www.parrot.org | Prepare for 1.6.0: Improve test coverage for NameSpace and Continuation PMCs / Stability! / Port tests to PIR / Performance! / No more big branch merging until after 1.6.0
Set by moderator on 8 September 2009.
00:00 bacek joined
Whiteknight parrot builds with clang if we configure with --without-gmp 00:01
kid51 Whiteknight: What exactly do you mean by "builds with clang"? 00:03
japhb Is there any way to initialize array or hash in NQP, without resorting to PIR? 00:05
Whiteknight kid51: as in, I can create a parrot executable with the clang C compiler 00:14
japhb: no, at least not last time I looked at NQP 00:15
kid51 Whiteknight: I've been reading thru trac.parrot.org/parrot/wiki/JITRewrite 00:18
I think it would benefit from more definitions of terms and acronyms at its outset 00:19
dalek TT #1006 created by whiteknight++: Does not build on clang with GMP 00:20
Whiteknight kid51: yes, definitely
and there are some clarifications that we need to add now from Allison's email 00:21
dalek rrot: r41255 | whiteknight++ | trunk/PLATFORMS:
update PLATFORMS to reflect the current status of the clang build. See TT #1006
 00:23
kid51 JIT, machine code, call frame, NCI, call signature, op/ops would all benefit from definition ... 00:24
... even if defined elsewhere in Parrot
Whiteknight yes, I take a lot of those terms for granted 00:25
kid51: I'm glad to hear somebody is reading it! 00:31
00:34 Andy joined
kid51 Whiteknight: Here are two terms from Allison's post to list that would also benefit from definition: runcore; dynamic compilation strategy 00:40
Whiteknight okay
ash_ Whiteknight: any knowledge on the status of trac.parrot.org/parrot/ticket/757? 00:55
Whiteknight ash_: no, I don't even know why I was the owner 00:56
ash_ oh, hehe, its kind of old, but its still relavent, i know Tene in perl6 was trying to implement some threading stuff in rakudo and ran into that problem 00:57
Tene Whiteknight: looks like you set yourself to the owner. 00:59
ash_ well, um... should i ask in the mailing list maybe? incase someone has done anything for it? just don't want it to be forgottne
Whiteknight Tene: yeah, and I have no idea why I did that
Tene Whiteknight: I think that we were chatting about it shortly after I created the ticket, and you volunteered to try to fix it.
Whiteknight Maybe, but i must have forgotten about it shortly thereafter 01:00
Tene Whiteknight: I also read and enjoy your blog posts. Thank you for them.
Whiteknight Tene: Thanks! I try not to be boring with it 01:01
ash_ Have anyone of you guys seen grand central dispatch? 01:07
Whiteknight heard of it, but haven't seen it 01:27
ash_ apple open sourced it under the apache license 01:29
it requires a C compiler with block support (llvm-gcc, clang or apple gcc) but its pretty cool for a way to thread without having to thread in C 01:30
kid51 Running make fulltest on Darwin/PPC, the only two test failures I got were those I reported last night in trac.parrot.org/parrot/ticket/1004 from fulltest on Linux/i386. 01:37
mikehh ping
Whiteknight Can you post in TT #1006 output of failed builds with clang/gmp -- so that we can see what you mean by "several of the GMP symbols are defined more then once"? 01:47
01:48 rhr joined
Whiteknight kid51: I'm about to sign off, can you ask in the ticket so I don't forget? 01:49
kid51 k
Whiteknight thanks. Goodnight
01:49 sri_ joined 02:20 kyle_l5l joined 02:35 janus joined 02:57 nathanmccauley joined
jdv79 Coke: around? 02:59
03:23 jhelwig joined
dalek rrot: r41256 | NotFound++ | trunk/src/pmc/scalar.pmc:
[pmc] assign Undef for subclasses of Undef in Scalar.assign_pmc. Rakudo test S12-introspection/walk.rakudo pass with this change. This function needs more review, though
 03:29
03:44 mokurai joined 04:21 dukeleto joined 04:23 nathanmccauley joined 04:27 theory joined 04:30 Andy joined 04:38 Zak joined, patspam joined 05:41 nathanmccauley joined 05:44 flh joined 06:07 uniejo joined 06:46 mj41 joined 06:50 barney joined
dalek kudo: e4c7a05 | moritz++ | build/PARROT_REVISION:
bump PARROT_REVISION to get fixes in Scalar.assign_pmc
 07:04
07:19 chromatic joined
dalek rrot: r41257 | mikehh++ | trunk/src/pmc/scalar.pmc:
codetest failure - tab in leading whitespace
 07:41
japhb use.perl.org/~geoffrey/journal/39622 07:52
Because why get the sleep I need, when I can write about hacking?
