Parrot 0.9.0 "From Outer Space" Released | parrot.org/ | migration to svn.parrot.org begins at 19:00 UTC on Thurs
Set by moderator on 26 January 2009.
00:07 davidfetter joined
dalek r36042 | jkeenan++ | trunk/t/pmc: 00:07
: Applying patch submitted by Ron Schmidt in
: trac.parrot.org/parrot/ticket/230. Uncomment several tests which are
: now passing.
review: www.parrotvm.org/svn/parrot/revision?rev=36042
00:09 AndyA joined 00:12 kid51 joined
particle \\o/ i've got NaN/Inf support working here :) 00:31
Coke finds free wifi at his son's bball practice.
particle too bad my vm doesn't have net access. sigh. 00:34
nopaste "particle" at 76.121.106.245 pasted "inf/nan handling patch (working)" (382 lines) at nopaste.snit.ch/15430
dmajnem2_ hmm, I think I can replicate test failures on other runcores using more powerful code instrumentation
00:35 Andy joined
rg particle: are you sure you want to define the int constants? i'm getting a floating point exception when trying to divide integers by zero. 00:43
particle rg: i may not need the int consts...
rg after all, there is no integer inf/nan. 00:44
particle yeah, i don't really need them
i'm updating trunk after a vm reboot, to build and run tests now 00:45
i've removed the intval stuff locally, rebuilding now 00:47
00:50 tetragon joined 00:53 bacek_ joined
bacek_ seen chromatic 01:01
purl chromatic was last seen on #parrot 3 hours, 33 minutes and 33 seconds ago, saying: I think it's part of GCC.
chromatic pong 01:02
bacek_ chromatic There is typo in Modern::Perl POD: See L<http://www.modernperlbooks.com/mt/2...l.html>
shorten bacek_'s url is at xrl.us/bedvt7
bacek_ double http://
chromatic Where?
bacek_ chromatic: in url :) 01:03
chromatic Oh I see.
01:08 tetragon_ joined 01:10 tetragon joined
Coke chromatic: see this: code.google.com/p/partcl/source/bro...s/fact.tcl 01:14
shorten Coke's url is at xrl.us/bedvwu
Coke that generates 10 virtually identical anonymous subs for the [expr] invocation.
chromatic How "virtually identical" is "virtually identical"? 01:16
01:32 tetragon_ joined 01:57 skv joined 02:07 skv joined
Coke sorry, laptop power died before I could nopaste. 02:16
nopaste "coke" at 72.228.52.192 pasted "evals, this time in order, for fact.tcl" (3157 lines) at nopaste.snit.ch/15431
Coke look at EVAL_53 (... 54, ...55...) 02:17
chromatic That's fairly identical. 02:22
Do you have an expr builtin? 02:23
Coke yes. [expr] is a sub in runtime/builtin/expr.pir 02:26
chromatic Ah. 02:27
Coke er. not a builtin like a perl builtin.
chromatic Do you always have to store the arguments to [expr] as a string, or can you parse them to AST nodes? 02:28
I mean, at compile time.
Coke wonders if "EVALS.out" is some kind of special filename. 02:30
chromatic: I don't think I have any infrastructure to support that at the moment, but I think it's doable. 02:31
02:31 tetragon joined
chromatic That seems like a fruitful line of research to me. 02:32
You may have to invoke the toNumber words on the appropriate nodes at runtime, for example, to abuse terms of art from different languages, but you shouldn't have to reparse strings every time. 02:35
Coke i suspect a switch to pct will help facilitate such things. 02:37
my evil plan from some time ago was to put in certain functions into the parse; cheat and make them work like were language syntax. then if someone changed one, update the grammar and rebuild it. 02:38
chromatic Factor does something similar. 02:40
Coke ig going to have to start blocking game requests on facebook. 02:42
This is crazy.
chromatic Or you could do what I do, and ignore Facebook.
Coke is attempting to be vaguely social, you scumsac. 02:43
Tene IRC counts as socialization. 02:45
... it had better, at least.
cotto This is odd. When the tcl code is called with 1/7.0, assigns the correct value to a float, then the wrong (rounded) value to a different float. 02:48
with 1.0/7.0 it only creates one extra float, with the correct value 02:49
02:52 Tene_ joined
Coke cotto: presumably because with 1/7.0, we're using tclint's multi for int/float division 02:54
Coke checks that bit of code.
hurm. the Multi is just for (Integer, dest0 02:55
hurm. the Multi is just for (Integer, dest)
so that's probably going to Integer.pmc's divide.
src/pmc/integer.pmc:777; anything look dodgy there? 02:57
hurm. setting a breakpoint, that isn't invoked. 02:59
cotto Coke, my breakpoint says it is. 03:01
Coke just noticed I was running the wrong code.
did you pick 1/7.0 because it takes longer to repeat? =-) 03:03
cotto just a bit
Coke Now that we're not storing the data in the union, using gdb to poke at PMC guts sucks.
is the resulting PMC a TclFloat or a Float? 03:05
chromatic It should be easier to cast the PMC_data member to the appropriate storage type. 03:06
cotto It's just a little more typing. 03:07
The correctly calculated value is a Float; the rounded value is a TclFloat. 03:08
p *(Parrot_Float_attributes*)pmc->data 03:09
tab completion helps a little
tewk we really need a set of gdb macros that look at the type id and automatically do the cast and print the object 03:10
Coke how do you get the type of a pmc?
Coke overrides the visit vtable temporarily to help figure it out. 03:11
tewk they all have a typeid, or at least a string constant VTABLE->whoami, VTABLE-type
cotto Coke, p *pmc->vtable->whoami->strstart
Coke cotto: aha. 03:12
you moved the set_string_native out of the tclobject.
er, set_number
originally, when you called that on a tcl it, it did a morph, then assigned.
cotto (until the strings refactor)
Coke *tclint
now, it's not's doing the morph; so you get ints morph. which morphs to a float. which stringifies differently than a tclfloat. 03:13
so if you add a set_number* to tclint that morphs (like the old one on object did), it should work.
I'm afraid this refactor you're doing obviats the whole 'mixin vtables from tclobject' I was doing. 03:14
cotto I suppose ATTR multiple inheritance would un-obviate it? 03:16
cotto checks thesaurus
cotto feels stupid
s/un-obviate/allow/
Coke, it looks like that suggestion fixes the simple test case. 03:21
Coke I hate pmc2c; set_number_native isn't getting translated into /ANYTHING/ in the corresonding c file, it's getting dropped on the floor.
cotto running make test now
Coke cotto: for parity, should also override set_number* for TclString 03:22
cotto++
cotto makes sense
Coke, the VTABLE function works fine in TclObject 03:24
Coke ;you mean, putting it back there? (didn't your original patch remove it?) 03:25
(when I said mine was getting dropped on the floor, I mean a local mod I was trying to add just now. 03:26
nopaste "cotto" at 96.26.202.243 pasted "likely tcl float fix" (16 lines) at nopaste.snit.ch/15432
cotto running make test...
Coke cotto: ok. (we have the type id of tclfloat lying around, I'll sub that in.
cotto of course 03:27
I had copy/pasted from when I put it in TclInt
Coke, do you have something like spectest_regression for tcl? 03:28
Coke cotto: no; I have no way atm to disable particular tests inside a test file. 03:30
(for the tcl spec tests)
but if you run one of the test files, you could compare the results against the output in docs/*current*
"tools/spectcl mathop" - mathop.test was the one with the 8 new failures.
cotto running... 03:32
Coke, you might also change TclObject's set_integer_native to use VTABLE_set_integer_native instead of PMC_int_val. 03:33
Coke mmm, avoid revisiting this in the future. =-) 03:34
cotto That's what all of this debugging boils down to.
Coke changed, retesting core tests... 03:35
cotto What version of tcl are you targeting? 03:36
03:39 TiMBuS joined
Coke 8.5.6 03:40
dmajnem2 is there any procedure to debugging a runcore? 03:41
Coke cotto: if I change set_integer_native to use VTABLE_set_integer_native(INTERP, SELF, value); 03:42
instead of:
PMC_int_val(SELF) = value; 03:43
I fail tragically. "prove t/cmd_list.t" just dies, e.g.
cotto nuts 03:50
I'll have to play around with that.
Coke reverted to a fix that includes the float stuff though, retesting to make sure I didn't hose it.
cotto++ 03:51
04:05 tetragon joined
dmajnem2 I have a build failure using --m=32 on a 64 bit machine using icc 11 as my compiler. Is this unheard of? 04:14
chromatic I don't know if anyone else has tried it. 04:16
nopaste "dmajnem2" at 64.198.244.3 pasted "icc build failure with --m=32 on a 64 bit arch" (21 lines) at nopaste.snit.ch/15433 04:17
dmajnem2 chromatic: what should is the best thing for me to do? 04:18
should I try to triage it myself?
chromatic Yeah, the more information you can get out of the segfault the better. 04:20
dmajnem2 it seems that debugging symbols are not compiled in by default :/ 04:21
chromatic They should be, unless you used --optimize. 04:22
dmajnem2 nope
cotto Coke, it looks like that patch (set_number_native) makes mathop.test pass more tests than in spectest.current (158 sv 151)
Coke yes, because I ran the spec test after we started getting this failure. 04:25
so this fixes that error (which i hadn't verified was coming from this work, but probably was.)
