
22:09 rurban_ joined
plobsing cotto_work: what happens if we become able to collect/copy/compact code segments? then a pointer into a code segment is no longer a valid PC representation. if we have multiple code segments, neither is an INTVAL a valid representation. 22:10
PC needs to be something like struct { PMC* segment; UINTVAL offset; } 22:11
but visible to GC
22:12 rurban left, rurban_ is now known as rurban
cotto_work I could use a hash of the segment's name, but that'll get expensive pretty quickly. 22:16
22:29 fperrad left, Andy left
dalek rrot/gen_gc2: d2c99cc | bacek++ | src/pmc/hash.pmc:
Put write barrier into Hash.clone. This one wasn't easiest to catch.
 22:48
tadzik bacek_at_work: all those fresh write barriers are bugfixes? 22:50
bacek_at_work tadzik, yes 22:51
22:51 cosimo_ joined
cotto_work bacek_at_work: I thought all tests were passing. 23:00
23:15 whiteknight joined 23:17 lucian_ left
Coke ... if it's that hard for the person who wrote the GC to catch these, I despair of our users knowing when to use them. 23:30
23:32 toei joined
plobsing if NCI becomes useful (yes, I know my attempts thus far have fallen far short), hopefully very few users will need to be aware of these (only extenders should need to worry about the barriers) 23:33
mikehh Coke: re GCI news, we havew a section in NEWS for Community - shouldn't it go there 23:34
23:34 kid51 joined
whiteknight Coke: write barriers are notoriously difficult. This isn't a Parrot-only thing 23:41
dalek p: 1307625 | jonathan++ | src/pmc/dispatchersub.pmc:
Fix up dispatcher sub cloning.
p: 86697e2 | jonathan++ | src/metamodel/how/NQPMu.pm:
Mark the Str proto as also being the parrot_vtable handler for get_string. This means say(SomeTypeObj) works.
23:52 cotto left
bacek_at_work cotto_work, yes. Unfortunately our test suite doesn't provide enough stress for GC. 23:52
and many-many different cases can't be reliably tested. 23:53
E.g. "trigger GC during cloning of hashes"...
seen masak
clunker3 masak was last seen on #parrot 1 year, 10 months, 22 days, 13 hours, 3 minutes and 31 seconds ago, saying: oh well. it's early in bot history still.
aloha masak was last seen in #perl6 2 mins 8 seconds ago saying "anyway, it's as if someone would say "Will have to check it out, haven't heard of Chess before. Looks a lot like Crazyhouse."".
plobsing bacek_at_work: would it be a good idea to run the same program repeatedly with GC somehow configured to run collections at differing times to try to detect such edge cases? 23:56
Tene there was chromatic's gc-debug runcore oslt, yeah? 23:57
bacek_at_work plobsing, yes. Something like this will be helpful.
Tene run gc at every possible time
bacek_at_work Tene, it doesn't cover all possible cases.
plobsing gc runcore runs GC between all ops. better to run gc between all allocations. even then it won't catch all cases.
Tene ah, yeah, it was just between all ops. 23:58
plobsing to cover all cases you have to run the program 2^N times where N is the number of allocations 23:59
23:59 contempt left
plobsing randomizing is a good enough approximation I think 23:59
cotto_work We're going to need a bigger farm.

Powered by Cro and the Raku® Programming Language.

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

Thank you!