mikehh All tests PASS (pre/post-config, smoke, nqp_test, fulltest) at r41257 - Ubuntu 9.04 amd64 08:11
08:27 einstein joined
mikehh msg kid51: certainly - I am just reporting results for amd64 - when I test on i386 I report results for that platform 08:33
purl Message for kid51 stored.
08:44 KatrinaTheLamia joined
mikehh partcl r731 builds on parrot r41257 - make test PASS - Ubuntu 9.04 amd64 08:47
decnum_dynpmcs r181 builds on parrot r41257 - make test PASS - Ubuntu 9.04 amd64 08:50
08:51 patspam joined
mikehh rakudo (e4c7a05) builds on parrot r41257 - make test / make spectest (up to 28238) PASS - Ubuntu 9.04 amd64 09:01
rakudo - t/spec/S03-operators/arith.rakudo - TODO passed: 120, 131-132
rakudo - no Segmentation faults reported in make spectest
cardinal builds on parrot r41257 - rake test:all - reports same 3 failures - Ubuntu 9.04 amd64 09:04
ok time to do some i386 tests - bbiab 09:05
09:10 bacek joined 09:36 mikehh joined
mikehh ok back on i386 09:36
09:44 gaz joined 09:54 mokurai left 09:59 payload joined
dalek rrot: r41258 | bacek++ | trunk/t (2 files):
[cage][t] Mark tests failing under JIT with todo. TT#983
 10:00
TT #983 closed by bacek++: JIT code incorrectly access registers 10:08
mikehh bacek: I am not getting any failures in make testj at r41257 on i386 10:09
bacek mikehh: erm... You should have them... Did you run make testj? 10:10
dalek rrot: r41259 | bacek++ | trunk (2 files):
[core] Implement Undef.logical_not. TT#816.
mikehh bacek: dat is wat I sed 10:11
bacek mikehh: it's... suspicious.
bacek@icering:~/src/parrot$ perl t/harness -j t/pmc/sub.t
t/pmc/sub....18/68
not ok 36 - caller introspection via interp # TODO broken with JIT (TT #983)
(This one of tests that I todoed) 10:12
(It should be article somewhere in previous phrase :)
mikehh let me finish testing at r41257 and I'll see what happens 10:13
bacek mikehh: ok. 10:14
10:20 MoC joined
mikehh All tests PASS (pre/post-config, smoke, nqp_test, fulltest) at r41257 - Ubuntu 9.04 i386 10:21
however in testg - t/op/exit.t - TODO passed: 6 10:22
partcl r731 builds on parrot r41257 - make test PASS - Ubuntu 9.04 i386 10:29
10:32 donaldh joined
mikehh decnum-dynpmcs r181 builds on parrot r41257 - make test PASS - Ubuntu 9.04 i386 10:34
10:38 kjeldahl joined
mikehh rakudo (e4c7a05) builds on parrot r41257 - make test / make spectest (up to 28238) PASS - Ubuntu 9.04 i386 10:46
rakudo - t/spec/S03-operators/arith.rakudo - TODO passed: 131
rakudo - make spectest reports 126 Non-zero wait status: 11 (Segfault after passing tests)
rakudo - Note t/spec/S03-operators/arith.rakudo TODO passes 120, 131-132 and NO Segmentation faults reported on amd64 at sa,e revision
same
bacek seen jrtayloriv
purl jrtayloriv was last seen on purl 11 hours, 15 minutes and 27 seconds ago, saying: <private message>
moritz seen on purl. Ts ts. 10:49
bacek :)
mikehh can't get cardinal to build at the moment on i386 - will check later - it builds on amd64 10:52
10:52 kid51 joined
mikehh bacek: re-testin' now 10:52
bacek moritz: ping 11:07
mikehh: any luck? :) 11:08
moritz bacek: pong
bacek moritz: anything particularly annoying to Rakudo from Parrot? I'm just looking for some "small" fixes that I can try to do. 11:10
moritz bacek: don't know if that's a parrot or a rakudo issue, but some errors don't produce backtraces 11:11
can this be changed from within parrot?