committed. 04:26
always happy to get patches for tcl. =) 04:29
cotto The segfault happens because Parrot_TclObject_set_integer_native calls itself recursively. For some reason TclFloat has Float's set_number_native, so it's fine. 04:46
04:50 ChrisDavaz joined
cotto If I reverse the order of the "extends" lines in TclInt, TclFloat's set_x_native VTABLE functions become infinitely recursive. 04:51
05:12 dalek joined 05:15 dalek joined 05:16 dalek joined 05:17 rg1 joined
s1n can i do % (mod) directly in PIR like you can +? 05:25
TiMBuS i don't think so s1n 05:27
05:28 dalek joined
s1n TiMBuS: what kind of data types do i get in PIR then if i have to compute this manually? 05:28
TiMBuS parrot has a mod opcode, you just cant use it inline like + - etc 05:29
you'd have to use 'mod $I0, $I1'
s1n you have to manually spell mod like that? 05:30
TiMBuS ooor, maybe you can use it inline.
its how i use it
i was wrong though, you can use it inline with the % operator, i just checked the pir pdd 05:31
s1n TiMBuS: i'm trying out your first suggesting now 05:32
error:imcc:syntax error, unexpected PARROT_OP ('mod') 05:33
TiMBuS that odd 05:37
how did you use it
05:39 dmajnem2 joined
TiMBuS s1n, you'd get that error if you weren't putting your ops inside a subroutine block. 05:45
dalek r36043 | petdance++ | trunk/src/gc: 06:07
: one teeny const
review: www.parrotvm.org/svn/parrot/revision?rev=36043
06:12 TimToady joined 06:14 russell_h left 06:28 alvar joined 06:41 HG` joined 06:44 mberends joined 06:45 tetragon joined
dalek r36044 | petdance++ | trunk/languages/perl6/src/pmc (3 files): 06:54
: consting and localizing
review: www.parrotvm.org/svn/parrot/revision?rev=36044
06:59 Gerd joined 07:01 Ademan joined
TiMBuS hey woah, the new make html output is pretty 07:09
cotto apparently I can control the economy (.com) for a mere $175,000 07:10
TiMBuS sounds like a good deal to me 07:11
pay in zimbabwe dollars 07:12
cotto TiMBuS++ #Ironic as it is, I don't think they'd accept that.
TiMBuS :> 07:15
07:19 Zaba joined
dalek r36045 | allison++ | trunk (45 files): 07:33
: [pdd28str] Merging in the pdd28str_part1 branch. Large file renaming in strings
: subsystem for greater maintenance sanity.
review: www.parrotvm.org/svn/parrot/revision?rev=36045
r36046 | allison++ | branches: 07:34
: Removing first strings refactor branch from the repository
review: www.parrotvm.org/svn/parrot/revision?rev=36046
07:36 Phurl joined
Phurl hi all 07:36
anyone using pipp?
moritz seen barney
purl barney was last seen on #parrot 18 hours, 29 minutes and 15 seconds ago, saying: A usable PIR lib for filepath manipulation would be nice.
moritz Phurl: barney is the main developer, but it seems he's not around at the moment 07:37
you can always send a mail to the list (parrot-dev@lists.parrot.org) if have some questions 07:38
cotto Phurl, what's your question?
Phurl well, I have it now installed
built tests
and want to know if there is any chance of running anything larger with it
on the joomla/index.php i get : 07:39
Couldn't find constant __FILE__
cotto It's definitely not ready to run any full apps
Phurl pastebin.com/f13342886 07:40
ok
well, I can report bugs if you want
Phurl is a good bug reporter
cotto excellent
Phurl ok
then I will see what I can do 07:41
trac.parrot.org ?
cotto You shouldn't have trouble finding any at this stage. ;)
moritz yes, that one :/
cotto yup. Use Parrot's trac for now.
Phurl :)
Phurl condenses a testcase 07:43
quite simple : define('JPATH_BASE', dirname(__FILE__) ); where __FILE__ is not defined. 07:45
:>
TICKET_CREATE privileges are required to perform this operation
uid:h4ck3rm1k3
cotto Annotations (i.e. filename and line #) were only recently added to Parrot. It'll be a little while before they filter to the HLLs like Pipp. 07:46
Hmm. I thought all registered users could submit tickets.
Phurl ahh 07:47
i had to verify
doh
ok
i will add that to the ticke cotto 07:48
07:56 iblechbot joined
cotto actually now that I think about it, I think they also need to be integrated into PCT before the PCT-based HLLs can use them 07:57
Phurl ok 07:58
i will add that too 07:59
trac.parrot.org/parrot/ticket/232
so my workaround just replace the __file__ with values an continue testing 08:00
is there an #ifdef parrot?
;)
cotto Phurl, you don't need to add what I say to the ticket. Barney follows Parrot too and should know the status of annotations. 08:04
Phurl so you can now add annotations when you with some syntax reference the ticket number in the chat? 08:06
ok, so I got the page to process, /home/mdupont/2009/01/php/parrot/parrot src/common/pipp.pir /home/mdupont/2009/01/conference/code/joomla/index_parrot.php 08:07
cotto you're missing a verb
Phurl but it just emits the php code
cotto I'd be beyond surprised if it worked. It's very much a work in progress. 08:08
Phurl ok
cotto If you want to help, there should be lots of functions in src/common that need implementing. 08:10
They're written in PIR, but it's not too hard apart from being fairly low-level.
Phurl i see
ok, but the ./pipp should in theory do something 08:11
./pipp ./t/php/arithmetics_1.php
ok that works
just not the more complex suff 08:12
well enough
at least I tried! now to try the phc
:)
cotto Are you playing with alternative PHP implementations?
Phurl If I can get something running, then it is possible to port the needed functions
I am just looking at what workds
it would be nice to help parrot along the way of course 08:13
cotto Are you using Parrot from svn?
Phurl yes
i just built it, as indicated in the ticket
trunk
I dont mind helping
but i would like to try phc next
so i can see what works 08:14
./pipp --variant=phc ./t/php/arithmetics_1.php | sh: phc: not found | Creating XML-AST with phc failed 08:26
ahh, it can use phc 08:27
?
08:34 Eevee joined 09:28 barney joined 09:36 masak joined 09:39 tomyan joined 09:41 Ademan_ joined
dalek r36047 | bernhard++ | trunk/examples/pir: 09:58
: [examples] fix a typo
review: www.parrotvm.org/svn/parrot/revision?rev=36047
10:01 kj joined 10:14 Zaba joined
dalek r36048 | bernhard++ | trunk/config/auto/backtrace: 10:52
: [codingstd] svn properties
review: www.parrotvm.org/svn/parrot/revision?rev=36048
r36049 | bernhard++ | trunk/src/pmc (2 files): 11:00
: [codingstd] c_parens.t, trailing_space.t, linelength.t
review: www.parrotvm.org/svn/parrot/revision?rev=36049
r36050 | bernhard++ | trunk/src: 11:03
: Let svn ignore object files in src/string
review: www.parrotvm.org/svn/parrot/revision?rev=36050
r36051 | bernhard++ | trunk: 11:04
: Regenerate MANIFEST.SKIP
review: www.parrotvm.org/svn/parrot/revision?rev=36051
11:05 tomyan joined
dalek r36052 | bernhard++ | trunk (2 files): 11:06
: Let svn ingnore external language pjs
review: www.parrotvm.org/svn/parrot/revision?rev=36052
cotto Hmmm. It appears that check_named reads past the end of an array. 11:11
Sigh. One stupid hack deserves another (and a bug report). 11:21
11:30 pancake joined 11:44 TiMBuS joined 11:46 UltraDM joined 11:55 register joined
TiMBuS anybody know how to get a pmc from a specific namespace from inside a pmc? 12:07
i was using Parrot_find_name_op but i've moved the pmc it was getting to another namespace and now im stuck =/ 12:08
ehh maybe ill just write a sub in pir to do that then call it from the pmc.. 12:09
register I can't build parrot on windows. 12:12
I am working with vs 2008 and strawberry perl
perl Configure.pl returns an error in test_6784.exe 12:13
then after nmake the build process starts and continues for a while
until i get an error on the PGE.pbc file 12:14
it happens that the parrot vm is correctly built
but the make process doesn't terminate correctly
TiMBuS did you try to build optimized? 12:16
register no 12:18
i just ran perl Configure.pl
TiMBuS you might need to make a bug report. 12:19
register i'll write on the mailing list 12:21
TiMBuS oh wait, is the error
Perlbinperl.exe: command not found mingw32-make
register mingw32-make.EXE: *** [PGE.pbc] Error -1073741819 12:22
is this one
TiMBuS it might be related to this? trac.parrot.org/parrot/ticket/114 12:23
since strawberry comes with mingw but you're using visual studio it sounds like this is the issue
register could be 12:25
i'll try to set the make variable
dalek r36053 | cotto++ | trunk (11 files):
: [pmc] update *IntegerArray to use ATTRs
review: www.parrotvm.org/svn/parrot/revision?rev=36053
r36054 | rurban++ | trunk/tools/dev: 12:27
: Fix TT #208 pbc_to_exe --install needs to use the path to the installed libparrot
: - Use rpath_lib instead of rpath_blib on --install
review: www.parrotvm.org/svn/parrot/revision?rev=36054
pancake is there any example or tutorial about how to embed parrot in an application, or how to create a parrot binding for a library? 12:31
12:34 rdice joined 12:35 rurban joined 12:37 jan joined
register this is really funny 12:39
:D
i forced to use nmake by editing MAKE=nmake in the makefile 12:40
the build process went farther than before
but now is looping on nmake docs
TiMBuS =/
take out make docs? 12:41
register done 12:42
different error in different area
:D
well
tonight i will try to build with activestate
and i will update the bug report pointed out bu timbus 12:43
TiMBuS if you use activestate, dont forget to fully remove strawberry perl first 12:44
or it'll probably get worse =/
register sure sure 12:45
rurban TiMBuS: Where is your nmake bug report? 12:48
TiMBuS i didn't make one, it was yours 12:49
rurban I'm just trying now a MSVC6 with nmake build
TiMBuS oh wait, not yours, i got you confused with rblasch
trac.parrot.org/parrot/ticket/114 <- thats the report I found 12:50
rurban I get an undefined opengl_lib error in config/gen/makefiles/root.in 12:51
Otherwise it works fine. The reporter obviously forgot to do make realclean 12:53
12:59 Gerd left
mj41 cotto: r36053 probably broke t/examples/shootout.t - see tt.ro.vutbr.cz/report/pr-Parrot/do?...hat+I+mean 13:07
shorten mj41's url is at xrl.us/bedw9s
rurban matisse misses TAP::Harness::Archive on his BuildBot
What is Matisse Enzer's email? 13:08
cotto mj41: I can't access your server, but it does appear to be quite broken. 13:09
13:11 bkuhn joined
masak rakudo: say 'test' 13:23
polyglotbot OUTPUT[testā¤]
masak rakudo: class A { has $.b; }; while shift [A.new( :b(0) )] -> $a { say $a.b; $a.clone( :b($a.b + 1) ); say $a.b; last; }
polyglotbot OUTPUT[0ā¤1ā¤]
masak submits rakudobug
Coke_z masak;when reading those bug reports, it would help me if it showed the expected output. 13:27
Coke is only casually aware of p6 syn.
