[Coke] updated gist.github.com/coke/8250608 00:03
diakopter I'm trying to run a spectest file individually but I can't because it can't find Test
(I guess because rakudo's nmake install isn't completing) 00:04
(anyone have any tips?) :) 00:17
ergh. 00:29
talk about wheel-spinning
00:42 colomon joined
diakopter how do I fudge things for just the moar backend? 00:48
diakopter updates moar's UCD to 6.3 01:11
dalek arVM: d2bcf56 | diakopter++ | / (3 files):
update UCD to 6.3 and add u1names when needed
 01:15
diakopter [Coke]: that should fix at least one of the unicode things :) 01:16
01:29 colomon joined 01:34 jnap joined
[Coke] will do a rebuild and test some. 01:46
diakopter hopes it doesn't break stuff 01:47
[Coke] moar: 73.40%
(today's run, not just now)
diakopter is that better than it was?
[Coke] slightly. Go to: github.com/coke/perl6-roast-data/b..._rates.csv - put "moar" in the search box. (just above line 1) 01:52
it was 73.33% - slight improvement over a day.
diakopter ooo 01:58
02:03 colomon joined
timotimo diakopter: oh, you added the missing strings to the unicode database? 02:05
diakopter yah 02:06
timotimo will build the stuff 02:07
oh
you checked in a version of the unicode database that adds a \r to every line ending
diakopter gah 02:08
timotimo would you mind re-making that commit with non-wrong line endings so it doesn't show up as 100% removed, 100% added?
diakopter I'm not the right person to do that
since I've never force pushed or modified a commit
timotimo i can do it for you
diakopter but regardless, that file will be almost totally changed anyway 02:09
timotimo is it ok if i show up as the committer and you as the author? 02:10
diakopter yes :P
timotimo | 3 files changed, 17366 insertions(+), 17321 deletions(-)
| 3 files changed, 34432 insertions(+), 34387 deletions(-)
which one is yours, which one is mine? :)
diakopter do the .h also
[Coke] should I kill my test run and wait for the new commit? 02:11
diakopter (plz)
timotimo ERROR: Permission to MoarVM/MoarVM.git denied to timo.
:(
[Coke]: no, it only does whitespace changes
diakopter I emailed jnthn to add you a week ago but he didn't reply yet; he's on vacation from Overlord Inbox ;)
timotimo i suspect i'm only not allowed to force-push 02:12
diakopter oh
timotimo nope 02:13
i also can't push a new branch
diakopter well go ahead and push it; we'll erase the history of that file at some point anyway
we already did it before
timotimo hm. should i push it to a new fork on github? %) 02:15
i already have one
this way you at least have the commit
diakopter why does Moar's Configure.pl tell me --no-optimize is not a valid option but the help output says it is 02:16
02:36 colomon joined 03:28 colomon joined 04:00 colomon joined 09:20 V_S_C joined
FROGGS [Coke]: is this always the one from the last run? feather.perl6.nl/~coke/moar.out 09:23
11:18 V_S_C joined
V_S_C I got moar built on Windows. but not nqp/Parrot. 11:22
Where should I look next to try nqp/moar? 11:23
masak argh! can't we just .gitignore all of 3rdparty/dyncall ? 13:26
hm... that doesn't seem to actually help... :/ 13:27
arnsholt What's the problem? 13:31
We bundle dyncall, so ignoring the whole thing seems a mite excessive =) 13:32
14:13 hoelzro joined, harrow joined 14:19 colomon joined
masak arnsholt: the immediate problem is that it keeps showing up in my status. despite not being dirty from my point of view. 14:20
arnsholt: the reason I can't .gitignore it is that it's a submodule, or something like a submodule. 14:21
14:56 ggoebel118 joined
masak ./moar: error while loading shared libraries: libmoar.so: cannot open shared object file: No such file or directory 15:38
this is from a freshly cloned, Configure.pl'd and make'd MoarVM.
libmoar.so is in the current directory.
ah, oh. 15:39
'make install' fixed it.
I think 'make' alone ought to work, though.
can I submit an issue about this?
arnsholt masak: It's weird that dyncall is weird about being changed or not. IIRC it's just a pile of files included in our repository 15:48
masak 3rdparty/dyncall is a submodule. I just checked. 16:18
probably building it causes it to be changed, though.
modified: 3rdparty/dyncall (untracked content)
meaning there are generated files that aren't properly .gitignored inside of the submodule. 16:19
if I descend into 3rdparty/dyncall and 'git clean -f', the thing doesn't show up in our status anymore.
so probably such a step can be added to the build process. 16:20
dalek arVM: 41db275 | masak++ | .gitignore:
ignore generated install/ directory
 16:22
masak ok, I want to append ' && git clean -f' to line 431 in the Makefile. but the Makefile is generated. do I need to change build/Makefile.in ? 16:40
FROGGS yes 16:41
masak I checked, and the corresponding line in that file (line 425) only says '$(CMD)@dcrule@ $(NOOUT)' 16:42
and I don't know where to find @dcrule@
FROGGS dcrule if in %config in Configure.pl or the build perl 5 module
s/if/is/
it is a key in that hash
masak aha. 16:43
but I don't find it in either of those. odd.
FROGGS hmmm
masak 'git grep' only turns up the occurrence in build/Makefile.in .
