[Coke] nwc10: I can build and install moar on os x, but then NQP dies horribly trying to use it. 00:00
(it's looking in bin, not lib, for libmoar)
lee__ [Coke]: i was able to get around that by using --static when configuring moar 00:02
though it isn't really fixing the problem
[Coke] anyone working on getting nwc10++ 00:40
's patches applied?
02:23 colomon joined 02:42 lue joined
diakopter [Coke]: I'll look at tonight 05:06
cxreg is moar going to be the first rakudo with non-blocking i/o? 07:01
looks to me like the jvm doesn't support that feature 07:05
moritz there's NIO 07:38
08:31 odc joined
nwc10 FROGGS: is there any easy way to specify optimisation levels such as -Os or -Og using MoarVM's Configure.pl? It seemed that it wanted them to be numbers. 11:22
FROGGS nwc10: let me fix that 11:30
dalek arVM: cda331c | (Tobias Leich)++ | / (2 files):
allow string options for --optimize and --debug
 11:35
12:05 nwc10_ joined 12:06 ingy1 joined
nwc10_ Gah: 12:11
(gdb) p handle
$1 = (uv_handle_t *) 0xf56bad86e0ad88c0
13:23 cognominal joined 14:50 eternaleye joined
FROGGS uhh, 56bad 15:29
that is like really bad
15:34 jnap joined 15:50 jnthn joined 16:03 colomon joined
dalek arVM: 36e958e | nicholas++ | src/core/hll.c:
In MVM_hll_set_config() need to root config_hash, as GC might move it.

The calls to MVM_string_utf8_decode() allocate, which means that they can trigger GC.		 16:56
FROGGS nw++ 16:57
nwc++
hmmm, I should try v5 on moar these days... 16:58
jnthn ooh, yes :) 17:05
jnthn is looking at the other two nwc10++ patches too
jnthn wonders if some of these will get us able to build on the roast machine 17:07
Assumiung that ain't already happening.
dalek arVM: 6cb9a3a | nicholas++ | src/6model/sc.c:
MVM_sc_get_sc() was missing a write barrier.
 17:08
FROGGS ohh, hi jnthn :o)
v5 has a build system since about 30 hours... so it is the right time to try moar (and jvm) 17:10
17:15 colomon joined 17:25 jnap joined
dalek arVM: 9ef1890 | nicholas++ | src/io/fileops.c:
MVM_file_set_encoding() needs to root oshandle

MVM_string_find_encoding() can allocate, and hence can cause a GC run.		 17:33
arVM: 8be359e | jonathan++ | src/strings/ops.c:
Missing MVMROOT, avoid GC-able-on-stack bugs.