bacek moritz: any examples?
moritz a typical example is the "Method 'foobar' not found for invocant of class 'Int'ā¤ 11:12
bacek ouch... 11:13
mikehh bacek: the TODO's PASS in testj for me 11:16
bacek mikehh: on i386?
moritz: does Rakudo show backtrace on any circumstances? 11:19
moritz bacek: yes 11:20
11:21 donaldh joined
moritz for example die('foo'), but also lots of other errors 11:21
mikehh bacek: yes on Ubuntu 9.04 i386 11:23
testj don't run on amd64
bacek mikehh: oh... I can mark those tests with SKIP to avoid false positives. But they are definitely failing on my Debian Lenny/i386. 11:24
mikehh let me do a clean checkout and check again 11:25
bacek moritz: sigh... It's PCC related. Major difference between Do_something_from_c and Do_something_from_op.
moritz :) 11:30
s/)/(/
so rakudo has to catch those exceptions and re-throw them with a stack trace?
bacek it will not help afaiu. backtrace already lost when exception is thrown. 11:33
moritz ouch.
11:39 fperrad joined 11:43 tetragon joined
mikehh bacek: BTW g++ doesn't like src/pmc/undef.pmc 11:44
bacek moritz: btw, get_foo_same VTABLEs can be removed src/pmc/objectref.pmc now
mikehh: oh. What it complains about? 11:45
mikehh - ./src/pmc/undef.c: In function ā€˜VTABLE* Parrot_Undef_update_vtable(VTABLE*)ā€™: 11:46
- ./src/pmc/undef.c:274: error: invalid conversion from ā€˜INTVAL (*)(parrot_interp_t*, PMC*)ā€™ to ā€˜PMC* (*)(parrot_interp_t*, PMC*, PMC*)ā€™
- ./src/pmc/undef.c: In function ā€˜VTABLE* Parrot_Undef_ro_update_vtable(VTABLE*)ā€™:
- ./src/pmc/undef.c:295: error: invalid conversion from ā€˜INTVAL (*)(parrot_interp_t*, PMC*)ā€™ to ā€˜PMC* (*)(parrot_interp_t*, PMC*, PMC*)ā€™
dalek TT #1007 created by fperrad++: load_language creates wrong search paths 11:47
bacek mikehh++ #
mikehh: hmm... I didn't touched Undef.is_equal. 11:50
11:51 PacoLinux joined
bacek mikehh: r41260 12:00
dalek rrot: r41260 | bacek++ | trunk/src/pmc/undef.pmc:
[cage] Fix g++ failure reported by mikehh++. pmc2c-- for not reporting it.
 12:02
12:18 whiteknight joined
whiteknight good morning #parrot 12:20
mikehh hiya whiteknight 12:23
whiteknight hello mikehh 12:24
mikehh whiteknight: been keeping up with your blog - very informative
whiteknight thanks! 12:25
parrot is becoming a pretty big project with a big ecosystem, and not everybody always knows what's going on with the core development 12:26
12:29 payload joined
mikehh whiteknight: in terms of JIT as I understood it - it only should "compile" when a function or op is called the first time as opposed to your ahead-of-time 12:29
whiteknight AOT is technically a different thing from JIT 12:30
I lump them together because ideally a good JIT compiler will be able to do AOT stuff as well
mikehh sure but that is what we seem to have been doing
whiteknight sortof, yes
mikehh All tests PASS (pre/post-config, smoke, nqp_test, fulltest) at r41260 - Ubuntu 9.04 i386 12:42
however t/op/exit.t TODO passed: 6 in testf and testg and t/op/calling.t TODO passed: 52 and t/pmc/sub.t TODO passed: 36 in testj
that is is with a clean checkout on Ubuntu 9.04 i386 12:43
12:45 quek joined
szbalint it's always a pleasure to read whiteknight's posts :) 13:04
whiteknight++
whiteknight thank you!