masak Coke: acknowledged. 13:28
Coke keep in mind that i'm probably not the one who is going to be /fixing/ the bug, so it doesn't help from that perspective. =-) 13:29
s/probably//
cotto Looks like a Heisenbug. It fails in with prove but works fine when I run the pir manually.
masak Coke: sometimes I think the expected output is evident. sometimes I write a descriptive subject line instead. 13:30
but you're right, I should be explicit. 13:31
Coke cotto: what are you working on now?
cotto my last commit somehow broke t/examples/shootout.t 13:32
Coke we use to break that all the time, when I was your age!
cotto but only if it's run via prove
Coke built character! 13:33
prove uses the harness, no?
perhaps that's passing an extra arg to the parrot executable?
Coke can never remember.
cotto Bah. I neglected to clean out a bunch of code that uses macros that depend on the UnionVal. 13:44
masak Coke: in the above case, a .clone should never change the original object. but it does. 13:45
not in most cases, but in this particular one.
13:49 iblechbot joined 14:18 gryphon joined 14:25 PerlJam joined 14:38 rdice joined 15:00 tetragon joined 15:13 Theory joined 15:30 uniejo joined
Coke urk? in tcl, unsetting the last element of an associative array apparently undefs the array. 15:33
urk? I'm hallucinating. good! 15:35
dalek r36055 | pmichaud++ | trunk/languages/perl6/docs: 15:54
: [rakudo]: spectest-progress.csv update: 295 files, 6546 passing, 0 failing
review: www.parrotvm.org/svn/parrot/revision?rev=36055
Coke fixes one tcl spectest for his break. 15:57
16:01 Andy joined 16:07 galf joined 16:18 dmajnem2 joined
Coke pmichaud: very very slight preference for svn, but only because I don't use git on a regular basis. 16:20
dalek r36056 | julianalbo++ | trunk/src:
: fix a const issue that disallows c++ build
review: www.parrotvm.org/svn/parrot/revision?rev=36056
16:21 Tene joined
jonathan hi all 16:24
jonathan is back from his birthday break
Coke jonathan: happy birthday!
purl for (('to you', 'dear '.shift)[0,0,1,0]) { print "Happy birthday $_" }
pmichaud Coke: noted, thanks. 16:25
Coke of course, since I broke spec test the last time I committed anything, you can safely ignore me. =-) 16:26
jonathan wonders who wanted him, way way back in the backscroll...
Coke might have.
PerlJam pm: you're taking a poll?
Coke has no idea.
PerlJam: check yer mail.
pmichaud PerlJam: yes, see email. Also spreadsheets.google.com/ccc?key=pZ7...0Uf1-ouPHQ
shorten pmichaud's url is at xrl.us/bedxsq
PerlJam pm: I prefer git, but I'm fine with svn. 16:27
16:30 uniejo joined
jonathan OK, whoever wanted me, it's so far back in the backscroll that I don't see it. But I'm sure if it's important you'll ping me again. ;-) 16:31
pmichaud: How's things going? Any Rakudo news?
pmichaud jonathan: not from my end -- had sick kids + wife for this past week, and I wasn't feeling so hot myself so didn't get much done. 16:32
jonathan pmichaud: OK, sounds un-fun. :-( 16:33
Hope everyone is well again soon.
Coke pmichaud: wasn't a stomach bug, wozzit?
pmichaud Coke: no, mostly sniffles and fevers for the others
for me, I just felt... "out of phase"
wasn't able to concentrate well; in such cases it's often best to not be doing code :-) 16:34
Coke pmichaud: like on en.wikipedia.org/wiki/The_Next_Phase ? =-)
pmichaud Coke: yes, something like that. 16:38
Except in my case I was trying to avoid others instead of get their attention. :-)
16:39 uniejo joined
PerlJam pm: are you going to add a "weight" column to the spreadsheet? :) 16:40
pmichaud no.
particle jonathan: how many rakudo spectest failures do you get these days?
pmichaud there is a weight, but it's a hidden factor that will enable me to get the outcome I want. As soon as I know what that outcome is. :-P 16:41
particle checks out a clean parrot
particle has 17 rakudo spectest failures with his NaN/Inf patch atm
Coke particle: down from? 16:42
pmichaud particle: note that it's very unlikely you'll be able to do "svn switch --relocate" and be done with it.
PerlJam pm: why is that?
particle down from 60
pmichaud because rakudo is very unlikely to continue living in the languages/perl6/ directory of parrot.
particle pmichaud: will you rely on installed parrot, then? 16:43
16:43 uniejo joined
PerlJam pm: yeah, but he can still switch that subtree I'm fairly sure. 16:43
pmichaud particle: we'll probably have a local copy of parrot in the rakudo repository.
particle i should be able to switch the subtree to a subtree of the rakudo repo
pmichaud PerlJam: my point is that rakudo may no longer be in languages, period. And that the Configure.pl/Makefile will be different.
Coke pmichaud: if you go with svn, perhaps an svn:externals link tied to a specific revision would be better.
particle the parrot subtree
Coke (just a thought)
jonathan particle: Will check - not built Rakudo since maybe Friday or so. 16:44
particle we previously outlawed svn:externals because it doesn't play well with svk.
Coke particle: that was for parrot.
PerlJam particle: have you ever used git?
particle Coke: ah, right!
purl ah, right! is that documented well enough?
jonathan particle: Building latest now.
pmichaud in other words, regardless of svn or git, I don't think that the process will be "check out a copy of parrot, check out a copy of rakudo into languages/..."
particle perljam: i tried checking something out once with git. i don't know if it's msys, git, my lack of focus, or age creeping up on me, but i can't get a handle on git 16:45
pmichaud the process is more likely to be: "check out a copy of rakudo, run Configure.pl in that directory"
particle pmichaud: and... set this env var to the location of parrot? 16:46
PerlJam particle: okay, so there's a learning curve associated with git that wouldn't be there with svn. That's another important factor for you.
particle because i can set ../..
pmichaud particle: if not using the copy of parrot from the rakudo repository, yes.
particle pm: i thought you expected core rakudo hackers wouldn't use the parrot in rakudo repo 16:47
would instead use parrot head
pmichaud core rakudo hackers who are also hacking on parrot will probably use parrot head
but I think there are only 3 or 4 of us
particle that's me!
pmichaud there will be some "core rakudo hackers" who will work with the latest designated version of parrot.
jonathan Are we still pondering having the "current" copy of Parrot that Rakudo is being developed against duplicated into the Rakudo repo? 16:48
particle ok, so call me a bleadrakudo hacker
pmichaud jonathan: yes, that's still my intent.
jonathan OK
I would likely just use that.
pmichaud either a copy, or the parrot subversion number to be checked out.
PerlJam the intersectiong of "core rakudo hackers" with "core parrot hackers" will have a slightly more complex time of things if one of parrot or rakudo moves to git. 16:49
s/ng/n/
particle i like subversion number, because then i can change from 34567 to HEAD
jonathan That way, Parrot breakage doesn't impact Rakudo hacking so much.
particle: Yeah, good point.
pmichaud the point being that a typical person obtaining rakudo shouldn't have to worry about the version of parrot
jonathan Agree.
particle if i *need* git to work on rakudo, i'll get it, and figure out how to use it. 16:50
pmichaud we'll undoubtedly publish instructions.
particle i really just want the distraction to be over with quickly.
16:51 gryphon joined
Coke pmichaud: (version of parrot) If you really want to enable casual usage, I'd recommend having a "stable" version of rakudo that targets the last released version of parrot. 16:51
particle let me apologize to everyone here if i come off as cranky, i haven't slept well.
last released isn't new enough, frequently. 16:52
Coke then you can have a stable branch of rakudo that's guaranteed to work against that version of parrot, and a bleed version that works against trunk.
particle but, a tagged version or revision number, yeah.
Coke particle: define "new enough". you don't need to have all your recent progress avaialable on stable.
particle coke: note that some rakudo progress relies on pct progress
Coke particle: so?
purl Yeah, so?
pmichaud last released version of parrot will often not be new enough for rakudo. 16:53
Coke if you want to track a months worth of progress, yes.
why not have a stable version that doesn't get all the new features right away?
pmichaud because people currently expect the new features
Coke <shrug> then they need to use trunk.
pmichaud no, not necessarily.
Coke Ok. That's my 2c. 16:54
pmichaud sometimes parrot trunk breaks rakudo. This is more likely to happen when the repos split.
I don't want rakudo progress to halt just because parrot trunk changed.
Coke pmichaud: I know the deal. I've already done this.
pmichaud so, we'll track along behind trunk a ways. 16:55
particle ...and you're targeting 0.8.1 because 0.9.0 broke rakudo.
