Parrot 2.1.1 Released! | parrot.org/ | Tasks: PCC deprecations hackathon on Saturday, TT #389 branch
Set by moderator on 2 March 2010.
00:00 AndyA joined
allison plobsing: I'm not entirely opposed to the idea of callbacks, but I'm also not convinced yet that we need the feature 00:01
plobsing: and would rather do the PCC refactor without this added complexity, then see if the callback strategy makes any sense after the refactor 00:02
plobsing: it may be that this is a feature of Parrot::Embed, rather than of the base Parrot API
plobsing allison: ok, it can wait I guess. 00:03
allison: this isn't just about Parrot::Embed. It's about any forwarding interface of parrot.
allison plobsing: we're aiming to merge the hackathon branch back in after the weekend, so it's not a long wait
00:05 eternaleye joined
plobsing sweet. can't wait. 00:06
Whiteknight Austin++ 00:08
dalek TT #1497 created by Austin_Hastings++: Expose 'meth_cache' attribute of class.pmc 00:12
00:14 AndyA_ joined
Austin What does VTABLE_instantiate call? 00:14
nm, found it
particle coke: i have configured mollum on parrot.org for spam protection. if you need to adjust, see www.parrot.org/admin/settings/mollum 00:21
00:37 ash_ joined
Whiteknight question: does Object.mark need to mark the Class object, when the Class is already in the VTABLE, which gets marked? 00:51
00:54 parthm joined
Austin What about anonymous classes? 00:54
00:54 payload joined
Austin $P0 = newclass 00:54
Whiteknight we support that? 00:55
Austin or $P0 = new 'Class'
Yeah, that's how P6object creates the protoclass for the protoobject of every class.
The protoclass is an anonymous class that multiply inherits from P6protoobject and (your class).
Pm uses this to insert special protoobject behavior in front of your class' methods, except for some special ones. 00:56
Look in $_RTL/P6object.pir, at or around line 506 - "## create an anonymous class for the protoobject" 00:57
Anyway, I'm not real clear on the difference between 7 $P0 = newclass and 7 $P0 = new 'Class' , so I'm not so much telling you anything as suggesting that if there's a weird case, this is likely where you can find it. 00:59
01:03 abqar joined
parthm hello. i see that squaak grammar inherits from PCT::Grammar while the mk_language_shell generated skeleton inherits from HLL::Grammar. 01:07
Austin squaak is very old
parthm Austin: thanks. i will go with HLL::Grammar then. I was trying to locate its source, is it nqp-rx/src/stage0/HLL-s0.pir? 01:09
Austin I think it's in src/HLL/Grammar.pm
Whiteknight Austin: I don't see a newclass opcode with only one parameter 01:11
parthm there doesn't seem to be a src/HLL dir. These are the Grammar.*s that i found pastebin.com/Ew0y5xUv 01:12
Austin It'll have to be new 'Class' then
Parthm, are you looking at a parrot distro, or nqp-rx?
If so, that looks right. $_PARROT/ext/nqp-rx/src/state0/HLL-s0.pir 01:14
parthm Austin: I am using the 2.1.1 tarball. ok. will go through HLL-s0.pir
Austin: thanks. 01:15
Austin parthm, that's one big file, combining some hand-written and machine-generated code. Try this instead: github.com/perl6/nqp-rx/tree/fc0d84...01/src/HLL 01:16
parthm Austin: thanks for the pointer. HLL-s0.pir did look a little scary :-) 01:17
Austin :)
01:20 payload left 01:33 atrodo joined
Austin WTF? Why is add_parent a vtable op? 01:38
01:41 parthm left
allison Austin: because different object models behave differently when adding a parent 01:42
Austin And so it has to be an op rather than a method?
allison methods are slow as mud 01:43
Austin Sure, and you're going to call "add_parent" how many times, exactly?
allison which is to say "eventually, no, but for now, yes"
depends on the system 01:44
kid51 I guess it's appropriate that on every box on which I do a smolder test, I get a different set of TODO passed tests from t/dynoplibs/random-range.t ;-)
Whiteknight Austin: I've got plenty of complaints about the list of current VTABLEs. We can toss yours on the pile :)
Austin Whiteknight: I figured you'd be all over this one. There's already a method - it calls the vtable function. Just reverse 'em, deprecate the vtable, and you're done. 01:45
kid51: I've been thinking about that random-range test. I think another layer of statistics might quiet the thing down a bit. 01:46
Whiteknight by the way, I was *very* happy about allison's email earlier today about how we should be getting rid of delegate objects in Object PMC eventually 01:49
if we could fix the inheritance problem between Objects and PMCs properly, so many bugs will disappear
Austin Can anyone think of why 'new' would ignore a find-method override in a brand-new subclass?
Whiteknight what do you mean? 01:50
Austin Well, I *think* I'm creating a p6object class, with its associated baggage. And I'm fairly certain none of the parent classes has a vtable override. But when I lookup the pir::class() of the protoobject, and install a vtable override find-method on it, and then call $protoobject.new(), it goes directly to a sub, rather than visiting my find-method override. 01:52
mikehh kid51: t/dynoplibs/random-range.t seems to be rather random in it's results :-} - it gives different results every time it is run 02:12
try running prove t/dynoplibs/random-range.t a few times to see what I mean 02:15
Austin: I played around with the test a few times and could not get consistent results 02:17
Austin mikehh: I think the right thing to do is to make the current-day "test" ops into an accumulative result, and check if the result fails more than a certain percentage. 02:18
(FWIW, I believe it will fail that meta-test...)
mikehh also tried altering the parameters a bit but still had a few failures
bear in mind it is a test foir randomness in cryptographic systems which generally have much more stringent tests 02:20
s/foir/for/
and furthermore, a truly random system could still allow for 50 heads in a row or more 02:23
Austin You're right, of course. 02:25
Do you have a suggestion for a better test?
mikehh the test is usefull for looking at randomness of the pseudo-random number generators but I am not sure it should be part of the standard test suite 02:27
Austin Is there a non-standard test suite? 02:29
dalek kudo: 2d9808d | (Solomon Foster)++ | t/spectest.data:
Turn on S32-io/file-tests.t.