13:09 donaldh left
mikehh rakudo (e4c7a05) builds on parrot r41260 - make test / make spectest (up to 28238) PASS - Ubuntu 9.04 i386 13:11
rakudo - t/spec/S03-operators/arith.rakudo - TODO passed: 131
rakudo - make spectest reports 91 Non-zero wait status: 11 (Segfault after passing tests)
rakudo - re-ran make spectest and got 115 Non-zero wait status: 11 (Segfault after passing tests) 13:24
the Segmentation faults in rakudo seem to be preety random 13:28
in the first run at r41260 - 91 counted t/spec/S03-operators/arith.rakudo was one of them - in the second run (115 counted) it wasn't 13:29
whiteknight mikehh: on average, how many tests are failing like that each run? 13:39
and is it only specific tests that fail, or is it a random smattering? 13:48
moritz back in the days (last week) when the same occured on amd64 it was random
whiteknight moritz: has there been any progress on this then, or is it still as bad as it was? 13:49
whiteknight hasn't been around much this weekend, and doesn't know if anything got fixed 13:50
moritz whiteknight: on amd64 I don't observe that bahviour anymore, but it's been reported on i386 and macos x (see TT #998) 13:52
13:52 Andy joined
whiteknight the problem is, I think, that a good number of our developers are on amd64 right now as their primary platforms 13:53
mikehh whiteknight - it does not fail for me on amd64 but on i386 the last 3 make spectests had 121, 91 and 115 (today) some repeats but a lot different 13:56
whiteknight urg 13:57
I've looked over a few of the big changesets, and don't see anything that should have made this such an issue 13:58
mikehh looking back I got 112 at r21220 but only 2 at r41221 again on i386 14:02
r41220 14:03
bbl
whiteknight well that's weird as hell. What a small change! 14:04
particle hi all 14:30
talking about segfaults?
14:35 jsut|work joined 14:41 theory joined 14:45 Psyche^ joined 14:58 allison joined 15:00 quek left
NotFound Some rakudo boy wanting to try a workaround agains segfaults? 15:04
moritz raises hand 15:05
NotFound Or girl ;)
No discriminations!
Insert 'exit 0' before line 378 in perl6.pir
moritz but don't you discriminate discriminations? 15:06
NotFound Oh, no discriminants may also try.
moritz NotFound: thanks for the workaround. I have to run now, but if we don't have a proper fix in time of the release I'm sure we'll use it. 15:09
NotFound moritz: I think that the remanining problems are inner loop/exception handler related, so I don't hope to have them fixed shortly. 15:13
15:15 cognominal joined
whiteknight NotFound: you saw those emails I sent to Allison? 15:20
einstein continuation.pmc marks a non pobj obj, Parrot_gc_mark_PObj_alive(interp, (PObj *)cc->seg);
whiteknight einstein: cc->seg isn't a PMC?
einstein no it is not
purl i already had it that way, einstein.
15:20 donaldh joined
whiteknight it? 15:21
purl or a matter ot tuits
whiteknight it?
purl or a matter ot tuits
whiteknight no it?
purl i think no it is not
NotFound whiteknight: yes, but lacked the time to elaborate some comment
whiteknight okay, just making sure you saw it
I think that will be the best permanent solution to this problem 15:22
of course, it will be hard to get people like Rakudo to completely switch to using the new handler type
NotFound whiteknight: but I don't think that a specific type of sub is the better solution.
einstein shall i make ticket for that one?
whiteknight NotFound: Why not? I think it's a pretty damn good solution
einstein: if it's causing a problem, es 15:23
yes
NotFound whiteknight: I think that is the solution is to handle the exception in a sub, just use a plain sub.