Coke no, I'm targeting 0.8.1 because I'm lazy.
if anyone actually cared about partcl, I'd bother updating to use the latest parrot release for my stable branch.
pmichaud my target audience gives me less room for laziness.
Coke look, I made a suggestion. You don't have to take it. arguing with me isn't going to convince me, and you don't /need/ to convince me. =-)
You've said no. I've moved on. 16:56
pmichaud no problem
I figured we were just exploring. Thanks, opinions noted.
particle coke: i saw a CF/Perl job last week, thought of you.
...and shuddered.
Coke I get that alot.
particle i think it was in atlanta. 16:57
the perl6str get_number inf/nan stuff doesn't seem to be working on windows. i think that's the last of what i need to fix before applying this patch 16:58
pmichaud particle: perl6str simply passes the work off to parrot. 16:59
particle no, it looks for lc(nan) or lc(inf)
i think that's failing. i'll know more soon 17:00
pmichaud sure, it looks for those strings, but the actual setting of nan or inf is done by parrot, iirc.
17:00 gryphon joined
pmichaud and it actually looks for nan/inf regardless of case 17:00
(it's a case-insensitive compare)
particle yes 17:01
sorry, i typed that wrong.
ok, i'm running spectest on a clean parrot/rakudo now, will see how many failures in trunk head
pmichaud: is there a reason not to use tools/test_summary.pl as the spectest make target? 17:02
it's *so* much better than t/harness results
pmichaud particle: people are very used to the traditional harness; I haven't been wanting to push a break from that yet 17:03
moritz particle: IMHO it's a bit too verbose for casual checking
pmichaud well, we could certainly add some switches to make it less verbose :-)
17:03 uniejo joined
particle ah, i run spectest with HARNESS_VERBOSE=1 anyway, so i don't notice that it's noisier 17:03
pmichaud parrotsketch in 84 17:06
particle if i can just get this to work, i'll have fixed inf/nan: is +'Inf' == Inf 17:07
jonathan particle: Sorry, got distracted here - spectest is running now. 17:26
particle jonathan: i got 60 failures in my clean wc
i have one question... should we standardize on Inf/NaN, or inf/nan. 17:27
i may bring it up at #ps 17:28
pmichaud perl5 uses inf/nan
particle perl5 doesn't have a portable inf/nan impl
so, linux uses inf/nan.
17:29 cognominal joined
particle windows native is 1.#INF and -1.#IND 17:29
pmichaud I think that glibc uses inf/nan
particle i was thinking it might be better to use lc, since it doesn't look like a type then
Coke tcl uses Inf
17:30 dmajnem2_ joined
pmichaud Perl 6 appears to use Inf and NaN 17:30
Coke (are you trying to decide for parrot or rakudo?)
dmajnem2_ where is parrot supposed to do strength reduction? At the assembler or the JIT?
particle trying to decide for parrot.
ok, since tcl and perl 6 are caps, i'll go with that. 17:31
there's also the INF/NaN and INF/NAN crowd, but i'll ignore them for now
java is INF/NaN iirc 17:32
Coke wonders how to generate a nan in tcl.
particle 0.0 / 0.0
Inf - Inf 17:33
Coke "divide by zero"
domain error: argument not in valid range
tcl's too smaht for you. =-)
particle a = 0; 0 / a
NotFound āˆž
Coke particle: it's a runtime check too. =-) 17:34
17:35 hercynium joined
jonathan particle: Failed 4/295 test scripts. 58/8358 subtests failed. 17:35
Need details?
particle jonathan: no, thanks, that'll do 17:36
pmichaud the Open Group says that infinity/nan should convert to "inf" and "nan" :-)
particle with my patch, it's:
Failure summary:
S03-operators/smartmatch.rakudo 81 - NaN ~~ Nan is True
pmichaud or "infinity"
particle pmichaud: link? 17:37
purl link is crap
pmichaud www.opengroup.org/onlinepubs/000095...rintf.html
shorten pmichaud's url is at xrl.us/bedx2w
Coke no purl, link is <reply>
purl okay, Coke.
17:37 Theory joined
jonathan particle: Is this a patch not checked in yet? 17:38
particle jonathan: not yet, but very close now. lemme nopaste
jonathan OK, great!
pmichaud particle: if your patch results in only one failure I don't mind regressing.
jonathan will be happy to see these fails go away
pmichaud we can fix that one regression easily enough.
jonathan Aye.
particle pmichaud: should be a straightforward fix 17:39
ah, hrmm...
nopaste "particle" at 76.121.106.245 pasted "portable Inf/NaN" (391 lines) at nopaste.snit.ch/15435 17:42
particle i think there are some spectest updates, with 4 or so newly passing tests
if that could be tested on linux, i'll commit shortly
pmichaud particle: does this only allow "inf" and "nan" as strings? 17:43
i.e., case-sensitive?
particle Inf and NaN
case sensitive, yes
pmichaud case-sensitive?
oooh, that's bad. Most libraries accept case insensitive values. 17:44
particle that can be fixed later, then.
Coke particle: I think that the single statement if blocks are a coding std violation, fyi.
pmichaud at least, most of the ones I've found. Rakudo should continue to be case-insensitive, at any rate.
Coke (can just write them without braces). otherwise looks like a small, applyable patch.
particle coke: if you're talking about spf_render.c, yes, i put braces in there because the else isn't a single statement, and i wanted the paragraph to use consistent formatting. i think pdd07 allows that. 17:45
Coke (is there already an Inf test?)
pmichaud does an installed parrot also install its header files? 17:46
I need to set up a test vm where I can play with make install.
particle i sure hope so
inf test, haven't checked...
Coke patch seems to think your patch is malformed. 17:48
particle it's direct from svn di. hrmm 17:51
Coke missing header for unified diff at line 135 of patch 17:52
can't find file to patch at input line 135
particle ah, it's probably some ^M chars, i'll renopaste 17:54
NotFound pmichaud: you can play with make install using --prefix 17:56
pmichaud NotFound: that's good idea too. 17:57
nopaste "particle" at 76.121.106.245 pasted "portable Inf/NaN (no Ctrl-M)" (314 lines) at nopaste.snit.ch/15436 17:59
NotFound A chrooted environment is another way, but needs more work
PerlJam I little while ago I replied to a Daniel Ruoso on perl6-compiler and that email hasn't shown up. Does that mailing list silently drop email from addresses that aren't in the subscriber list? 18:01
Coke particle: did you ever see my post to the list regarding "why named tests" ? 18:02
(speaking of dropped msgs.) 18:03
pmichaud PerlJam: they get moderated
PerlJam: I've even had my messages get moderated from time to time. it's weird.
18:03 chromatic joined
NotFound Who moderates the moderators? 18:03
chromatic Alan Moore 18:04
particle coke: yes, i did, thanks 18:05
i do understand the desire for named and even unique tests, but damned if i'm going to let it happen! ;)
Coke particle: it's one thing I've always disliked about TAP. 18:07
particle: that version applies. 'make test': 100%
particle yay, thanks! 18:08
Coke (but the potential lack of Inf tests concerns me slightly)
particle i'll write one up, then commit
Coke (i tested on osx/86, btw) 18:09
PerlJam Coke: the main problem with named tests that they require the human to come up with the names.
Coke PerlJam: which is demonstrably not hard. 18:11
in fact, you could code it. =-)
PerlJam heh
Coke chromatic++ # have you seen the fake news footage of doctor manhattan?
chromatic No, I try only to watch movies that won't suck. 18:12
Coke chromatic: www.youtube.com/watch?v=nd5cInmK6LQ
18:12 barney joined
PerlJam chromatic: you won't watch many movies do you? 18:12
18:13 jhorwitz joined
chromatic True. 18:13
18:15 geof joined 18:17 allison joined
Coke chromatic: I am going to have to buy the damn bound edition to re-read. 18:17
Pretty sure I have the original issues still in bags/boards. 18:18
chromatic I re-read it recently. Its structure depends on the comic format.
NotFound xlibtest.perl fails with: Method 'HOW' not found for invocant of class 'Xlib;Display' 18:23
s/perl/p6
dalek r36057 | particle++ | trunk (11 files): 18:24
: [core] portable Inf/NaN support
: ~ adds configure code to set min/max constants for FLOATVAL
: ~ modifies Parrot_sprintf_format and string_to_num to handle two-way string/number conversion
: ~ adds basic Inf/NaN math test
: note: currently parrot accepts 'Inf' and 'NaN' values in N registers, which disagrees with the open group's spec, but agrees with Tcl and Perl 6 specs 18:25
review: www.parrotvm.org/svn/parrot/revision?rev=36057
r36058 | particle++ | trunk/languages/perl6/src/pmc: 18:26
: [rakudo] make Inf/NaN support portable
: +58 tests with msvc
review: www.parrotvm.org/svn/parrot/revision?rev=36058
moritz particle** 18:28
particle++ I meant ;-)
kj ps? 18:32
purl ps is probably postscript or process status or see "parrotsketch" or non-vector?!
PerlJam yeah, it looks like ps is a little late in getting started 18:33
chromatic I was eating toast is why.
moritz how dare you?
particle oh crap, that time already? 18:34
Infinoid Toast has realtime scheduler priority.
dalek r36059 | moritz++ | trunk/languages/perl6/t: 18:37
: [rakudo] add a few more passing tests to t/spectest.data.
: Restore alphabetical ordering.
review: www.parrotvm.org/svn/parrot/revision?rev=36059
Coke src/inter_call.c:1404: warning: nested extern declaration of 'GETATTR_FixedIntegerArray_int_array' 18:38
Coke bets that's cotto.
chromatic jhorwitz, you could make the idx parameter to Parrot_get_vtable_name UINTVAL. 18:42
Coke kj: parrot should never segfault, even on invalid pbc. 18:43
jhorwitz chromatic: easy enough. but conceptually i'm doing the right thing here?
chromatic Yeah, I think so.