02:35 parthm joined 02:40 Andy joined
Coke Austin: when you remove a file, you need to update the manifest. 02:42
Austin Fixing. Sorry. 02:43
02:45 redbrain joined
dalek rrot: r44635 | Austin_Hastings++ | trunk/t/dynoplibs/random-range.t:
Resolving "unexpected TODO passes" in standard test suite.
 02:47
rrot: r44636 | Austin_Hastings++ | trunk/MANIFEST:
Updating MANIFEST
Coke anyone mind if I merge rm_cflags back? 02:53
Austin What, that old thing? 02:54
Tene Coke: that would be great.
Coke ok. only several dozen conflicts to resolve that I thought I already took care of. =-) 02:55
kid51 Coke: Branch built successfully for me tonight on Darwin/PPC: smolder.plusthree.com/app/public_pr...ils/32496. Also passed make codetest and make buildtools_tests 03:00
Also passed make test on Linux/i386: smolder.plusthree.com/app/public_pr...ails/32493
So, no objection here. I think this was the most intensive look at/rethinking of our configuration and build in 5 years! Coke++ 03:01
Coke heh 03:03
Austin Is there a way in `less` to temporarily hide certain lines? 03:13
Coke Austin: that seems like an odd question to me. 03:29
also, subversion sucks.
Austin Coke: while paging through a bunch of trace output, I'd like to hide the "current instr.:" lines, and particularly the "*** switching to" lines. But I know that when I find what I'm looking for, I'll want to see them again. So ... 03:31
plobsing Austin: I hardly use less. I pipe straight to my editor (vim) instead. Maybe that could work. 03:36
Coke mm, vip
Austin Maybe so. Does vim support masking?
plobsing it supports s/// and undo
Austin heh
I think you mean d 03:37
ash_ what do you mean masking?
parthm Austin: vim support folding (:h folding, also try selecting visual range and zf)
Austin That might work. I'll rtfm 03:38
ash_: I mean "don't show me lines like this" 03:40
Coke ok. the only thing stopping me from merging back rm_cflags is... svn.
ash_ ah, well i'd follow parthm's suggestion, you can fold the lines you don't wana look at
Austin Coke: I could swear we've been here before...
03:42 snarkyboojum joined, cotto joined
Austin Okay, it looks like vim's expression based folding is entirely capable of doing what I want. 03:42
Sadly, I'm not smart enough to use it. 03:43
Coke suggests just using grep.
Austin (Or, put another way, I'm not dumb enough to waste the brain cells necessary to figure it out.)
Yeah, that's the plan.
ash_ i have been slowly learning vim for like 2 years 03:44
i try to at least once a week read another part of vimtutor or the help section, but i don't even know how to code fold based off an exression :P
Austin I've been slowly forgetting vi for about 20, now.
ash_ expression
purl well, expression is a set of directions to an automaton.
Austin :)
Coke ok. svn merge ^/branches/foo gives me a gazillion conflicts. 03:45
(most of which I already fixed going the other way.)
Austin This will make a fold out of paragraphs separated by blank lines: 12 :set foldexpr=getline(v:lnum)=~'^\\\\s*$'&&getline(v:lnum+1)=~'\\\\S'?'<1':1
Coke svn merge --reintegrate ^/branches/foo fails with a ton of error messages like: 03:46
Missing ranges: /trunk/t/oo/root_new.t:43789-44631
Austin Hmm..
03:47 janus joined
Tene Coke: you might take notes about this and add them to the notes about svn merging problems. 03:51
Coke would be happy with "instructions on how to use svn properly." 03:52
the way we do it seems to be recommended against by one of the core svn developers. 03:53
(with no suggestion as to the right way, sadly.)
Austin Coke: did you merge trunk into ^/branches/foo first? If so, things are up to date and you can probably merge --ignore-ancestry instead.
Tene Coke: when I've had to merge svn branches, I just give up and do a squash merge in git, discarding the history, and then delete the branch.
04:13 parthm left 04:24 atrodo joined 04:55 parthm joined
dukeleto 'ello 05:07
dalek rrot: r44637 | dukeleto++ | trunk (2 files):
[docs] Update tests.pod a bit and remove a wrong comment from t/op/sprintf.t
 05:31
05:38 dns joined, hercynium joined 05:41 eternaleye joined 06:00 dns left
dukeleto jhelwig: trac.parrot.org/parrot/ticket/949 06:01
06:02 petdance joined 06:24 cotto joined 06:27 janus joined 06:41 parthm left
jhelwig Tene: I'm pretty sure that you can actually just do the merge normally in Git, and dcommit that with a recent enough Git, and SVN server. 06:58
07:06 chromatic joined
chromatic DOS is now done. 07:09
dalek rrot: r44638 | chromatic++ | branches/pcc_hackathon_6Mar10/compilers/imcc/pcc.c:
[IMCC] Changed IMCC's PCC generator to emit get_results *after* the
rrot: r44639 | chromatic++ | branches/pcc_hackathon_6Mar10/compilers/imcc/instructions.c:
[IMCC] Modified the opcode introspection for detecting instructions which write
chromatic The good news is that it only required changes in two places, and I knew where both of them were with a bit of searching. 07:16
The bad news is that function calls now do not work. At all.