einstein it is causing problem in my build, i changed some things my own build, i can simply remove the line, but i don't know we are stuck with a memory leak 15:24
NotFound Or better, use anythink that has VTABLE_invoke
whiteknight NotFound: Continuations have invoke, ExceptionHandlers are continuations, and THAT is the problem now
it can't just be something that invokes, it has to be something that can return too 15:25
because we need to be able to resume
NotFound whiteknight: right, then the solution is: ExceptionHandlers are not continuations and not sub, but contains one.
whiteknight That's a semantic issue
NotFound And create a continuation to run it, just like any other sub. 15:26
whiteknight these are just details, the important parts are the same
it's fun thinking about exception handlers being coroutines :)
NotFound whiteknight: I don't think so. Tell me for example how so set a 'can_handle' method in a Sub 15:27
whiteknight NotFound: use a property
.sub 'myhandler' :can_handle("stuff")
NotFound whiteknight: More pir sugar?
whiteknight or, a new push_eh
push_eh 'my_sub', 'can_handle...' 15:28
lots of ways to do it
NotFound whiteknight: lots of way to have subtle bugs
whiteknight the best kind of bugs!
NotFound We'll have enough without the need to open highways for them 15:29
einstein I will do a test to see whether there are more non pmc injected into the mark Parrot_gc_mark_PObj_alive 15:30
NotFound einstein: good catch! 15:31
einstein :)
15:32 donaldh left 15:36 darbelo joined
NotFound einstein: Definitely open a ticket for it 15:37
cognominal whiteknight, is this a feature or a bug, the ad for the scientology in wknight8111.blogspot.com/2009/09/jit-plan.html
JIT scientology? 15:38
whiteknight cognominal: I don't see an add for scientology there
einstein yes i will do, first i will a test to see whether there are more non pobj entered into Parrot_gc_mark_PObj_alive 15:39
NotFound einstein: good idea
purl NotFound: Good Idea: Giving a small child a balloon. Bad Idea: Giving a small child a bunch of balloons.
cognominal whiteknight, I suppose I have been selected by some AI for that treat 15:40
15:40 MoC joined
whiteknight I really should remove those advertisements entirely. Since I've had them I've managed to earn a whopping $2.09 from them 15:40
Finally! I don't have to worry about money anymore! 15:41
NotFound whiteknight: Did you won some 1.000.000 pounds in some Britanic Lotto?
whiteknight NotFound: I wonder if it would be better to have IMCC automatically add an "exit 0" to the end of the :main function? 15:45
Coke: Ping
cognominal how these ad guys know I feel unworthy and insecure to promote me the cure : scientology :)
whiteknight cognominal: I removed all adds from m blog now 15:46
cognominal :) 15:47
whiteknight I refuse to be advertising scientology
NotFound whiteknight: too specific, we can have problems of the same type with tailcall from main, :init or :load subs... 15:48
whiteknight well, add it to all those types of functions
we're looking for "quick and dirty solution"
and the dirtier, the better :) 15:49
NotFound I'm thinking about creating a special return continuation that inmediatley stops the current runloop.
whiteknight how would you do that? 15:50
NotFound whiteknight: just an idea for anow
whiteknight Once we fix exception handling, we shouldn't need any other fixes 15:51
dalek rrot: r41261 | NotFound++ | trunk/src/call (2 files):
[core] handle some corner cases in pcc retval settings, avoid some segfaults
 15:52
NotFound whiteknight: talking about "quick", I can read and modify continuation code a lot faster than imcc one 15:55
whiteknight true
irclogs? 15:56
purl irclogs is irclog.perlgeek.de/parrot/today or see also: infrared clogs
NotFound BTW nci mysqltest works again :)
davidfetter w00t 15:58
16:03 payload joined
Tene NotFound: did you change parrot or the nci test? 16:05
NotFound: and does just the .pir one work, or the Perl 6 version too?
NotFound Tene: changed parrot, and disabled last part of the example that was not well tested 16:06
Tene nods. 16:07
NotFound Don't checked yet the perl version
Tene NotFound: I'll try again to get it working from Rakudo, then.
NotFound Parrot problem was in passing a NUL pointer, it passed a PMCNULL instead of NULL 16:08
particle notfound: re r41261: you should be checking PMC_IS_NULL(retval), don't assume PMCNULL == null everywhere 16:15
everywhere meaning on every platform.