Infinoid kj: if pbc is portable to different platforms, I'd love to take a look at the crash under gdb 18:44
chromatic allison might want to look at it too.
I don't think you need PARROT_EXPORT though.
kj Coke: it's hard to tell what's happening in my normal windows shell, as it seems that Windows seg faults are not so spectacular as on linux. but my attempts on moritz++ 's linux box do clearly segfault, indicating memory errors (invalid or whatever, can't remember) 18:45
jhorwitz chromatic: was following along the lines of Parrot_get_vtable_index, which is PARROT_EXPORT. not surprising.... ;-)
allison chromatic: look at what? 18:46
chromatic TT #185
kj Infinoid: the generated PBC can be disassembled using pbc_disassemble, so pbc_dis. thinks it's valid PBC ;-)
chromatic NotFound, I agree with your idea in TT #210.
particle kj: windows will return a negative exit code on "segfault"
18:47 masak joined
kj particle: ah yes, that's what I was planning, write some kind of script to run the program and check the return value. thanks for reminding :-) 18:47
18:47 rurban joined
NotFound chromatic: I know 18:47
Coke_away I'm going offline. anyone has questions for me that are #ps related, email's best. 18:48
NotFound Maybe for the first time I have a question with a short answer X-)
kj Infinoid: shall I send the PIR script that's segfaulting, or do you want the PBC? PIRC builds without any problems on linux. 18:49
Infinoid I don't know how to use PIRC. instructions on how to make it crash are just as good 18:50
kj use pirc: (1) make (2) ./pirc -b <pir file> (3) parrot a.pbc 18:51
the default output file is "a.pbc"
Infinoid Ok, thanks. Is the pir script nopasted somewhere? 18:52
kj will be, soon. (preparing now, double checking)
nopaste? 18:56
purl it has been said that nopaste is at nopaste.snit.ch/ (ask TonyC for new channels) or rafb.net/paste or poundperl.pastebin.com/ or paste.scsys.co.uk/ or App::Nopaste or tools/dev/nopaste.pl or at www.extpaste.com/ or paste.scsys.co.uk (for #catalyst, #dbix-class, #moose and others)
clunker3 pasta.test-smoke.org/ or paste.husk.org/ or nopaste.snit.ch:8001/ or rafb.net/paste or poundperl.pastebin.com/ or paste.scsys.co.uk/
18:58 dmajnem2_ joined
nopaste "kjs" at 193.1.104.5 pasted "PIRC-generated file segfaults" (18 lines) at nopaste.snit.ch/15437 18:59
kj Infinoid: Just pasted it.
19:02 ask_ joined 19:09 dmajnem2_ joined
Phurl i finally got phc to compile 19:11
now i have a new bug report for pipp : trac.parrot.org/parrot/ticket/235
cause the latest version of phc does not work with pipp 19:12
19:18 Casan joined
Casan read that as Mooses 19:19
chromatic A prophet is without honor in his #home.
barney Phurl: I have removed the phc variant from pipp 19:25
s/the phc variant/all variants but pct/ 19:26
kj Allison: hi, I did some simple stuff for languages/pod just now, but no Makefile is generated, so can't really test here. Does it build correctly now? 19:28
Phurl barney: ok should i update?
when was this? 19:29
Infinoid kj: cool, I get a gc segfault from ./pirc -b crashy.pir 19:30
barney Phurl: I removed the variants after the the december release of Parrot 19:31
kj Infinoid: cool? :-P I'm not sure if it's so cool :-/
Infinoid: thanks for looking into it, btw.
nopaste "Infinoid" at 96.238.213.50 pasted "pirc GC backtrace" (58 lines) at nopaste.snit.ch/15440
barney If there is somebody, Phurl?, that want's to maintain it, i could add the PHC variant again 19:32
Tene barney: how complete is pipp?
kj Infinoid: b.t.w., if you run ./pirc with "-b -x" options (so, add -x), then it also executes the generated PBC code, so then you don't need to invoke parrot separately 19:33
barney very incomplete, 36 of 1490 tests passing, not counting the extensions
kj (actually, I'll be changing this so that when you specify -x, it will imply " -b -x")
Phurl barney: ok, lets say this, how can I invoke a removed feature ;) 19:34
i did not imagine it
Infinoid kj: it isn't getting that far... need to track down this gc issue first
barney Phurl: svn checkout the code for the Parrot 0.8.2 relaease and add 'src/phc' to svn head 19:36
Phurl ok, i got the head 19:37
barney then add the variant handling code to pipp.pir again
Phurl not the release
i am confused.
:(
look, how is this 19:38
i will just leave it alone for now
kj Infinoid: maybe that's triggered by pirc. You could check to just create a PIR file that just prints something, that should work fine.
Phurl i am going to test the php app on the phc first
Infinoid kj: it emitted pasm to stdout just fine until I RTFM'd and added -b
Infinoid is trying it under valgrind at the moment. 19:39
barney svn co svn.perl.org/parrot/tags/RELEASE_0_8_2/
kj Infinoid: yes, just printing PASM etc worsk fine. It's only the PBC generation that goes wrong *somewhere* 19:40
Infinoid well, in the case of this crash, it looks like glibc's internal memory allocation structures are being corrupted.
that's not necessarily related to your pbc emission issues...
barney Phurl: drop me a mail, if you want to play with. The transformation from phc output to PAST is done with XSLT 19:41
so XSLT hackers have an advantage
Phurl yes,
i can do all that
xml
xlst
asts
and am looking forward to helping 19:42
but first i am going to test the phc on joomla
and see if I can get the entire thing compiled
into xml
or anything
:)
the beast
barney my only reason for using xslt, was that i wanted to learn xslt 19:43
Phurl yes
chromatic Infinoid, what's pool->object_size and pool->objects_per_alloc in gc_ms_alloc_objects?
Phurl well, its all good
you know in the end. it is all machine code right? 19:44
barney A saner approach would be that phc generates the past and uses an embedded Parrot
In the end it is impossible to distinguish between machine code and magic :=) 19:45
Infinoid chromatic: 112 and 213
purl 325
chromatic Seems sane. 19:48
Phurl barney: please explain
it uses an embeded parrot to run the transformations? 19:49
Infinoid valgrind reports several errors about PackFile_Segment_pack writing to unallocated memory
Phurl that makes sense
NotFound 'src/charset/ascii.h' not found (referenced from 'src/string/charset/ascii.c') at tools/build/headerizer.pl line 505.
Phurl i was thinking about an embedded rdf engine
dalek r36060 | kjs++ | trunk/languages/pod/src/parser (2 files):
: [pod] implement some bits for languages/pod. Can't test, as no Makefile is generated on my box (windows; should it?)
: Needs more work, but it's a start.
review: www.parrotvm.org/svn/parrot/revision?rev=36060
Infinoid err, not unallocated, but already-freed
Phurl that is used to glue/transform models
basically map between two models
kj Infinoid: mmm, interesting.
Phurl using rdf model transformations based on logic 19:50
kj Infinoid: when you sick of it, could you post your results to parrot-dev list?
Phurl that are well defined
you can optimize them out most of the time
unify the models
Infinoid hmm, this looks like a possible 64-bit specific issue. 19:51
kj well it's certainly a 32 bit issue as well :-)
dalek r36061 | allison++ | trunk/config/gen:
: [pod] Add pod to the list of languages whose makefile is generated during
: configure.
review: www.parrotvm.org/svn/parrot/revision?rev=36061
Infinoid Invalid write of size 8 ... Address 0x83E5CA0 is 48 bytes inside a block of size 52 alloc'd
allison kj: I added pod to the configure process for generating language makefiles 19:52
kj allison: great
allison: not sure if i'm on the right track, but I think so :-)
Coke (tene, exceptions) see if you can fix rt.perl.org/rt3//Public/Bug/Display...l?id=57088 while you're in there. =-) 19:53
nopaste "infinoid" at 96.238.213.50 pasted "valgrind errors from "./pirc -b ../../crashy.pir"" (189 lines) at nopaste.snit.ch/15441
Infinoid oh, he left. 19:54
NotFound allison: 'src/charset/ascii.h' must be 'src/string/charset/ascii.h' ?
allison NotFound: yes
Tene Coke: can tcl generate pir yet?
Coke chromatic: I can deal with the release manager question. we have volunteers.
(svn move) I've got a note on parrot.org/downloads. 19:55
(have since the first time we tried the move.)
tene: yet? 19:57
it used to. it doesn't now.
Tene ah.
s/yet/again yet/
barney Phurl: As of Parrot 0.8.2 PHC worked like: .i let phc export XML ii. generated a nqp-script with xslt iii. run the nqp script and get a PAST data structure iv. execute the PAST
Coke and I'm not going to add it back in by hand. PCT or nothing.
I can provide you with a small example to duplicate that problem, if that would help. 19:58
Tene Dunno yet. It might. I can try to look at it this week, though.
barney Phurl: generation PAST within phc would simplify things. Embedding Parrot would remove the need for temporary files 19:59
Phurl ok 20:00
Coke oh, right, someone broke the build.
src/inter_call.c:1404: warning: nested extern declaration of 'GETATTR_FixedIntegerArray_int_array'
Phurl barney: sounds good.
so we just need a tool to transform into past basically given the xlst 20:01
so xlst -> parrot
and then run that on a parrot machine
really, what i am saying is : we can optimize these steps out
so that yes, the phc would emit PASTS into parrot 20:02
i need to look into this all.
but if phc can generate c,
then parrot could generate c as well at some point?