Austin Well, welcome to 1981. 07:17
What's DOS stand for?
chromatic Lotus DOSn't run. 07:19
Austin Heh
Well, I did my share of batch files. I can make do without function calls.
dalek tracwiki: v21 | chromatic++ | CallingConventionsTasklist 07:20
tracwiki: Made the IMCC get_results/invoke ops order changes when generating PBC
tracwiki: trac.parrot.org/parrot/wiki/Calling...ction=diff
Austin Alternatively, you could google svn revert and get them back.
chromatic This was always the plan, Ron Moore style. 07:21
07:53 tuxdna joined 08:03 parthm joined 08:09 iblechbot joined 08:17 parthm_ joined 08:19 parthm_ joined 08:26 riffraff joined 08:29 parthm left 09:12 payload joined 09:43 mikehh joined 09:56 mikehh joined 09:58 parthm joined
parthm i was trying (unsuccessfully) to find where the make function used (as in 'make $past') in actions.pm defined. any pointer? 10:07
allison parthm: it's a builtin of the language NQP 10:23
parthm allison: thanks. i see 'stage2/gen/nqp-actions.pir:.sub "prefix:sym<make>"', I suppose thats what you mean? what exactly does the method do? 10:26
allison Patrick can give more of the internal details, but it mainly acts to return the PAST structure you create as the result of the rule. 10:31
parthm allison: thanks. 10:33
10:46 parthm left 10:48 aukjan joined 11:05 mj41_ joined 11:08 snarkyboojum joined 11:21 payload joined 11:25 shockwave joined
dalek kudo: d4be43d | (Solomon Foster)++ | src/builtins/Int.pir:
Change infix:<===>(Int, Int) to return Bool.
 11:29
purl dalek: that doesn't look right
11:52 snarkyboojum joined
dalek rrot: r44640 | allison++ | branches/pcc_hackathon_6Mar10/src (3 files):
[pcc] Rework set_returns and get_results to use the same code paths as

set_args and get_results operate from the caller's context, while get_params and set_returns operate from the called sub's context). This change sucessfully makes basic PASM and PIR subroutine calls with the new opcode order.		 12:04
12:15 AndyA joined
dalek rrot: r44641 | allison++ | branches/pcc_hackathon_6Mar10/t/op/calling.t:
[pcc] Reorder the 'get_results' opcode calls in PASM calling convention tests.
 12:20
12:25 iblechbot joined 12:27 tetragon joined 12:41 whiteknight joined
whiteknight everybody ready for tomorrow's T3H M4J0R H4XX0RING tomorrow? 12:46
12:51 payload joined 12:53 mikehh joined 13:07 rblackwe joined 13:08 snarkyboojum joined
dalek izkost: 0fe41d3 | unknown++ | build/ (2 files):
Don't use := in makefiles, it's unportable.
 13:12
whiteknight chromatic++ (work on the hackathon branch already)
purl msg chromatic Would it be possible (or even worthwhile) to put together a test in t/compilers/imcc to prove that a PIR function call generates opcodes in the correct sequence? 13:15
purl Message for chromatic stored.
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#32505), fulltest) at r44641 - Ubuntu 9.10 i386 (g++ with --optimize) 13:23
dalek rrot: r44642 | mikehh++ | branches/pcc_hackathon_6Mar10/src/ops/core.ops:
fix codetest failure - trailing whitespace and copyright
 13:25
13:26 bacek joined
dalek izkost: ac53c46 | unknown++ | src/pmc/p5 (5 files):
Some hacks that prevent weird __attribute__ related errors when compiling on Win32, so I can actaully hack on Blizkost again.
 13:34
mikehh pcc_hackathon_6Mar10 branch make corevm ok, make fails make: *** [compilers/tge/TGE/Parser.pir] Error 1 13:38
make coretest - mega errors - finally gets lost in t/compilers/imcc/syn/regressions.t 13:40
Coke if TGE or PGE fails, that's usually a sign that some core thing without enough testing is failing.
dalek rrot: r44643 | allison++ | branches/pcc_hackathon_6Mar10/src/pmc/continuation.pmc:
[pcc] Remove a hack that's no longer needed, since all returns are
 13:42
mikehh rm_cflags branch - still failing post-config with g++ - i386 this time - t/steps/auto/warnings-01.t - Failed test: 17 13:50
13:50 atrodo joined
Coke mikehh: argh. ok. will fix that before mergeback. 13:59
mikehh rm_cflags branch: 14:02
post-config FAIL - t/steps/auto/warnings-01.t - Failed test: 17
all other tests PASS (pre-config, make corevm/make coretest, smoke (#32507), fulltest) at r44643 - Ubuntu 9.10 i386 (g++ with --optimize)
Coke ugh. perl Configure.pl --verbose-step complains about auto:warnings, but silently accepts auto:warning (neither of which is a valid config step) 14:05
IWBNI Configure steps had a way to depend on other steps, like, say, in a makefile. 14:14
dalek rrot: r44644 | allison++ | branches/pcc_hackathon_6Mar10/src/call/args.c:
[pcc] Remove code supporting hack for continuation parameters, since
 14:15
rrot: r44645 | allison++ | branches/pcc_hackathon_6Mar10 (2 files):
[pcc] Remove unused function to build a special return signature from
kudo: 0126a16 | (Solomon Foster)++ | t/spectest.data:
Turn on S32-list/pick.t.