16:15 nathanmccauley joined
NotFound particle: if it already is PMCNULL, no need to change 16:16
particle return retval ? retval : PMCNULL;
your logic is "if retval is null, return PMCNULL"
NotFound particle: if retval is PMCNULL , both branches give PMCNULL 16:17
particle ah, yes, ok
so, you mean to change null to PMCNULL there.
NotFound Yes, the point is to avoid a plain NULL to be returned, it can find a way to a PMC register. 16:18
particle yes, notfound++
16:29 mokurai joined
japhb cognominal, AdBlock Plus. For great sanity. :-) 16:31
cognominal japhb, I used it on firefox, can I use it from safari? 16:34
japhb cognominal, I wish. Lack of ABP is probably the single biggest reason I'm not running Chrome half the time.
(Second reason would be that I'm addicted to the Firebug, YSlow, and Web Developer add-ons for Firefox.) 16:35
16:56 parrot-poke joined 17:06 joeri joined
whiteknight (trolling the mailing list)-- 17:10
17:12 kyle_l5l joined
Tene whiteknight: One point about jit that I see coming up in conversations... 17:13
whiteknight: The reason we want to do better isn't "other people are faster" but "We can do so much better."
whiteknight yes, very much so 17:15
Parrot needs to do a lot of work before it can catch up with other VMs in terms of performance
but other VMs tend to be focused on individual languages
17:17 flh joined
japhb How do I get my Parrot Plumage u.p.o journal entries added to Planet Parrot? 17:21
whiteknight japhb: I'm supposed to be able to do that but I forgot how 17:22
Coke can though
japhb :-)
Coke: see above
NotFound The reason we want to do better is that the current one is a mess. 17:26
whiteknight that's only part of it. If it were good but messy we would create a branch and clean it 17:33
but it's wrong on so many levels: It's bad design AND bad implementation 17:34
szbalint whiteknight++ # blogging machine 17:35
whiteknight szbalint: No no, I've been lazy this month
szbalint if this is lazy, I wonder what's it like when you aren't lazy for a month... 17:37
Seriously, half the posts at planet parrot are yours :)
whiteknight I had 28 posts in May. Only 9 posts so far in september
szbalint I guess I think in rolling months though 17:38
17:39 kjeldahl joined 17:42 cotto_work joined
cotto_work ohai 17:42
i haz a job 17:48
17:51 chromatic joined
davidfetter cotto_work, are congratulations in order? 17:52
NotFound How is the perl6 syntax for using parrot modules now? 17:54
cotto_work davidfetter: I'm at work now (though I have yet to get a working non-live CD computer) 17:58
NotFound use Mysql:from<parrot>; ? 18:10
dalek TT #1008 created by jessevdam++: continuation mark function marks non pobj PackFile_ByteCode *seg 18:13
18:14 ruoso joined
dalek rrot: r41262 | NotFound++ | trunk/src/pmc/continuation.pmc:
[pmc] Delete wrong mark of packfile segment wich is not PObj , TT #1008 einstein++
 18:34
einstein ok thanks 18:39
NotFound Thanks to you :) 18:40
einstein: I think this fail can be the culprit of the packfile freeing problems we were worked around some days ago 18:49
chromatic Agreed.
einstein ok that is a nice extra catch for that one :) 18:51
dalek rrot: r41263 | jkeenan++ | trunk/src/call/pcc.c:
Enforce coding standards: no tabs in leading whitespace.
 18:53
18:57 bacek joined, cotto_work left 18:59 zostay joined
kyle_l5l why would an object disappear from the next_for_GC list for an arena? 19:03
whiteknight kyle_151: what do you mean?
kyle_l5l whiteknight: that I don't understand all that is going on with the gc? 19:06
whiteknight the next_for_GC pointers should form a linked list of objects that need to be marked
and they should stay on that list until the GC sweep phase
kyle_l5l and after I mark them, does this list go away or need to be re-created? 19:07
whiteknight it gets recreated for every GC mark phase, and cleared every GC sweep phase 19:08
or, that's what should happen
einstein I also looked at the code of gc and that piece was also not clear to me, i also could not find any documentation which the original intention of this code 19:11
whiteknight that's a good observation. It does require more documentation
feel free to send in a patch, or I will try to do it tonight.
chromatic I'm not sure it's useful. 19:12
The behavior, I mean. Patches are welcome. 19:13
whiteknight okay, yes
so what would be the behavior instead, marking the flags and all children immediately when Parrot_gc_mark_PObj_live() is called?
