Camelia, the Perl 6 bug

IRC log for #moarvm, 2013-10-24

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:40 BenGoldberg joined #moarvm
01:21 _sri joined #moarvm
07:08 FROGGS joined #moarvm
08:23 donaldh joined #moarvm
08:54 dalek MoarVM/config: 9ca0fd7 | (Tobias Leich)++ | build/Makefile.in:
08:54 dalek MoarVM/config: install moar- and 3rdparty-headers
08:54 dalek MoarVM/config:
08:54 dalek MoarVM/config: It is super verbose but it seems to work. Needs sugaring.
08:54 dalek MoarVM/config: review: https://github.com/MoarVM/MoarVM/commit/9ca0fd7354
11:22 rurban joined #moarvm
11:22 japhb joined #moarvm
12:54 flussence joined #moarvm
13:36 diakopter FROGGS: why 3rdparty headers
13:37 FROGGS diakopter: because these are included by moar.h
13:37 FROGGS and have have no user-friendly moar.h yet
13:37 diakopter oh
13:56 FROGGS I guess we could get rid of installing header for tommath already
13:56 jnthn FROGGS: 'tis verbose but looks like it should work :)
13:57 FROGGS and we'd need to put things like #include "core/threadcontext.h" into an internal header file
13:57 FROGGS jnthn: what are you referring to? my commit message?
13:57 FROGGS or are you already testing it?
13:58 FROGGS hmmm, putting core/threadcontext.h into an moarvm-internal header wouldn't even work, would it?
13:59 FROGGS because when we register an op in rakudo_ops.c or so, we always get/want/need the tc
13:59 jnthn FROGGS: 9ca0fd7
14:00 FROGGS k
14:09 benabik joined #moarvm
14:31 benabik joined #moarvm
14:48 jnthn FROGGS: Is the config branch based off ext?
14:48 FROGGS jnthn: yes
14:49 jnthn ok, which was now merged...but I think it's got the important commits
14:49 jnthn So I won't bother merging master into it
14:50 FROGGS jnthn: could be difficult due to the build-system refactor
14:51 jnthn How so?
14:51 FROGGS k, I've just grepped for the name from the extops markdown, but didnt found it
14:51 jnthn I was talking about config branch in MoarVM...is there one in nqp too?
14:51 moritz if you can tell me which branches I need, I can start to look into the build system for rakudo-moar
14:51 FROGGS jnthn: because I made changes to Makefile-Moar.in for rakudo
14:51 jnthn moritz: Currently, MoarVM master, NQP master, moar-support
14:51 FROGGS just tiny changes, but still
14:52 jnthn moritz: uh, last one is Rakudo branch
14:52 jnthn FROGGS: We didn't unify ConfigureMoar.pl into Rakudo's configure stuff yet.
14:52 FROGGS ahh, alright
14:52 jnthn But given moritz++ is planning to look at it, maybe it's better we merge your changes sooner rather than later.
14:52 jnthn To avoid re-work.
14:53 jnthn FROGGS: So is there an NQP branch I need for trying the config stuff?
14:53 FROGGS yes, config
14:53 FROGGS + creating a stage0
14:53 FROGGS I didn't push one just because of the repo size
14:53 jnthn k
14:54 FROGGS so, it is config/config/config :o)
14:55 jnthn k
14:59 jnthn FROGGS: Oh...the NQP config branch doesn't have the unified configuration stuff in it?
14:59 FROGGS no :/
15:00 jnthn Mind if I do that? Anything I should be aware of?
15:00 jnthn Looks like a trivial merge \o/
15:01 FROGGS go ahead
15:01 FROGGS I really added just three lines or so
15:01 jnthn Merged cleanly; building now with a fresh stage0
15:06 jnthn FROGGS: eek, on Windows it goes and turns the /foo compiler flags into \foo...
15:07 FROGGS huh
15:07 jnthn 'cus of the path thing
15:07 benabik joined #moarvm
15:08 jnthn FROGGS: Hm, it's unhappy...
15:08 jnthn cl : Command line warning D9027 : source file 'dynext/perl6_ops.obj' ignored
15:08 jnthn perl6_ops.c
15:08 jnthn C:/consulting/MoarVM/install/inc​lude/moar\platform/inttypes.h(2) : fatal error C1083: Cannot open include file: 'msinttypes/inttypes.h': No such file or directory
15:09 jnap joined #moarvm
15:10 FROGGS I thought the mix of \ and / would still work
15:10 jnthn The first warning there is fixed with
15:10 jnthn @moar::ccout@$(PERL6_OPS_OBJ)
15:10 jnthn (There used to be a space in there)
15:10 FROGGS ahh, okay
15:10 jnthn The missing header file I'm less sure about...something we simply don't install?