 14:20
kudo: 0728931 | (Solomon Foster)++ | src/builtins/Str.pir:
Switch .s to .filesize.
kudo: b348b3e | (Solomon Foster)++ | src/builtins/Str.pir:
.filesize became .filebytes before I could get it pushed.
mikehh rm_cflags branch: 14:27
All tests PASS (pre/post-config, make corevm/make coretest, smoke (#32508), fulltest) at r44645 - Ubuntu 9.10 i386 (gcc with --optimize)
dalek rrot: r44646 | allison++ | branches/pcc_hackathon_6Mar10 (2 files):
[pcc] Remove function only left in place to support now-removed
 14:32
14:52 rt7 joined 14:55 snarkyboojum joined 14:56 lucian joined
dalek rrot: r44647 | coke++ | branches/rm_cflags/config/auto/warnings.pm:
Be nice if the g++ warnings were actually used for g++.
 15:05
Coke mikehh++ #typo 15:07
mikehh++ # other commit with no builtin karma.
mikehh: try now.
Coke restarts the mergeback to trunk. 15:08
sloooooooooooooooooooow.
Coke wonders if this would be faster if he ran it on parrotvm.osuosl.org 15:09
Coke suspects that would negatively impact web services, so doesn't 15:11
15:15 mikehh joined
Coke ok. i'm merging back and ext/nqp-rx is getting updated? 15:17
... this'll be fun. :P
15:21 patspam joined
dalek rrot: r44648 | coke++ | branches/rm_cflags/t/steps/auto/warnings-01.t:
Make this test pass on g++, add note about how fragile this test is.
 15:21
mikehh Coke: looks good 15:23
15:25 Andy joined
Coke tries a run with --ignore-ancestry 15:27
15:30 bubaflub joined
Tene Coke: if that doesn't work, I'll try a merge from git-svn. 15:31
mikehh Coke: I get a problem if I add the --verbose flag to configure with both gcc and g++ - not the configure but with t/steps/auto/neg_0-01.t 15:32
it reports Parse errors: Tests out of sequence. - Failed 2/23 subtests - but reports ok too all tests 15:35
Coke mikehh: does that happen on trunk? 15:36
mikehh never tried it there - let me see
15:37 Psyche^ joined 15:39 clinton joined
mikehh Coke: yes - same result on trunk - same test gives error information with --verbose but the test actually seems to pass, not collecting the info properly 15:40
15:41 theory joined
Coke mikehh: excellent. (from my POV) 15:42
mikehh prove -v t/steps/auto/neg_0-01.t seems strange when configured with --verbose
15:49 riffraff joined
nopaste "mikehh" at 81.149.189.7 pasted "post-config test failure when configuring with --verbose" (67 lines) at nopaste.snit.ch/19856 15:54
mikehh I am not sure I understand the output from the test - it passes configured without --verbose 15:56
I did not see anything wrong with the --verbose output 15:58
bubaflub mikehh: i think it has to do with the output of "(no)" and then the TAP consumer is seeing that as a fail
in front of test 12 15:59
in front of test 10 is the same deal (i'm guessing) - output in front of the tap output is messing up prove
mikehh if you look at the output it somehow misses the (yes) as well as the (no)
bubaflub mikehh: yep, i think that's it. the (yes) and (no) in front of the tests are confusing prove and it thinks the tests are out of order. it seems like with the --verbose flag the output is being captured correctly. 16:03
mikehh and also the test passes if configured normally without the --verbose - it doesn't have the (yes) and (no) 16:04
bubaflub it looks like lines 121/122, and 104/105 for (no) and (yes)
in the test
mikehh agreed, but why does the test output it when the --verbose option is specified with configure - it does not normally 16:07
Coke tene: this is a PITA. I can probably do it, but if you can do it quicker with git-svn...
mikehh: because those tests are RUNNING configure bits. 16:08
so the test should disable verbose explicitly.
mikehh Coke yes - line 97 and line 114 in the test 16:12
Tene Coke: trying to push a normal merge commit now 16:19
mikehh the is/like on line 104/105 and 121/122 16:20
Tene looks like it's going through fine
Coke: r44649
Coke: any chance I could get you to estimate how long you spent trying to get the merge to work?
Coke ... several hours. 16:21
(on and off)
which is entirely likely due to: 1) myself not properly doing the merge or 2) other people screwing things up and leaving the mergeinfo stuff everywhere.
16:21 davidfetter joined
Tene IMO, that's not something a VCS should be asking its users to do in the first place. 16:22
16:23 cotto_work joined
cotto_work helo 16:24
Coke Tene: oh, I'm sure it's all supposed to work automatically somehow.
Tene: now, there's some minor properties cleanups I have to do here, i think, but it'll be simpler to do this as a two step process. 16:26
cotto_work What is this witchcraft? Make is telling me what it's actually running? 16:27
Coke hopefully this is seen as a net win. =-) 16:28
dalek rrot: r44649 | tene++ | trunk (57 files):
Merge branch 'rm_cflags'
Tene aw, now that it's merged, I can't merge rm_cflags into my exceptions branch to cause more merge troubles later.
Coke (I for one am glad that I can now cut and paste the command line to rerun something.)