NotFound Looking at the code, it seems there was a provision for order of destruction, or something like that. 19:14
whiteknight yeah, that's what I think they were going for
einstein I intended for some kind of prioritization of marking, but in that does not seem function as is should be, because the first phase does already everthing
whiteknight right 19:15
einstein It was intended for some kind of prioritization of marking, but in that does not seem function as is should be, because the first phase does already everthin
woops
chromatic There are probably better ways to manage GC-in-pieces.
whiteknight "probably"?
purl "probably" is correct
einstein it belongs to some kind of implementation of gc-in-pieces, but since the original used algoritm is unknow it migth becomes better to reimplement gc-in-pieces 19:18
or you have to find out what the original gc-in-pieces algoritm is 19:19
chromatic Right, it's tied too closely to mark and sweep. 19:20
whiteknight I think we could have next_for_GC ripped out entirely within a week if we needed to
darbelo whiteknight: Didn't you have a patch for that already?
whiteknight darbelo: my patch was flawed.
At the time I didn't realize exactly what next_for_GC was being used for. Now I think I know better 19:21
darbelo Hmm. mis-used for: "XXX this basically invalidates the high-priority marking of PMCs by putting all PMCs onto the front of the list." 19:23
einstein pmc_free.c also uses the next_for_GC field 19:26
whiteknight pmc_free.c?
einstein pmc_freeze.c sorry 19:27
whiteknight it shouldn't use it anymore, I thought darbelo removed them all
darbelo I thought that too.
einstein is that a short time ago?
whiteknight relatively
einstein ok that may explain 19:28
NotFound This month has been long :)
whiteknight yes it has!!!
darbelo Hmm, but I apparently left a reference in the embedded pod. 19:29
19:35 baest joined
dalek rrot: r41264 | NotFound++ | trunk/examples/nci/xlibtest.p6:
[examples] update xlibtest.p6 to new use syntax
 19:36
darbelo make codetest is sloooooooow 19:39
19:40 baest_ joined
dalek rrot: r41265 | darbelo++ | trunk/src/pmc_freeze.c:
Remove references to next_for_GC from pmc_freeze.c
 19:43
19:45 baest joined 20:00 joeri joined 20:07 kyle_l5l left 20:08 hercynium joined 20:12 ash_ joined 20:16 joeri left 20:38 payload joined 20:53 kjeldahl joined 21:22 kyle_l5l joined 21:23 cotto_work joined
darbelo Is anyone else seeing segfaults in test 18 of t/examples/shootout.t ? 21:25
dalek kudo: a9ff309 | pmichaud++ | docs/spectest-progress.csv:
spectest-progress.csv update: 436 files, 15500 (71.5% of 21679) pass, 0 fail
 21:29
chromatic Finally broke 70%. 21:30
21:35 Whiteknight joined, bacek joined
pmichaud and 15K tests :) 21:35
nopaste "darbelo" at 200.49.154.172 pasted "Segfault in continuation PMC set_pointer()" (25 lines) at nopaste.snit.ch/17960 21:37
chromatic The spectest has grown by 19%, but Rakudo passes 26% more spectests than in the August release.
darbelo, is pos invalid in that backtrace?
darbelo chromatic: yep. 21:38
cotto_work nobody's managed to fix those darwin failures that have been occuring since the pluggable_runcore merge, have they? 21:40
smolder?
purl smolder is sourceforge.net/projects/smolder or web-based smoke test aggregator used by developers and testers to upload (automated or manually) and view smoke/regression tests using the Test Anything Protocol (TAP). or smolder.plusthree.com/app/public_pr..._reports/8
chromatic Hm, that means the problem is in the exit op, specifically line 889 or 890 of src/ops/core.ops. 21:41
cotto_work hm. either the test got todo'd or it's passing now. 21:42
NotFound cotto_work: some of the recent argument fixes can have solved tahat. 21:43
chromatic darbelo, what's your example code?