FROGGS but it is even put in the generated config.c file... so it must be a key in that hash 16:44
see MoarVM/build/setup.pm:375: rule 16:45
masak oki 16:46
ah.
so... I'll just try adding it there.
FROGGS and the other appearances for other platforms/compiler
masak .oO( OCD-driven development )
oh, indeed. 16:47
I count three locations.
FROGGS three for dyncall, nod 16:49
masak doesn't seem to make a difference :/ 16:52
that is, I changed it in build/setup.pm, but the change doesn't propagate to the generated Makefile
FROGGS you did it for all three? 16:53
masak oh!
it's on L41 too.
that's probably the one used.
yep.
FROGGS ohh, take a look: 16:54
MoarVM/build/auto.pm:15: $defaults->{-thirdparty}->{dc}->{rule} =
/home/froggs/dev/MoarVM/build/auto.pm:33: $defaults->{-thirdparty}->{dc}->{rule} =
masak work'd! \o/ 16:56
[Coke] FROGGS: no, feather is for one offs. the latest "official" is on github/coke/perl6-roast-data 16:57
FROGGS [Coke]: k, thx
dalek arVM: 16deab9 | masak++ | build/setup.pm:
clean up after building dyncall

Otherwise, 3rdparty/dyncall shows up as "changed (untracked content)" in moar's
  'git status'.
masak my janitorial work here is done. 16:58
[Coke] I never did finish a second run last night. internet kept dropping connection, and I was too lazy to do it right. 16:59
today's daily run is already running, will just use that. 17:10
timotimo masak++
that has been very mildly annoying me as well
17:15 V_S_C joined 17:23 colomon joined 17:53 cognominal joined
nwc10 FROGGS: in 428cd37093, doesn't env_str also need to be set to NULL ? 17:54
jnthn: There is still some sort of mark bug as I'm getting tc->cur_frame->caller->caller->params.args[1].o ending up in fromspace 17:59
and tc->cur_frame->caller->caller->cur_args_callsite is non-NULL 18:00
this is somewhere in line 148 of 99-test-basic.t in Rakudo, which is 18:01
is_deeply $deeply, $deeply;
there is a very long chain of callers. I gave up pasting ->caller in gdb trying to work out how deep it was
[Coke] r: say 22311 - 20883 ; # moar improvement 19:01
camelia rakudo-parrot e5268b, rakudo-jvm e5268b: OUTPUTĀ«1428ā¤Ā»
19:29 moritz_ joined 19:31 chipdude joined 19:32 cxreg2 joined
diakopter \o/ 19:38
dalek arVM: 43c8856 | (Tobias Leich)++ | src/io/procops.c:
set env_str to NULL because we root it, nwc10++
 19:43
19:53 harrow` joined 20:02 colomon joined, tadzik1 joined, [Coke]_ joined
timotimo what does that fix, FROGGS? :) 20:20
FROGGS a potential wrong rooting I think
timotimo oh, just potential? :(
i'm longing for more implemented stuff on moar :)
but i suppose i'll have to wait for jnthn to make the continuation passing tsuff work 20:21
FROGGS the question is: what could we do?
timotimo dunno
i'm going to drive my car back and forth through the city now
20:24 brother| joined 20:26 lue joined 20:31 chipdude joined 20:35 sorear_ joined 20:38 retupmoca joined, timotimo joined, lee__ joined 21:01 krunen joined
masak src/moar.h L53: '/* Headers for APIs for various other data structures and APIs. */' 21:05
should be s/for APIs//, no?
moritz_ aye 21:08
masak fixes 21:23
dalek arVM: 8eaa45c | masak++ | src/moar.h:
improve comment
 21:27
masak my attempts so far with reducing MoarVM to something that I can instrument its GC through have resulted in me being able to segfault the VM consistently. 21:54
it's fun to delete code, though.
FROGGS >.< 22:02
22:11 benabik joined
masak (trying this by paring moar down to just a GC. the segfault probably means I removed something the GC needed.) 22:20
nope. seems it segfaults even without that stuff removed. 22:23
that means I can go to bed posting a very simply question:
why does MVM_gc_allocate(instance->main_thread, 100); immediately after MVM_vm_create_instance() segfault? 22:24
timotimo i have no idea what you're hoping to do with a moarvm without the vm but with the gc
masak timotimo: I want to allocate objects, cause GC runs to trigger, and observe the effects. 22:25
timotimo ah 22:26
so replacing the whole run loop with your own
masak yes.
(I would keep the original runloop, but I don't believe it will have the control/reach I want.) 22:27
ooh, I managed to trigger 'Unhandled exception: Could not map file 'input.moarvm' into memory: FIXME'
(by providing ./moar with an empty input file)
anyway, 22:32
'night, #moarvm
[Coke] biggest moar failure is now: 23:28
Cannot call method 'dispatch:<.^>' on a null object
Here's one way to get that: '(1 | 3) < 3' 23:36
timotimo that would be junctions not being implemented 23:37
i wonder how hard that is.
wait. not quite 23:38
there's one piece of junction autothreading implemented
23:49 colomon joined
timotimo inside the bootstrap, the find_best_dispatcher method apparently knows how to autothread 23:50
jnthn Well, the 'dispatch:<.^>' thing is really about backtrace generation failing I guess... 23:56

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!