Tene: *finger snap*
ttbot Parrot trunk/ r44649 MSWin32-x86-multi-thread make error tt.ro.vutbr.cz/file/cmdout/221179.txt ( tt.ro.vutbr.cz//buildstatus/pr-Parrot/rp-trunk/ ) 16:29
Tene Coke: does this also let make be more-intelligent about dependencies and rebuilding things? 16:31
or was that a different proposed build system cleanup?
Coke Tene: don't think there's anything particularly hitting that target in this merge. 16:32
(though killing most of auto::cgoto helps slightly)
Tene Okay, I must have confused it with something else.
particle wow, there's almost enough output during make on windows now to disguise the thousands of warnings... almost. 16:33
unfortunately, it doesn't disguise the error at the end. 16:34
Tene Yay, I get a segfault when building PGE now! 16:35
Coke I have a fix for that.
and this is why I asked for feedback on windows. =-)
ttbot++ 16:36
particle ttbot++
looks like an easy fix
Coke yup. just extra spaces in a few places. 16:37
particle but, it's too bad that the path fixup for display of files during compilation is gone
nopaste "tene" at 76.27.121.193 pasted "PGE segfault" (5 lines) at nopaste.snit.ch/19857
particle windows builds used to show 'src/gc/system.c' when compiling that file, now they show only 'system.c'
Tene lemme try to get a backtrace
Hmm. Looks like my realclean didn't run, actually. re-trying. 16:39
Coke particle: that's because that's what your compiler shows. 16:40
particle yes, and ron had provided a fixup for it
Coke in cflags, which is gone.
particle that was in cflags?
Coke yes.
particle odd
Coke it was the same thing that was hiding all the output.
er, the command line.
cotto_work outlook-- 16:41
particle ah, funny place for it.
cotto_work Those were not extra line breaks.
Coke not really. where else would it go?
particle cotto_work: i know folks on the outlook team. shall i shoot them for you?
Coke particle: are you using cl?
whiteknight particle: that would be handy
particle coke: yes
whiteknight I can't think of a piece of software I dislike more
Coke Looking at tt.ro.vutbr.cz/file/cmdout/221179.txt shows the path.
particle itunes is worse than outlook
ttbot Parrot trunk/ r44650 MSWin32-x86-multi-thread make error tt.ro.vutbr.cz/file/cmdout/221228.txt ( tt.ro.vutbr.cz//buildstatus/pr-Parrot/rp-trunk/ ) 16:42
whiteknight that may be true, but I don't use itunes
particle me neither :)
cotto_work particle, that's a kind offer but I think I'll decline.
Coke fix for /Fo in.
er, -Fo
particle cotto_work: my rates are fair. consider it and let me know.
Coke Can anyone translate the error message at the bottom of that last ttbot report? 16:44
whiteknight Coke++ the new makefile is going swimmingly on my system
dalek rrot: r44650 | mikehh++ | trunk/MANIFEST.SKIP:
regenerate MANIFEST.SKIP
rrot: r44651 | coke++ | trunk (2 files):
Fix the build for cl, which doesn't allow a space between the -Fo option and the value.
 16:45
Tene Coke: sorry for the false alarm. no segfaults for me.
whiteknight ACtually... build just failed 16:46
Tene build worked perfectly for me
whiteknight this.wsnProgramingPanel1.
bison -v -y compilers/imcc/imcc.y -d -o compilers/imcc/imcparser.c
conflicts: 1 shift/reduce
ttbot Parrot trunk/ r44651 MSWin32-x86-multi-thread make error tt.ro.vutbr.cz/file/cmdout/221252.txt ( tt.ro.vutbr.cz//buildstatus/pr-Parrot/rp-trunk/ )
whiteknight bison: cannot open file `compilers/imcc/imcparser.output': Permission denied
Coke are you running with --maintainer? 16:48
whiteknight Coke: yeah 16:52
I'm reconfiguring without
Coke might also want a realclean. 16:53
(the .output file there is kind of a hack.)
I just tried to force a --maintainer run, and nothing changed, so I think we're good, there.
particle coke: i have svn trunk head on my win32 box, need any thing run in particular? 16:55
running 'perl Configure.pl && nmake world && nmake fulltest' 16:59
ttbot Parrot trunk/ r44652 MSWin32-x86-multi-thread make error tt.ro.vutbr.cz/file/cmdout/221313.txt ( tt.ro.vutbr.cz//buildstatus/pr-Parrot/rp-trunk/ ) 17:00
dalek rrot: r44652 | coke++ | trunk (4 files):
fix svn:ignore, regen manifest.skip, rerun headerizer.
 17:01
Coke particle: danke. 17:02
if you want to start masking some of those warnings with windows, I can point you at the spot in config/auto/warnings.pm for it. 17:03
(though I'd rather just fix the warnings.)
particle build failure with default.obj
Coke ... I can fix that.
particle 7400+ lines of build output make it hard to copy/paste 17:04
i fear it'll be more than just one, lemme look closer 17:05
Coke it's the same space bug but with pmcs. 17:06
particle search tt.ro.vutbr.cz/file/cmdout/221263.txt for 'D9027'
ah, great, maybe that'll take care of it.
Coke particle: fixed. 17:07
44653
gotta run for lunch. feel free to queue up any more bugs you find. 17:08
particle danke
shockwave I know this is besides the topic. I would like to register my nick, since I'm basically a regular around here, by now. I've been Googling on how to do it on this server, but I haven't found any good matches. Can anyone point me to a location with some info about it.
Coke there is no nick registration on this irc server.
particle there's no nick registration on this network
shockwave oh, alrighty.
purl alrighty. is there a way to do that without having to start a new process?
shockwave Thanks, particle. :) 17:09
particle we now consider you 'vaguely human'
shockwave lol
I do have one question related to Parrot. I did a test, but just want to confirm it. Is,
particle opbots, trust shockwave
slavorg Ok
slavorgn Ok
bubaflub Coke, not sure if you need it, but I'm building trunk on Mac OS X 10.6.2 17:10
Coke bubaflub: hokay.
17:10 frodwith joined
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#32509), fulltest) at r44650 - Ubuntu 9.10 i386 (g++ with --optimize) 17:10
Coke and now, ->
shockwave 'null' is supposed to only be used to initialize variables, correct? Is not used in other ways, such as: foo(null), right? 17:11
Coke oh, and someone can remove rm_cflags branch, I guess.
null is an opcode.
particle coke: linker error
Coke ttbot? hurry up.
particle building is so much speedier, though. these errors come up much faster than before.
Coke particle: that's because we're not invoke perl to invoke cl. 17:12
bubaflub Coke: builds fine, i can run make coretest
Coke bubaflub: danke.
shockwave Coke, but one can do this: $P0 = null
oh, right.
Coke shockwave: that is sugar for null $P0
shockwave An opcode that has no operands.
I get it.
particle c:\\usr\\perl-5.10.1-x86\\bin/perldoc ...
misslash
shockwave Thanks, Coke.
Coke particle: thought I got all of those. checking...
Tene shockwave: one operand
shockwave Tene: True. 17:13
particle seems to be with maps_ops_switch
shockwave yay! I have an arroba in my name. I'm actually smiling right now.
Wow, I'm a nerd.
particle ah, no, just after... it's podifying src/packfile.c
bubaflub Coke: looks like i'm getting some errors on some tests under t/pmc 17:14
Coke particle: have a fix. 17:15
bubaflub Coke: nevermind; could be a problem that i didn't make realclean. i'm going to try building again and make coretest-ing again. if it comes up, i'll letcha know.
Coke one of things this merge did was eliminate the blind substitution of / to \\, so any place that needs a real local slash has to specify one. (perldoc didn't) 17:16
bubaflub I also get a warning when auto::arch runs - "Use of uninitialized value in pattern match (m//) at config/auto/arch.pm line 63."
particle coke++
ttbot Parrot trunk/ r44653 MSWin32-x86-multi-thread make error tt.ro.vutbr.cz/file/cmdout/221362.txt ( tt.ro.vutbr.cz//buildstatus/pr-Parrot/rp-trunk/ ) 17:17
Coke particle: try that and now I'm really going. =-)
->
dalek rrot: r44653 | coke++ | trunk/config/auto/pmc.pm:
another cl build fix - no spaces between -Fo and the file allowed.
rrot: r44654 | coke++ | branches/rm_cflags/config/auto/perldoc.pm:
Don't rely on auto-conversion of / to \\ anymore.
17:17 jan joined
particle sigh, coke made that change on the branch instead of trunk. 17:24
bubaflub Coke: false alarm on coretest; they all pass on R44653 on Mac OS X 10.6.3 17:25
dalek rrot: r44655 | particle++ | trunk/config/auto/perldoc.pm:
Don't rely on auto-conversion of / to \\ anymore.
 17:34
rrot: r44656 | particle++ | trunk/config/gen/makefiles/root.in:
fix yet another slashie
particle coke++ build success, testing now 17:37
ttbot Parrot trunk/ r44655 MSWin32-x86-multi-thread make error tt.ro.vutbr.cz/file/cmdout/221427.txt ( tt.ro.vutbr.cz//buildstatus/pr-Parrot/rp-trunk/ )
17:39 clinton joined
tewk Is there a C enum that maps op names to op number? 17:40
whiteknight Coke: i realcleaned and the bison problem went away 17:41
tewk: No, not that I am aware of
purl okay, whiteknight.
17:41 iblechbot joined
whiteknight tewk: you can look them up by string name. See the OpLib and Opcode PMCs for examples 17:42
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#32510), fulltest) at r44654 - Ubuntu 9.10 i386 (gcc with --optimize) 17:44
headinh back to amd64 now - bbiab
particle oh, come on! t/op/arithmetics.t makes false assumptions about win32, and i can't blame anybody because svn thinks this .t file is binary. 17:45
i don't want to go down this rabbit hole just to blame dukeleto, i'll just blame him now :P
17:48 kurahaupo joined
particle interesting that the file has svn:mime-type set correctly 17:48
wtf svn blame?
purl svn blame is just like p4 annotate, only better or just like git blame, only different
tewk whiteknight, I'm writing a runloop in c, note ops2c should generate a C enum 17:56
particle doesn't sound hard to emit 17:57
kurahaupo A while back the "topic" used to include the URL for the chat log, partly as a way of pointing out that it *is* logged. Which begs two questions: is this channel logged, and if ao where? 17:59
Tene irclog.perlgeek.de/parrot/today 18:00
yes
kurahaupo Thanks 18:01
Tene np
kurahaupo Tene++
Tene Feel free to add it to the topic yourself. 18:02
18:02 Austin joined
Tene It's good information to have available. 18:02
moderator #parrot Parrot 2.1.1 Released! | parrot.org/ | channel log: irclog.perlgeek.de/parrot/today | Tasks: PCC deprecations hackathon on Saturday, TT #389 branch 18:03
mikehh joined
moderator #parrot Parrot 2.1.1 Released! | parrot.org/ | Channel log: irclog.perlgeek.de/parrot/today | Tasks: PCC deprecations hackathon on Saturday, TT #389 branch
particle kurahaupo++ 18:03
Tene There used to be a bot that told everyone that the channel is logged. i've long-since ignored it, so I didn't notice that it was gone. 18:04
particle is it gone? i've ignored it, too
whiteknight yeah, the bot left
everybody hated it
Tene It seems that nobody could imagine a user that wanted to know if they were being logged, and if so, where. 18:05
;)
kurahaupo It had a pain in all the diodes down its left side... 18:06
cotto_work tewk, where's your code living? 18:07
dalek rrot: r44657 | particle++ | trunk/t/op/arithmetics.t:
[t] check for 'has_negative_zero' in config, rather than guessing at platform support
Austin Actually, Tene, I think it was more that the bot was (a) too verbose; and (b) hostile in its phrasing. 18:11
Tene I disclaim any ability to judge appropriate levels of hostility and verbosity. :) 18:12
Austin I doubt that - you ignored it, didn't you?
Tene Yes, I did.
tewk cotto_work, in my private git repo right now. 18:15
Coke particle: whoops, committed to wrong branch. danke. 18:43
particle: svn blame --force 18:45
particle i get failures in t/pmc/eval.t when i run during the test suite, but not with prove 18:46
heisenbugs--
Coke different parrot command line args.
try perl t/harness t/pmc/eval.t
ttbot? 18:47
purl ttbot is TapTinder build bot owned by mj41 and reporting tt.ro.vutbr.cz/buildstatus/pr-Parrot/rp-trunk build errors. or a master of timing.
Coke the bot is green.
plobsing particle: which tests in t/pmc/eval.t are failing? some of them are known issues. 18:54
particle i think 8-10, 13, but i'm rerunning everything now to be sure 18:55
i'm somewhere in t/op atm
Coke particle: the build is noticably faster on windows, you said, yes?
particle coke: so it seems, but i haven't built in quite a while. give me a previous rev to build, and i'll let you know 18:56
Coke I'd love to see some analysis between 44637 (last green one before the mergeback) and r44656, first green one after.
my timing checks on *nix were only about 1%.
Austin: note that repeated attempts to get a set of approved bot behavior failed. 18:58
19:00 chromatic joined
Coke sanity check: -Wunreachable-code is a valid option? 19:02
dalek rrot: r44658 | coke++ | branches/rm_cflags:
This branch has been merged to trunk, ttbot now shows green.
rrot: r44659 | coke++ | trunk/config/auto/warnings.pm:
build is clean on this warning, move it to the default set.
Coke dalek? 19:07
purl dalek is #parrot's spammy little rss bot or (see: dalek plugins)
19:08 szabgabx joined
chromatic We should try to be -Wunreachable-code clean. 19:08
cotto_work +1
purl 1
Coke chromatic: you'd need to fix ASSERT_ARGS. 19:09
as I just found out. (44659 is wrong, I reverted it)
chromatic How so?
Coke src/byteorder.c:511: warning: will never be executed 19:10
and that's the ASSERT_ARGS.
chromatic That probably needs decoration in include/parrot/exceptions.h:274 (cribbed from line 273). 19:15
plobsing 2 pmc2c questions: is it possible to get a dynpmc in a namespace (eq: myapp;widget)? is it possible to statically compile dynpmcs into an embedding app? 19:18
tewk plobsing, on question #1, last I knew that still didn't work but that is the long term plan. 19:20
cotto_work 2 shouldn't be a problem 19:23
It might require a little hacking, but not too much. 19:24
dalek rrot: r44660 | coke++ | trunk/config/auto/warnings.pm:
revert r44659, build is NOT clean on this warning (yet)
plobsing cotto_work: I'm more asking if calling Parrot_lib_myapp_group_load() is breaking encapsulation
dukeleto got postges to actually execute some PIR code last night 19:33
bubaflub hi-five dukeleto
Coke if -Wunused-function complains about a static function, it's 99% safe to remove, yes? 19:34
(src/gc/gc_inf.c has many) 19:35
dukeleto bubaflub: thanks! 19:38
chromatic Coke, those probably require some reading of the code.
Coke they don't seem to be referred to anywhere. 19:39
typically in this situation, there's a struct somewhere that contains the references to the functions.
(I don't even see that.)
how to run with the gc_inf core? 19:40
nopaste "particle" at 98.232.31.142 pasted "timed config/build, pre-rm_cflags merge" (7671 lines) at nopaste.snit.ch/19859
Coke nevermind, found it.
ugh. directions in src/gc/gc_inf.c are wrong. =-) 19:41
particle wonder what happens to postgres when you run 'hcf' in it... 19:42
Coke particle: that opcode was removed.
particle oh, come on! it was a great shortcut.
Coke ... hey, if you want to volunteer to answer all the tickets about why our test suite is coredumping, put it back. 19:43
the INF core is borked here anyway. 19:45
chromatic bacek said that a couple of weeks ago, but I thought he had it working again. 19:47
Coke TT#1498 19:48
whiteknight i thought he did too
Coke we need to test it or kill it.
(I'm all for the former.) 19:49
whiteknight It's purpose is purely academic
Coke right, but pedagogical code that breaks unintentionally doesn't help.
nopaste "particle" at 98.232.31.142 pasted "timed config/build, post-rm_cflags merge" (7622 lines) at nopaste.snit.ch/19860
Coke particle: danke. 19:50
fyi, there is some tool in the resource kit that is like 'time' on *nix.
can't remember the name of it.
particle yes, 'timeit', in server 2003 reskit 19:51
but that's not compatible with w7 iirc
Coke 44*60+07.73
purl 264773
Coke er...