darbelo test 18 of t/examples/shootout.t 21:44
NotFound They can also have been related with the marking of segment.
darbelo But it passes about half the time. 21:45
examples/shootout/spectralnorm.pir is the actual pir 21:47
Whiteknight how are we looking test-wise for tomorrow? I've been seeing a lot of clean smolder reports come through
cotto_work NotFound: that makes me happy.
mikehh All tests PASS (pre/post-config, smoke, nqp_test, fulltest) at r41265 - Ubuntu 9.04 i386
cotto_work nm. I read s/can have/have/
mikehh however t/op/exit.t TODO passed: 6 in testf and testg, and, t/op/calling.t TODO passed: 52 and t/pmc/sub.t TODO passed: 36, in testj 21:48
chromatic Those may be because I fixed some line numbers. 21:49
21:51 bacek joined
bacek Good morning 21:54
purl And good moroning to you, bacek.
duk3leto 'ello 21:55
mikehh hiya bacek
partcl r731 builds on parrot r41265 - make test PASS - Ubuntu 9.04 i386 21:56
decnum-dynpmcs r181 builds on parrot r41265 - make test PASS - Ubuntu 9.04 i386
cotto_work hi bacek 21:57
bacek cotto_work: congratulations :) 21:58
cotto_work I find myself in a twisty maze of offices, all alike. There are posters for SQL Server on the wall. 22:00
mikehh rakudo (a9ff309) builds on parrot r41265 - make test / make spectest (up to 28239) PASS - Ubuntu 9.04 i386 22:03
rakudo - t/spec/S03-operators/arith.rakudo - TODO passed: 131
rakudo - only 2 tests - t/spec/S12-attributes/class.rakudo and t/spec/S14-roles/basic.rakudo - Non-zero wait status: 11
had those before - obviously a different problem
22:05 fperrad left
mikehh ok had enough here - back to amd64 - bbl 22:05
dalek TT #1009 created by darbelo++: examples/shootout/spectralnorm.pir shows a segfault in the continuation ... 22:22
22:22 mikehh joined 22:24 bacek joined
bacek cotto_work: looks like it time to resurrect ops_pct branch 22:24
22:29 mikehh joined 22:31 darbelo_ joined 22:32 darbelo left
cotto_work bacek: great! 22:33
22:33 darbelo joined
cotto_work I'd still like to spend some time speeding up the profiling runcore, but I' 22:33
m glad to work on that too.
actually, I have a pretty big todo list for profiling. I'll have to think about that when I get home and have my todo list in front of me. 22:34
bacek cotto_work: ok. I'll try to remember where we stopped. It was looong time ago...
chromatic I think we need to speed up the postprocessing script and fix line numbers more than improving performance.
cotto_work chromatic: any luck optimizing the script?
chromatic Haven't done any work on it yet. 22:35
cotto_work I'm looking forward to seeing what you can do with it.
I'd much rather have more to do than I have time for than be bored. 22:37
22:42 rg joined
Coke tries to catch up on review. 22:43
... and done. 22:46
22:54 tetragon joined
duk3leto cotto_work: if you see a poster for SQL Server, run! 23:24
cotto_work duk3leto: that 23:26
's a lot of running
Coke jdv79? 23:36
purl jdv79 is, like, a purl hater
Coke seen jdv79?
purl jdv79 was last seen on purl 9 hours, 30 minutes and 33 seconds ago, saying: <private message>
darbelo purl haters? 23:37
purl hmmm... haters is nom nom nom
darbelo karma (purl haters)
purl (purl haters) has neutral karma
darbelo (purl haters)++
23:42 patspam joined
nopaste "darbelo" at 200.49.154.172 pasted "partcl, boy or man?" (7 lines) at nopaste.snit.ch/17963 23:55
darbelo Man, apparently (en.wikipedia.org/wiki/Man_or_boy_test) 23:56

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!