Discovered thanks to bug report by nwc10++.
jnthn nwc10: What size are we down to by now? :) 17:35
FROGGS the -O3 && BOOTSTRAP.pm error is still there fwiw 17:39
jnthn ugh :(
FROGGS -O1 still works on my box 17:40
jnthn FROGGS: Is that error still tehre with all the nwc10++ patches I just applied? 17:41
FROGGS I pulled and tried just before telling it
jnthn ok 17:42
I dunno if the nursery size thing is being done with Rakudo and NQP builds or just NQP at the moment...
diakopter .
jnthn FROGGS: What opt level do we default to? 17:43
FROGGS to 3 since a few hours
jnthn Should we not default to something that works? :) 17:44
FROGGS :o) 17:46
dalek arVM: 81c749c | (Tobias Leich)++ | Configure.pl:
default to -O1
 17:47
jnthn If I configure with --debug I get: 17:49
Option debug requires an argument
See --help for further information
But --help tells me...
FROGGS :(
jnthn --debug
--no-debug
Toggle debugging flags during compile and link. Debugging is off by
default.
timotimo yay moarvm commits 17:50
diakopter [Coke]: I guess I won't look at those tonight.. :) 17:54
FROGGS nwc10: do you need to be able to pass a string to --debug too? 17:55
diakopter has moar-support been merged? 17:59
jnthn no
timotimo there's still a segfault when trying to build Test.pm (but possibly after finishing the build ...) 18:12
(and again: --stagestats fixes it) 18:14
but i can make-^H m-install completely when i add --stagestats to all of the steps
`~j
make m-spectest tries to rebuild Test.pm, though ... so unless i edit the Makefile, it won't do the spectest run at all 18:15
18:15 svitich joined
jnthn Time for dinner :) 18:16
maybe bbl &
timotimo bon appetit :)
18:16 woolfy left 18:17 woolfy joined
timotimo whoops, more than one --stagestats is deadly 18:18
18:26 badlena joined
timotimo i do get a bunch of succeeding tests! 18:27
nand some hanging ones of course) 18:34
FROGGS you could fudge them while you're at it 18:38
timotimo oof 18:39
i'll try to look into a null pointer use in radix.t instead
FROGGS good luck :o)
timotimo actually, it would be good if i took a typing break
my wrists have had to endure some pretty bad posture in the last two weeks
FROGGS yeah, I had that too recently 18:42
19:08 arnsholt joined
FROGGS jnthn: we only have one lib_path which get prepended to .moarvm file lookups... but we need two :o( 19:16
nwc10 jnthn: NQP built with a 3K nursery in about 4 hours 19:18
the builds for 2K and 1K are still running
was not building Rakudo due to that nagging handle heisen-SEGV
but the second build passed
and SEGVs in the next step (Test.pm) with what looks like the side effects of a missing write barrier 19:19
(this is for a 64K nursery)
may not get a chance to look at that for 48 hours 19:21
19:44 colomon joined 20:19 arnsholt_ joined
nwc10 2K took 8h25 20:41
1K only started 6h20 ago so probably has a while to run.
FROGGS nwc10: do you need to be able to pass a string to --debug too? 20:44
nwc10 oh, sorry, no, not aware of any debug flags that are strings 20:48
just -g -g1 and -g3 IIRC
FROGGS k 20:50
then I will have a fix for -O defaults..
21:01 colomon joined
[Coke] recent changes have more-borked things on the daily run box, now it's failing in nqp again. 21:10
(for some time it was failing on loading Test.pm)
FROGGS [Coke]: that might be duie to the optimize level... I am about to push a fix 21:11
dalek arVM: ab5dc6c | (Tobias Leich)++ | Configure.pl:
make --optimize=1 and --debug off the default
 21:18
arVM/moar_libpaths: 8f0eef8 | (Tobias Leich)++ | src/ (4 files):
allow several --libpath options

And also searh for libperl6_ops_moar.so in these paths.		 21:49
FROGGS /dev/MoarVM$ perl6-m -e 'say "yay"' 21:50
yay
jnthn: one can invoke it fromm other directories now... please review the moar_libpath branches in moar and rakudo 21:51
21:55 grondilu joined
dalek arVM/moar_libpaths: 4f0df2e | (Tobias Leich)++ | src/core/loadbytecode.c:
forgot this file in "several libpaths" commit
 22:02
jnthn back 22:34
FROGGS: ooh, you did a "make install" thing? 22:35
FROGGS yes
btw, for parrot we install stuff like Perl6::Grammar to install/nqp/lib, but for jvm and moar in install/perl6/lib... shouldn't it go into nqp/lib because it is a nqp module? 22:36
I have trouble using Perl6::Pod and Perl6::Grammar in v5 because nqp-m does not find it :o( 22:38
jnthn FROGGS: Seems reasonable to me...they're to be loaded by the NQP module loader 22:39
FROGGS k 22:40
lue FROGGS++ (must recompile r-m now ☺) 22:43
FROGGS lue: I hope it works out
lue The only other issue I remember having was that the install directory structure didn't totally respect UNIX conventions (i.e. the <installdir>/languages directory), but I'll have to see what's changed first. 22:48
FROGGS: I need to switch from to moar_libpaths in MoarVM and rakudo it seems, right? 22:52
s/from//
FROGGS lue: only if you want to execute perl6-m outside of the rakudo directory 22:53
so, no, not now
lue FROGGS: actually, I do :) Additionally, I should probably update my perl6-p by now, which requires outside-install... 22:54
jnthn Got a fix for the latest nwc10++ bug report on perl6-compiler coming up :) 23:11
dalek arVM: 80a60fc | jonathan++ | / (3 files):
Fix op signature of SC WB ops.

They were declared as handing back obj, but actually gave int64. This could leave invalid pointers around in registers.		 23:20
jnthn To build with that, you must also get the latest NQP I just pushed.
dalek arVM/moar_libpaths: 48814c4 | (Tobias Leich)++ | src/core/loadbytecode.c:
protect orig/result from movement
 23:33
jnthn FROGGS: If you're going to make something non-static, it needs to get a MVM_foo_ prefix 23:57
FROGGS: And a better name than figure_filename, I suspect... :)
FROGGS kk 23:58

Powered by Cro and the Raku® Programming Language.

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

Thank you!