Phurl needs to look at all this good stuff first 20:03
barney Phurl: Yes, one could write a compiler, using the Parrot Compiler Toolkit, that generates C 20:04
just like pmichaud did for generating lolcode 20:05
Phurl ok
well, I would like to see in the end
really
all this be integrated into the gcc
cause the gcc has tons of optimizing
and that the llvm stuff would work with parrot etc 20:06
ok
enough dreaming
barney yes, compiling the gcc intermediate code to Parrot Bytecode would be a nice project 20:07
dalek r36062 | coke++ | trunk/config/gen/makefiles: 20:10
: document build dependency added in r36053; allow 'make -j' to work again.
review: www.parrotvm.org/svn/parrot/revision?rev=36062
r36063 | julianalbo++ | trunk (19 files):
: [cage] update headerizer info in charset and encoding files and update headerizing
review: www.parrotvm.org/svn/parrot/revision?rev=36063
20:11 Zaba joined
masak ok, I'm stumped. languages/perl6/src/classes/Grammar.pir:65 looks right to me, and sufficiently similar to languages/perl6/src/classes/Code.pir:66 -- how then is it that rt.perl.org/rt3/Ticket/Display.html?id=62704 still happens? I'd like to fix this, because it's holding a November branch hostage. 20:11
barney Phurl: are you coming to the German Perl Workshop ? 20:19
PerlJam looks like the gits are taking over rakudo ;) 20:22
Phurl barney: jup 20:24
i am a frankfurter
and frankfurt.pm. dem be my Homies
well, better said, 20:25
moritz will also come
Phurl i am a naturalized frankfurter
great
anyone here speaking?
moritz PerlJam: we haven't heard of jonathan, who might be a blocker for git
moritz
Phurl cool!
Coke jonathan? he's just this guy, you know?
moritz Phurl: I have a lightning talk about Perl 6 regex
Phurl nice i will look you up 20:26
masak I talked to jonathan yesterday, and we talked about git.
but I don't presume to represent his opinion in any way.
barney I hope to make it, from Munich, but I'm starting a new, freelance project on thursday, so I might be busy
PerlJam masak: does he seem pro-git in your estimation?
masak PerlJam: he's not actively for it, no. he seems slightly pro-svn due to his familiarity with it.
he said he's not a big brancher anyway. 20:27
PerlJam I figured he and particle would skew things towards svn. 20:28
masak I tried to tell him that he didn't like branching because he hadn't tested git or equiv, but he remaind understandably skeptical :P 20:29
particle it's like the fate of the universe hangs in the balance. ;)
moritz I'm sure we could make a good thriller out of it ;-)
Infinoid I'll wait for the scifi serial version. 20:30
PerlJam particle: if I were in Pm's shoes, I would stick with svn because those people that want to use git can and those people that don't want to use git won't have to suffer (especially when those people are frequent contributers)
But since I don't have to make the decision, I'm voting for git and hoping that pm is slightly insane :) 20:31
particle pj: i see it the same way. rakudo isn't going to switch to a linux-kernel-style dev model anytime soon
Infinoid Though rakudo certainly wouldn't mind receiving hundreds of incoming patches per day :)
PerlJam Infinoid: it would if they were made by millions of monkeys! 20:32
Tene It's probably a good idea to weight preferences by contributions
I have no objections to jonathan and pmichaud having a much larger vote than I do.
NotFound Enlarge your vote! 20:34
PerlJam particle: but ... you could switch to git anyway and suffer the pain of changing things all at once :) 20:35
particle i need to learn git sometime, for sure.
PerlJam (git is a good pain. It builds character. It puts hair on your chest. etc. ;)
chromatic I think you're nuts to change version control systems before a major release. 20:36
Infinoid I'm pretty sure I pulled out some hair, during my first couple days of learning git.
chromatic Of course, I think you're nuts to switch repositories before a major release.
PerlJam chromatic: come on! that's just sanity talking. 20:37
chromatic Yeah, you'd think I know something about project management!
Infinoid My first rule of project management: don't get anywhere near it. 20:38
PerlJam chromatic: but then again ... with the current release schedules, when *aren't* we just before a "major" release?
chromatic Some majors (1.0) are more major than other majors.
moritz chromatic: I think pm didn't *want* to switch before parrot 1.0...
PerlJam chromatic: or would you "burn a release" just for the switch?
Phurl my shameless plug : groups.google.com/group/free-softwa...onference/ you are all invited to come to the new free software / open source / wikimedia / voip conference we are organising this year! please spread the word 20:39
chromatic I'm just saying that the risk of a major frak-up before a major release is greater, in my mind, than the value for any developer so immature that he flips the bozo bit on Parrot because its Subversion repository has "perl" in the URL at the 1.0 release.
PerlJam putting parrot where it's going to live "forever" before announcing 1.0 seems like a good idea to me. I don't know about doing it just before the 1.0 announcement though.
chromatic 'cuz honestly, I don't care about patches from whiny brats. 20:40
Tene "just before" is rather better than "just after", fwiw
20:40 Zaba joined
moritz chromatic++ 20:40
PerlJam Tene: indeed.
Coke chromatic: I am under the impression it is as close to a marketing decision as we've ever made.
PerlJam woo! go marketing. 20:41
chromatic I have no problem with marketing, as long as we make marketing decisions for adults.
moritz oh, speaking of which... is parrot.org safe for being slashdotted?
Coke chromatic: I'm not sure who was driving the switch. I'm guessing allison.
moritz because I can very well imagine we'll make it into the /. headlines with a 1.0 release
PerlJam chromatic: sorry, but you're not marketing *to* adults.
chromatic "Hi there, I'm interested in Cardinal! Parrot looks great! OH YOU HOST ON PERL.ORG SERVERS YOU SUCK I HATE YOU I AM GOING TO GO WATCH THE HILLS AFTERSHOW NOW INSTEAD!"
I try to keep wankery *out* of my project management decisions. Personally. 20:42
Infinoid ...and start a 100 page flame war on a random (unrelated) forum along the way
chromatic 'cuz if something goes wrong with the switch and the actual contributors we have who actually contribute are unable to make actual contributions, well, that's a bigger problem for me than the pre-adolescent Ruby wanker. 20:43
PerlJam chromatic: it's a bigger problem ... that will be fixed before anything else. If there's a problem, better for it to happen per-1.0 than post-1.0 20:44
s/per/pre/
Infinoid Okay. Is the old repository getting the guillotine regardless of whether the new one works, then?
PerlJam anyway, the die has been cast, has it not? we're not going to affect anything by babbling about it.
Infinoid I'd say if something goes wrong, then strategy is reformed but life goes on...
chromatic Or we could avoid the "It's a bigger problem" problem totally. 20:45
dalek r36064 | julianalbo++ | trunk/t/op:
: [cage] spourious char in test
review: www.parrotvm.org/svn/parrot/revision?rev=36064
chromatic You know, make the 1.0 tarball.
Turn off the old repo.
Switch to the new repo.
Make the release announcement.
Take a day off to switch repos, if necessary. 20:46
Infinoid And in the meantime, the ticketing system can accumulate patches, no problem.
chromatic Exactly. 20:47
Infinoid You know, through this whole discussion, I've been having a really difficult problem understanding why this is as big a deal as everyone seems to be making it...
s/problem/time/ 20:48
chromatic I don't want to frak up 1.0.
Infinoid That makes sense. But I don't really see how that could happen. 20:49
Lets say things go horribly wrong and the new repo is unusable. So... keep using the old one?
chromatic In theory, we shouldn't be dealing with the fallout of switching ticketing systems several weeks after the switch either.
Infinoid I guess what I'm saying is we aren't working without a net. 20:52
The old repo is still there, in case everything goes wrong...
chromatic Sure, and the old ticket system is still there doing things that the new ticket system still doesn't do.
moritz so is rt
Infinoid The switch to trac is a Bigger Deal than switching to a new svn server, in my opinion. This is just a hostname switch 20:53
chromatic Hostname, accounts, permissions, checkouts....
smoke testers.... 20:54
Coverity....
purl coverity is a commercial tool for Automated Error Prevention and Source Code analysis, See, www.coverity.com/main.html or it has been used to measure the quality of the LAMP stack and other major source projects
chromatic syndication feeds and bots.... 20:55
Hands up everyone who wants me babysitting those instead of fixing bugs? 20:56
Hands up everyone who wants allison, pmichaud, jonathan, NotFound, etc... babysitting those instead of...?
Infinoid *chirping crickets*
chromatic That's what I mean about risk.
Infinoid Hmm, ok. Thanks for the list of things to fix, by the way. 20:57
chromatic You're welcome. 20:58
jonathan hi folks, back
chromatic I wish we'd had this discussion earlier instead of hearing "The repository shall move!" fiat speech.
jonathan re git - I'm only going to whine lots if we move to it *and* it gets in my way lots. :-)
I think there's enough people here who do use it, that if I need help I can get it. 20:59
NotFound Regarding marketing: if you have fear that some people can think we are too perl oriented you just need to show them some of the pirric examples X-)
jonathan So I'm certainly not against. But my lazy/liking familiarity/we have it now and it works-ish side doesn't mind svn at all either.
Tene NotFound: where are pirric examples in the repo? 21:00
Infinoid NotFound: "And look at the nifty things we've written in lolcode!" "And here's some great befunge stuff!" "Wait, where did he go?"
NotFound Tene: t/examples/pir/*.bas
Ops, forget the t/ 21:01
dalek r36065 | julianalbo++ | trunk (6 files): 21:05
: Implement Parrot_str_repeat and replace usages of string_repeat, TT # 210
review: www.parrotvm.org/svn/parrot/revision?rev=36065
chromatic Look out, jonathan -- masak is trolling you on use.perl.org! 21:22
masak :)
21:25 kj joined
Coke chromatic: rt.perl.org/rt3/Ticket/Display.html?id=57088 now has a pir-only version of the error. 21:34
chromatic Excellent. 21:35
dalek r36066 | fperrad++ | trunk/tools/dev: 21:52
: [tools] add the generation of 'Configure.pl'
: for new language
review: www.parrotvm.org/svn/parrot/revision?rev=36066
kj does anybody know whether the :vtable-implies-self issue has been resolved? 21:58
NotFound Self-resolution? X-) 22:09
kj it seems to be fixed..
dalek r36067 | kjs++ | trunk/compilers/pirc/src: 22:10
: [pirc] fix adding of "self" parameter with :vtable and :method flags. If either of these has been set, then don't add "self" again.
review: www.parrotvm.org/svn/parrot/revision?rev=36067
Coke kj: it's not fixed. 22:11
kj the example in the ticket works
Coke show me?
er, which ticket?
appears broken for me in Revision: 36059
kj the RT ticket, 47674
nopaste? 22:12
purl i heard nopaste was at nopaste.snit.ch/ (ask TonyC for new channels) or rafb.net/paste or poundperl.pastebin.com/ or paste.scsys.co.uk/ or App::Nopaste or tools/dev/nopaste.pl or at www.extpaste.com/ or paste.scsys.co.uk (for #catalyst, #dbix-class, #moose and others)
clunker3 pasta.test-smoke.org/ or paste.husk.org/ or nopaste.snit.ch:8001/ or rafb.net/paste or poundperl.pastebin.com/ or paste.scsys.co.uk/
Coke ah, slightly different version of the bug.
kj if only my browser would start up I'd show 22:13
ah there it is. 1 min.
Coke If you don't reference self in the vtable def, the sub definition is created with the wrong # of params.
if you /do/ reference self, it generally works.
kj oooooooh
Coke so if I override does with a simple ".return(0)", and then invoke does, it will fail.
but if I have .return(0)\\n$P0 = self, that works. (even though I never actually use self.) 22:14
kj eh, i have a sub :vtable, that only returns (42) 22:15
and that works
Coke for which vtable?
does?
nopaste "kjs" at 193.120.116.183 pasted ":vtable + self seems to work..." (31 lines) at nopaste.snit.ch/15443 22:16
kj get_integer
purl get_integer is Perl context only
kj purl, forget get_integer
purl kj: I forgot get_integer
dalek r36068 | fperrad++ | trunk/tools/dev:
: [tools] update the Makefile template
: for language
review: www.parrotvm.org/svn/parrot/revision?rev=36068
nopaste "coke" at 72.228.52.192 pasted "this fails." (16 lines) at nopaste.snit.ch/15444 22:19
Coke a sub with no params doesn't trigger argument handling. 22:20
jonathan Coke: That sounds...broken. :-|
Coke jonathan: yes. it's a fairly simple fix, as I recall; just need to have :vtable automagically setup for self, not only when self is referenced.
(like :method already does.) 22:21
if you add $P0 = self after the return there, it works.
(huh. although it prints 1 instead of 42. =-) 22:22
something must be booleanizing the retval.
PerlJam jonathan: quick! svn or git? :-)
moritz I'd say "yes" to that question ;-)
PerlJam well, his preference still isn't listed in the spreadsheet, so I thought I'd ask 22:23
jonathan PerlJam: A *lot* of people seem to like git. I don't think my laziness is a good reason for not switching. 22:24
PerlJam jonathan: you should still register a preference.
22:25 Theory joined
lathos I think it's probably time I had a good reason to learn git. 22:25
Currently I don't use it, and therefore it must suck.
jonathan PerlJam: At this point, I don't have much of a preference either way.
Enough people seem to rate it, that I'm willing to go with it. And enough people here know it, that I'm happy I can get hints as needed. 22:26
And I'm satisfied we can make a read-only svn frontend if need be, to keep folks happy who just want to quickly grab a checkout.
And don't have git yet (since svn install base is wider). 22:27
PerlJam if you're ambivalent then pm has to weight Coke and particle (and possibly himself) against everyone else. If a change is to be made the odds are looking better for git than I thought :) 22:28
lathos: when ever you do get around to using it, have a look at github and imagine that integrated with CPAN somehow (imagine it in a good way if you can muster it :-) 22:31
kj I might have a solution for the :vtable + self problem 22:32
PerlJam (not in a mandated, you must use this sort of way, but in a this is an option that opens up possibilities for sharing)
particle jonathan: sub with no params was a common idiom in parrot in the old days 22:34
there were so many examples of it in the repo, it wasn't turned on, so parrot would continue to build, run, and pass tests
coke just reminded me of that sad fact
Coke there's a ticket to turn it on. 22:35
I got pretty close at one point.
kj Coke: I'm adding it now. Slow computer. Slow connection. :-(
Coke which? a fix for my bug, or to always enable arg processing? 22:36
kj no, for the 'does' vtable example you gave me
Coke excellent.
kj to always enable arg processing is a different issue
I'd rather not go there at this point in the evening
Coke that will let me rip out a single line of code from partcl. =-)
kj not sure if it works properly, please double check 22:37
Coke is it committed?
kj no, I send it as a patch
i'm updating the ticket in trac now 22:38
done. 22:39
22:39 Limbic_Region joined
Coke rant: trac turns your nice patch into something completely unappliable when I click on it. 22:39
chromatic Yes, but it doesn't have perl.org in the URL! WIN! 22:40
kj someone with a fast pc, please run make test after applying that patch :-) I'd rather do stuff instead of staring at a scrolling screen
Coke (and if I want it applyable, I have to click in chunks.)
kj really?
purl well, really is it bad?
Coke kj: would I lie to you? (aaaaaa yah.)
kj hahah. I really should get internet THROUGH A CABLE 22:41
Coke ah. way down on the bottom. "download in original"
... but I didn't want to download it. i wanted to F(*#$& VIEW IT.
NotFound Coke: doesn't has a link at the bottom...
Yes
chromatic Coke prefers fruit on the bottom.
Coke NotFound: that link should be at the top.
NotFound Coke: buy a bigger screen :P
Coke chromatic: so, yah, I don't pretend to understand the switch here. 22:42
parrot.org over parrotcode.org I get, as it made it easier for people to contribute.
chromatic That's because no one explained it.
Coke it == the other switches?
chromatic Yes.
Coke mm.
kj I wonder why the patch has all sorts of extra semicolons 22:43
must be my bison
Coke moo.
... ew. now I am reminded of working on MOO.
kj what's MOO? 22:44
purl cranks up the chainsaw.
Coke kj: e.g. en.wikipedia.org/wiki/LambdaMOO
chromatic Cold Fusion for Cows
particle kj: i've got 22mb/s up 2.6mb/s down on my cable now :)
strike that, reverse it.
chromatic Wow, git pull is slow.
Coke there's a project for someone who is bored. port MOO to parrot.
chromatic: are you gitpulling all of the repo? 22:45
PerlJam Coke: dude, bored people should work on rakudo :)
Coke fair enough,.
kj particle: aah yeeezzz. i'm on a 3.6Mbit "3G" connection. Too bad it switches back to HSPDA all the time
chromatic I'm trying to update my bleadperl checkout.
Coke btw, I have two different people interested in contributing to APL.
chromatic I know Adin Falkoff; want me to invite him? 22:46
Coke (people who actually are coming at it from an APL background)
kj also bored people?
Tene Coke: great!
Coke should fork APL off to googlecode.
particle coke: that's two more that the people interested in contributing to tcl!
Coke particle: you are lucky you're far enough away that I cannot rochambeau you. 22:47
particle what would be the apl symbol to run foreign code?
kj APL is more popular than Tcl?
Infinoid I just did a "git pull" to pull in the last week of parrot changes; it took about 10 seconds.
Is bleadperl really that volatile, or is the server just slow?
chromatic Receiving objects: 43% (105784/244821), 45.49 MiB | 89 KiB/s
Infinoid shiny. big *and* slow. 22:48
Coke chromatic: if one were interested in contributing to perl5, is there a shiny web page?
chromatic perldoc.perl.org/perlhack.html perhaps; haven't tried it.
kj Coke: are you testing the patch I sent?
Coke kj: no. I cannot easily apply it here. 22:49
will do so later this evening.
kj oh ok
okido
Coke anyone care to opine on a clever name for "APL on parrot" ?
NotFound Apalot
PerlJam heh
NotFound A-palot in pseudo-chinese X-) 22:50
PerlJam I was thinking something like apple-rot myself
chromatic Dirty Bird
lathos Urgh. Rakudo is doing something weird. My loop variable is being rewritten half way through the loop.
Infinoid prap
dalek r36069 | fperrad++ | trunk/lib/Parrot/Docs/Section: 22:51
: [docs] add language POD
review: www.parrotvm.org/svn/parrot/revision?rev=36069
PerlJam lathos: show the code
Infinoid hmm. I'm thinking about parrots and apples, but not much is coming to mind.
NotFound iAPL X-) 22:52
kj chromatic: can you spare a minute to test that :vtable/self patch?
Infinoid hah, iParrot
chromatic Oh, now it has to RESOLVE the deltas.
Yes, this is much simpler and faster.
moritz HAIparrot? ;-)
Coke chromatic: I wonder sometimes what you'd do if life were sane.
chromatic kj, I can in a few minutes.
NotFound I dream with inventing a thing called 'Diot' and expect Apple do his version 22:53
lathos PerlJam: I think it's just the usual assignment versus copy oddness.
kj chromatic: great. don't want to interrupt too much here
Coke chromatic: ah well, :
"Because you cannot use the Perforce client, you cannot easily generate diffs against the repository"
chromatic Coke, it's out of date though.
You can use git now.
NotFound BQM
chromatic If you have a couple of hours to check out the sources.
(My repo's about a month out of date.)
lathos If I say "for (0.. $l-1) -> $char { $savechar = $char; wibble($char); ... }" then $savechar goes back to 0. If I say "$savechar = $char + 0" it's fine. 22:54
Coke if I do "git clone git://perl5.git.perl.org/perl.git", where does it put the clone?
Tene ./perl
PerlJam Coke: in a dir called perl 22:55
Coke danke.
kj lathos: can you nopaste the generated PIR for that?
Tene Coke: you can add the desired destination as an additional argument
lathos kj: This is a fairly big program. 22:56
NotFound www.traveljournals.net/explore/phil...palot.html
shorten NotFound's url is at xrl.us/bedzff
lathos But hey, why not try it yourself? It's in the repo.
kj lathos: what program is it? 22:57
chromatic ARGH, and the merge failed, due to a file I have not even touched.
dalek r36070 | simon++ | branches/strings/pseudocode (2 files):
: First stab at char indexing of NFG strings.
review: www.parrotvm.org/svn/parrot/revision?rev=36070
lathos kj: strings/pseudocode/ParrotString.pm
kj oh right, that's in the branch...
lathos Yes.
You can just check out the pseudocode dir. 22:58
kj good point
lathos vn co svn.perl.org/parrot/branches/strings/pseudocode string-prototype
^s
kj besides a slow pc and a slow connection, I'm also low on disk space :-P
lathos Well, I've hacked around it, so it's no big deal to me. You just sounded interested. :)
kj well, it just seems to me that there's somethign wrong with the generated PIR 22:59
Infinoid chromatic: "couple of hours"? I just did a clone of git://perl5.git.perl.org/perl.git, took 3 minutes 24 seconds. 23:00
chromatic Apparently I'm not using the same git. I'm using something called git-haunted-will-eat-your-soul, which is almost entirely unlike the Second Coming of Obama git that the rest of you use.
Infinoid (I'm not trying to be argumentative, I'm just not seeing the same thing you are.) 23:01
chromatic Yeah, that's pretty much my experience with git.
Sort of like watching Star Wars for the first time as an adult. 23:02
Infinoid heh, that's accurate.
dalek r36071 | particle++ | trunk/src:
: [core] remove commented out code, thanks for noticing nick!
review: www.parrotvm.org/svn/parrot/revision?rev=36071
Infinoid My first few experiences with git were along the lines of "What the hell does that big ugly command mean?" and "Why would anyone be crazy enough to use this stuff?" 23:03
I warmed up to it, but it took a while.
dalek r36072 | particle++ | trunk/config/init/hints:
: [config] add hint for icc, to induce correct floating point behavior instead of broken default behavior
review: www.parrotvm.org/svn/parrot/revision?rev=36072
chromatic My experience is "Why isn't it checking out?" 23:04
"Why isn't it merging?" 23:05
"Why are there conflicts in files I haven't edited?"
"Why won't it update?"
Tene I seem to have yet a third version of git. 23:06
Infinoid What's the conflict? (Wondering if maybe they added a file that used to be autogenerated, or something.) 23:07
23:07 solarion joined
Tene If you're getting conflicts when just trying to update, without changing anything, you're doing something very wrong. 23:08
chromatic Yeah, apparently.
particle this all just worked when perl used p4
:P
chromatic I cloned the repository, made some changes, and committed them. Then I tried to update my clone. 23:09
KAPOW
PerlJam chromatic: what version of git?
Tene 'git fetch; git rebase origin', right?
chromatic 1.5.6.3
PerlJam just checking.
chromatic I did a git pull first.
PerlJam older gits did have some weirdness
Tene "git pull" is "git fetch; git merge origin" 23:10
That's not what you want.
chromatic Apparently.
Infinoid (it's fine for local bounce trees, but often bad for development trees.)
NotFound git do what i want
Infinoid infinoid@chirp perl % git do what I mean
git: 'do' is not a git-command. See 'git --help'.
Someone should implement that. 23:11
NotFound Is as silly as purl
git make me a sandwich
chromatic sudo !!
Infinoid I'm guessing I'm the only one here who uses stgit? 23:12
NotFound Star Trek git?
Infinoid stacked git. sorta like quilt 23:13
moritz Infinoid: wasn't stgit's development stalled?
Infinoid I don't have any idea. my version is dated 09 Jun 2008 23:14
moritz then it was probably revived ;-) 23:15
particle anyone else getting 403 from hiveminder.com?
chromatic Okay, so now I have a patch in a git checkout that's completely frakked. 23:16
lathos Hiveminder is working for me albeit very slowly.
Tene chromatic: are you complaining or asking for help? 23:17
chromatic Depends. Is it possible to recover?
Tene yes
Infinoid Should be.
git reset --hard origin
Tene will lose your commit. you could get it from 'git reflog', though. 23:18
Infinoid ah, make that origin/HEAD
particle ping hiveminder.com
Tene you can always run 'git reflog' to go back through every state your repository has been in. 23:19
purl 10 packets transmitted, 0 received, +9 errors, 100% packet loss, time 9046ms
Tene Then 'git checkout 123abc456...' to return to that state.
kj particle: ping hiveminder.com is fine here
particle dns?
purl well, dns is the way to hack the Internet
lathos Same here.
kj Reply from 208.70.245.184: bytes=32 time=517ms TTL=40
Reply from 208.70.245.184: bytes=32 time=504ms TTL=40
Reply from 208.70.245.184: bytes=32 time=544ms TTL=40
Reply from 208.70.245.184: bytes=32 time=1201ms TTL=40
purl 10 packets transmitted, 0 received, +9 errors, 100% packet loss, time 8998ms
particle ok, well purl and i are hosed
Infinoid it's dropping packets here... 30% packet loss 23:20
kj Infinoid: did you find anything when running the buggy pbc file? 23:27
eh, 'crashy' i think it was called 23:28
Infinoid kj: yeah. if I remember correctly, PackFile_Segment_pack was writing to memory that had already been free()d. There was a nopaste, one moment and I'll dig it up. 23:29
kj: Ok, that wasn't quite it. But a couple of buffer handling problems in general... nopaste.snit.ch/15441 23:31
23:32 TiMBuS joined
chromatic kj, I'm rebuilding and testing with self.patch right now. 23:33
Infinoid I was going to play around with pirc some more this evening.
23:33 japhb joined
kj chromatic: great, thanks. There's some tests failing, but they should, I guess, given that that's what the patch does 23:33
Infinoid: thanks, this is useful, I think
(these reports) 23:34
Infinoid kj: if nothing else, it looks like create_codesegment() isn't allocating buffers big enough
But this is all happening *way* before the problems you were describing. This is crashing while creating the pbc file, I haven't even gotten to the point of running parrot on it yet. 23:35
kj well, that's easy to test :-) I'll multiply it a few times
Infinoid I still suspect a 64-bitness issue.
kj but i'm on win32
and on linux, 32 bit it fails too
Infinoid right, and you're seeing a different crash than I am. Right? 23:36
kj well on windows I don't seee anything, except that the expected outcome doesn't come
and on linux it's memory error
Infinoid that's when generating the .pbc, or running it?
kj generating is fine
Infinoid Not here.
kj running is not ok 23:37
oh, really
oh ok, that's an issue too then
:-(
Infinoid Yeah. I'm just not sure it's the same one :(
kj well at least just emitting pasm works :-) 23:38
so the core is ok
Infinoid I'll see what I can do about fixing the issue I'm seeing, and hopefully that'll get me up to the point where you're seeing parrot crash. 23:39
kj ok, thanks a bunch
Infinoid heh, nice comment. /* allocate enough space. XXX I *think* bytes is /always/ codesize * 4. */ 23:41
kj well, it's true :-)
but multiplying that by 10 didn't help 23:42
Infinoid I'm not so sure. Shouldn't it just be sizeof(opcode_t)?
kj good point. I was just copying stuff from IMCC
that certainly smells like a 64 bit issue!
Infinoid yeah, it is. opcode_t is Parrot_Opcode which is long, which is 8 bytes for me 23:43
kj if sizeof (opcode_t) is 8 bytes, and only 4 is allocated...
Infinoid It might not help your issue, but it might help mine.
Infinoid tries it
kj feel free to commit
codesize*4 calculates the size in btes 23:44
bytes
Infinoid yeah, I replaced 4 with sizeof(opcode_t). Rebuilding now
As a nice benefit, I think I can get rid of your XXX
kj and, to give some background info, codesize is the number of opcode_t's needed in the codesegment
1 for each op, and 1 for each operand 23:45
Infinoid++
Infinoid Makes sense. I ran into a weird issue while writing the Packfile PMCs, where PDD13 didn't actually mention what format the ops were in. 23:46
I had to do a bit of searching to find opcode_t.
kj ah yes. everything's an integer, basically 23:47
in the codesegment
chromatic kj, only two tests failed for me. t/pmc/namespace.t #62 and t/pmc/parrotobject.t #9-11. 23:48
kj yes, for me too
and after inspecting them, I wasn't surprised
are you?
chromatic Not at all. 23:49
kj heh :-)
that's sounds promising then 23:50
Tene there are tests passing with pirc?
chromatic Tene, this is the vtable-implies-self patch. 23:51
Tene ... reading comprehension fail 23:52
Infinoid kj: Ok, I've got a patch to fix the buffer allocation thing. I'll commit once I get a couple of other things sorted out.
(and I verified it works; now it generates pbc and I'm able to see your parrot crash as well) 23:53
kj ah nice!
pretty good, 1 fix to make it 64bit compatible :-)
Infinoid Yeah, better than average :) 23:54
s1n TiMBuS: what do you mean "putting my ops inside a subroutine block", where else will it go? 23:56
TiMBuS oh hey
purl que tal, TiMBuS.
kj chromatic: you think it's ok to commit that vtable thingy? 23:57
TiMBuS well i meant, if you were just doing the op outside any block, the op would be unexpected by the imcc compiler
chromatic kj, if we can make those tests pass.
kj sure, eh. shall i send a patch for those?
then if they pass for you as well, then should be fine, no? 23:58
chromatic If they pass for you, I trust them.
There should be nothing platform-specific about the fixes. 23:59

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!