15:12 jnap1 joined #moarvm
15:12 FROGGS aren't these system headers?
15:12 FROGGS #ifdef _MSC_VER
15:12 FROGGS #include <msinttypes/inttypes.h>
15:12 FROGGS ahh, ohh
15:13 FROGGS there is 3rdparty/msinttypes
15:13 jnthn ah
15:13 FROGGS looks like I forgot about this one
15:13 jnthn ok, that'd be the one :)
15:14 jnthn I pushed the other fix...hopefully it doesn't bust it on other platforms
15:14 FROGGS jnthn: can I push to MoarVM?
15:14 jnthn sure
15:15 jnthn to config branch, yes?
15:15 FROGGS yes
15:15 jnthn go ahead, I'll try it
15:15 dalek MoarVM/config: 539d089 | (Tobias Leich)++ | build/Makefile.in:
15:15 dalek MoarVM/config: install msinttypes headers too
15:15 dalek MoarVM/config: review: https://github.com/MoarVM/MoarVM/commit/539d089326
15:16 * jnthn pulls
15:21 jnthn FROGGS: Needed an addition in Rakudo repo too, but now the C file compiles
15:21 jnthn And it fails to link
15:22 jnthn Progress... :)
15:22 jnthn OK, fixed that too
15:22 FROGGS cool
15:22 jnthn Unfortunately, you put the rule first in the makefile so now it *only* builds the dynops :P
15:22 FROGGS *g*
15:23 FROGGS I recognized that too a few minutes ago :o)
15:23 * FROGGS .oO( `make all` is your friend )
15:24 jnthn oh, and it wants the path transform now :/
15:24 FROGGS hmpf
15:26 FROGGS so you have to fiddle $config{prefix} in Configure-Moar.pl and the rest in Makefile-Moar.in=
15:26 FROGGS ?
15:28 FROGGS[mobile] joined #moarvm
15:29 jnthn yeah
15:29 jnthn something else weird is going on
15:30 FROGGS sorry, but I have to move home... see you in a bit
15:31 jnthn oh, WTF
15:31 jnthn We were setting PREFIX twice o.O
15:32 FROGGS[mobile] I think I added one in moarvm's config
15:38 jnthn FROGGS[mobile]: when you get chance, please can you check current state doesn't bust stuff on non-Windows?
15:47 FROGGS[mobile] Will do then
16:10 jnthn FROGGS[mobile]: So, I pushed some stuff in perl6_ops.c; the line commented out is what needs to work :)
16:19 TimToady joined #moarvm
16:27 benabik joined #moarvm
16:28 FROGGS joined #moarvm
17:13 yoleaux joined #moarvm
17:17 yoleaux joined #moarvm
19:01 nwc10 not sure if this is interesting: http://arxiv.org/abs/1310.2300
19:14 arnsholt moritz++ # Unified build
19:16 jnthn nwc10: Looks like it :)
19:16 moritz fwiw I'm now working on --gen-moar in nqp's Configure.pl
19:18 diakopter moritz: you're like build systems extraordinaire lately
19:19 nwc10 jnthn: they test using code from the benchmark game site (or whatever it is called) because they want pure-Python Python 3 code
19:19 nwc10 (which they can also run as Python 2 code on PyPy, with no or minimal modification)
19:19 nwc10 so I've no idea if their "this is looking cool" numbers are only really good on numeric code
19:19 nwc10 or, at least, simple(ish) code using only the builtin types
19:19 moritz diakopter: not really; but now I've done so much with them that it's simply efficient if I tackle the remaining TODOs
19:20 arnsholt Interesting
19:20 nwc10 the latter may be particularly relevant, as it looked like their type-specialisation system used only 3 or 4 types
19:20 arnsholt I've been really happy with PyPy, incidentally. Reduced the runtime of a program by an order of magnitude
19:21 timotimo yes, pypy is amazing, i can't say that often enough
19:22 moritz somehow my nqp-on-moarvm build segfaults now :(
19:25 jnthn FROGGS[mobile]: When building I see:
19:25 jnthn config.c
19:25 jnthn src\gen\config.c(139) : warning C4129: 'P' : unrecognized character escape sequence
19:25 jnthn src\gen\config.c(139) : warning C4129: 'p' : unrecognized character escape sequence
19:26 dalek MoarVM: 104e4eb | (Tobias Leich)++ | / (11 files):
19:26 dalek MoarVM: added bakendconfig op
19:26 dalek MoarVM:
19:26 dalek MoarVM: Which return the hash $config from Configure.pl.
19:26 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/104e4eb539
19:26 jnthn FROGGS[mobile]: Anyways, config merged in MoarVM
19:26 dalek joined #moarvm
19:27 diakopter nwc10: cool; where'd you run across that?