(44*60+07.73) - ( 40*60+35.08)
purl 262337.92
whiteknight Coke: I can't decide whether it's worthwhile to keep the INF core in the build or move it out as some sort of example code 19:52
obviously we don't support dynloading of GC cores, and we may never 19:53
Coke ok. windows time post change is 212.65s; pre change is 283.59 19:54
283/59/212.65
purl 0.0225563610133625
Coke 283.59/212.65
purl 1.33359981189748
Coke not shabby. 19:55
(and that includes Config time.)
chromatic 25% faster
Coke particle: don't say I never did anything for ya. =-)
dalek TT #1498 created by coke++: gc_inf core breaks build. 19:56
particle coke++ :)
Coke wonders why -Wno-pointer-sign is in --cage, if pointer-sign is already implied by -Wall 19:58
chromatic Your -Wall is different from mine. 19:59
Coke fair enough. so shouldn't we have -Wpointer-sign in cage instead of shutting it off? 20:00
(the stuff in -case is kind of a hodge-podge) 20:01
*--cage
chromatic GCC adds things to -Wall sometimes.
Coke and if -Wall is different for different users, we shouldn't use it, neh? 20:02
(we should be explicity)
*explicit
(looks like we are -Wpointer-sign clean anyway) 20:05
dalek rrot: r44661 | coke++ | trunk/config/auto/warnings.pm:
we are -Wpointer-sign clean;
 20:13
rrot: r44662 | coke++ | trunk/config/auto/warnings.pm:
The build is -Wmultichar clean.
20:15 joeri joined
Coke whiteknight, particle: can you verify if trac.parrot.org/parrot/ticket/838 is resolved? 20:29
dalek rrot: r44663 | coke++ | trunk/lib/Parrot/Pmc2c/PCCMETHOD.pm:
[cage] remove unused label in generated C code.
 20:30
whiteknight Coke: I'll try to check it out later 20:31
Coke whiteknight: danke.
20:32 davidfetter joined
particle coke, whiteknight: it's fixed, ticket updated 20:35
whiteknight w00t
Coke do we have a need to support dmake anywhere? 20:46
(looking at TT #775; don't need it for strawberry)
dalek TT #838 closed by particle++: Configure.pl mangles forwardslashed-flags for CL
Coke the manifest stuff is probably clean-able, though. 20:47
20:49 bacek joined
Coke any reason not to delete 'sub truth' from lib/Parrot/Configure/Compiler.pm ?? 20:51
dalek a: 080b7bb | fperrad++ | dynext/pmc/luathread.pmc:
move Parrot::Coroutine to Parrot;Coroutine
 21:02
21:31 fperrad joined
shockwave I'm looking at: examples/tutorial/33_hashes.pir 21:33
Is there another syntax for adding and fetching elements of hash? 21:34
A syntax more like: $I0 = add 1, 2
I'm also looking at src/ops/*.ops, and docs.parrot.org/parrot/latest/html/...h.pmc.html 21:36
but nothing jumps out at me.
(which would indicate another syntax)
whiteknight for elements of a hash you can write $P0 = myhash["hey"] 21:37
or, myhash["key"] = $P0
internally, it's all the set opcode, I think
set myhash, ["key"], $P0
or set $P0 myhash ["key"]
shockwave whiteknight: I'll give those a try. That's the syntax I'm looking for, in order to be consistent with the way I'm currently outputing the other code. 21:38
Thanks.
cotto_work moving tiem 21:39
21:41 bubaflub joined
Tene shockwave: just remember that PIR does *not* support nested expressions. 21:46
Coke particle: did you ask them to tranform my captcha request into a mollom install, or did they just do that on their own? 21:50
21:50 snarkyboojum joined
particle coke: i said whatever tech works 21:51
do you have a preference?
Coke particle: I had opened a ticket. 21:52
(directors were cc'd on that.)
I said "please install foo" and they replied "ok, bar is installed". =-)
just wondering if I missed some conversation somewhere. 21:53
particle ah, sorry about that. feel free to make them redo, and blame it on me.
there was conversation in #osuosl
Coke no, I don't mind. I'm just wondering how that request got changed.
because, as usual, I am out of the loop. =-)
particle ok, then, my fault. :)
time for lunch &
Coke now I know not to complain to them. danke.
22:19 iblechbot joined
dalek TT #1499 created by arnsholt++: Null pointer dereference in imageio.pmc 22:40
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#32518), fulltest) at r44663 - Ubuntu 9.10 amd64 (gcc with --optimize) 22:53
23:23 kurahaupo joined
dalek rrot: r44664 | chromatic++ | branches/pcc_hackathon_6Mar10/t/compilers/imcc/imcpasm/optc.t:
[t] Reclaimed one failing test which relied on the old ordering of get_results
 23:31
rrot: r44665 | mikehh++ | branches/pcc_hackathon_6Mar10/include/parrot/call.h:
remove defines etc removed from call/args.c
23:34 kid51 joined
mikehh pcc_hackathon_6Mar10 branch at the moment make coretest hangs in t/compilers/imcc/syn/regressions.t 23:35
after some errors
23:36 snarkyboojum joined
mikehh make corevm ok - make fails 23:37
chromatic It won't build for a while yet. 23:39
23:48 bacek joined 23:53 snarkyboojum 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!