19:30 moritz moar build fail
19:30 moritz I configured with
19:31 moritz /home/moritz/perl5/perlbrew​/perls/perl-5.18.1/bin/perl Configure.pl --prefix="/home/moritz/p6/rakudo/nqp/install" --make-install
19:31 moritz /home/moritz/perl5/perlbrew​/perls/perl-5.18.1/bin/perl -MExtUtils::Command -e mkpath /bin
19:31 moritz /home/moritz/perl5/perlbrew​/perls/perl-5.18.1/bin/perl -MExtUtils::Command -e cp moar /bin
19:31 moritz /home/moritz/perl5/perlbrew​/perls/perl-5.18.1/bin/perl -MExtUtils::Command -e chmod 755 /bin/moar
19:31 moritz Cannot chmod 755 /bin/moar:No such file or directory at /home/moritz/perl5/perlbrew/perls/perl​-5.18.1/lib/5.18.1/ExtUtils/Command.pm line 274.
19:31 moritz make: *** [install] Error 2
19:31 moritz somehow it sets the prefix to /
19:32 moritz ah no, it was a double quoting fail
19:35 diakopter nqp-m: 2
19:35 camelia nqp-moarvm: OUTPUT«(signal SEGV)»
19:39 jnthn FROGGS[mobile]: config merged/pushed in NQP too
19:50 nwc10 diakopter: 5th comment: http://www.blogger.com/comment.g?blogID=39712​02189709462152&amp;postID=7145890443443707910
19:59 jnthn FROGGS[mobile]: and in Rakudo
19:59 dalek joined #moarvm
20:09 diakopter today's like the highest velocity day in a long time
20:10 PerlJam it still seems to be progressing at one second per second to me.
20:18 lue PerlJam: did you take into account gravity variances?
20:23 cognominal joined #moarvm
20:27 dalek MoarVM: 9cd824b | jnthn++ | src/core/ext.h:
20:27 dalek MoarVM: Add MVM_ext_register_extop to public API.
20:27 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/9cd824b1f3
20:28 dagurval joined #moarvm
20:36 jnap1 left #moarvm
20:41 dalek MoarVM: 72699f6 | jnthn++ | / (2 files):
20:41 dalek MoarVM: Fix install of a shared MoarVM build on Win32.
20:41 dalek MoarVM:
20:41 dalek MoarVM: This gets us far as being able to build an NQP with it, and Rakudo no
20:41 dalek MoarVM: worse off.
20:41 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/72699f6179
20:42 woolfy left #moarvm
20:42 woolfy joined #moarvm
20:43 diakopter jnthn++ tons of others++ amazing work
20:49 jnap joined #moarvm
20:56 dalek MoarVM: 9d2b258 | jnthn++ | / (2 files):
20:56 dalek MoarVM: Install moar.dll implib on Windows.
20:56 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/9d2b25840d
21:45 jnthn Is it me, or has LTO suddenly started taking a long time when doing an optimized build with MSVC?
21:45 jnthn Cannot iterate object with MVMString representation # now I improved the error...
21:45 diakopter jnthn: me too
21:46 dalek MoarVM: 83bfa5e | jnthn++ | src/6model/reprs/MVMIter.c:
21:46 dalek MoarVM: Include uniterable REPR in error message.
21:46 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/83bfa5e4a3
21:46 diakopter jnthn: well, past couple weeks anyway
21:57 jnthn argh
21:57 jnthn It looks like if you have an exe and dll of the same name it generates a pdb of the same name
21:57 jnthn And then debugging is hosed... :/
21:57 diakopter jnthn: /Fd
21:58 diakopter jnthn: /Fdother_name.pdb
21:59 jnthn yeah
21:59 jnthn Oh, though I used /pdb:othername
22:00 diakopter I don't see that flag
22:01 jnthn odd, shows up in my link /?
22:02 diakopter oh, link
22:02 diakopter /Fd is for cl
22:04 diakopter jnthn: maybe it needs named libmoar.dll or whatever
22:06 jnthn diakopter: nah, if you /pdb:moar.dll.pdb it picks it up just fine :)
22:06 jnthn I just hacked the makefile though; probably we should do this automatically
22:07 diakopter heh, interesting
22:12 benabik joined #moarvm
22:27 jnthn I'll hunt this problem more tomorrow.
22:27 jnthn I just found one missing thing.
22:27 jnthn (P6opaque handling of auto_viv_container is NYI)
22:27 jnthn Though it doesn't really explain The Bug.
22:35 BenGoldberg joined #moarvm
22:36 diakopter .oO( nothing solves a problem like another layer of misdirection.. )
22:44 TimToady in that case, I pick nothing.
23:12 jnap joined #moarvm
23:37 benabik joined #moarvm

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary