Camelia, the Perl 6 bug

IRC log for #moarvm, 2013-09-20

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

All times shown according to UTC.

Time Nick Message
00:38 wsri joined #moarvm
00:55 jnap joined #moarvm
01:19 _sri joined #moarvm
01:49 benabik joined #moarvm
02:08 d^_^b joined #moarvm
02:26 jnap joined #moarvm
02:42 dalek MoarVM: e2fb9f6 | jimmy++ | src/6model/reprs/P6bigint.c:
02:42 dalek MoarVM: s/mp_toradix_n/mp_toradix/, (Ben Goldberg)++
02:42 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/e2fb9f6214
02:44 dalek MoarVM: 4db966e | jimmy++ | src/6model/reprs/P6bigint.c:
02:44 dalek MoarVM: fixed a bug
02:44 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/4db966e15f
03:36 foo_bar_baz joined #moarvm
05:05 dalek MoarVM/gcorch: f832f13 | diakopter++ | src/ (4 files):
05:05 dalek MoarVM/gcorch: wip9
05:05 dalek MoarVM/gcorch: review: https://github.com/MoarVM/MoarVM/commit/f832f1366f
05:24 dalek MoarVM: b1efd97 | (Gerhard R)++ | src/core/oplist:
05:24 dalek MoarVM: Fix op annotation docs
05:24 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/b1efd97dd4
05:24 not_gerd joined #moarvm
05:25 not_gerd o/
05:25 yoleaux 19 Sep 2013 06:51Z <JimmyZ> not_gerd: I can't get 5172c20deb758cfc95f5745556595c21c788eb67
05:26 not_gerd JimmyZ: what's the problem with that commit?
05:28 diakopter o/
05:36 JimmyZ not_gerd: I don't know where is double freed
05:36 yoleaux 05:17Z <BenGoldberg> JimmyZ: Thank you very much! :)
05:36 yoleaux 05:17Z <BenGoldberg> JimmyZ: Here's the result of branching and making a pull request: https://github.com/BenGoldberg1/libtommath/co​mmit/49642b6dbb2d86cf3f1e93f501faad17bdf8c31a
05:42 not_gerd JimmyZ: MVM_HASH_DESTROY() already calls free() on the hash entries
05:42 not_gerd I manually did so again
05:42 not_gerd a nice recipe for memory corruption
05:43 JimmyZ not_gerd: oh
05:44 JimmyZ I thought MVM_HASH_DESTROY didn't
05:44 not_gerd that was my assumption as well before I actually looked at its definition ;)
05:47 JimmyZ hehe
05:48 JimmyZ diakopter: Is there any reason ch'uv_thread_t thread' to 'uv_thread_t *thread'?
05:48 JimmyZ s/ch/to change/
05:49 benabik joined #moarvm
05:50 diakopter JimmyZ: yes... because surely libuv keeps a pointer to that struct
05:50 diakopter ... but our gc moves that memory
05:51 JimmyZ It could be a good commit message
05:51 diakopter yeah, when the branch is done
05:58 not_gerd diakopter: using uv_thread_t directly instead of by pointer should be fine, imo
05:58 not_gerd both pthreads and winapi back them by opaque types with value semantics
05:58 diakopter why
05:58 diakopter oh
05:59 diakopter yeah
05:59 diakopter but
06:00 JimmyZ pthreads uv_thread_t is actually a int type ?
06:00 JimmyZ where winapi is HANDLE
06:01 not_gerd I believe POSIX required pthread_t to be an arithmetic type, but no longer does so
06:01 not_gerd regardless, value semantics should hold as they get passed around by value in the API
06:04 diakopter In the iSeries implementation of Pthreads, the thread ID is a 64-bit integral value and the pthread_t is an abstraction (structure) that contains that value and others.
06:05 JimmyZ hmm, I didn't see  libuv keeps a pointer all the time. if we doesn't do alloc, it's fine
06:05 diakopter but it looks like the pthread functions take pthread_t by value not pointer
06:06 JimmyZ diakopter: WinAPI doesn't too
06:06 not_gerd exactly (except for eg create where it's an out parameter)
06:06 JimmyZ WinAPI take handle as value
06:16 diakopter not_gerd: we need an efficient portable semaphore abstraction.. from pthreads I'm sure, where we have that
06:16 diakopter and I'm sure there are quite a few in windows available
06:18 diakopter seems we'll need to include pthreads.h
06:18 diakopter pthread.h
06:25 diakopter not_gerd: er, oops
06:25 diakopter seems libuv has it
06:51 not_gerd diakopter: is there a reason why MAST nodes come with DUMP() instead of dump()?
06:51 not_gerd it breask --target=mast
06:54 diakopter I don't know
06:54 diakopter change it if you like
06:54 diakopter it doesn't "break
06:54 diakopter "break" anything
06:54 diakopter it was created without any knowledge of --target at all
06:54 not_gerd ;)
06:58 FROGGS joined #moarvm
06:59 diakopter not_gerd: adapt it to --target if you like :)
06:59 not_gerd diakopter: just renaming it makes it work
06:59 not_gerd I can push that
06:59 FROGGS o/
06:59 not_gerd \o
07:00 diakopter not_gerd: even though it strongly begs to be refactored/redone, don't spend time fixing DUMP_lines :)
07:01 FROGGS .tell JimmiZ Re e2fb9f6214, please let us stay in line with nqp@parrot if they share identical code
07:01 yoleaux FROGGS: I'll pass your message to JimmiZ.
07:05 JimmyZ FROGGS: why?
07:05 FROGGS not_gerd: you would need to rename it to dump for nqp itself too, no?
07:06 FROGGS JimmyZ: because if you have the chance to have identical codepaths, it makes it easier to maintain
07:06 diakopter FROGGS: well, to be precise, would rather match the jvm impl
07:06 FROGGS JimmyZ: I mean, let us use bn_mp_toradix on nqp@parrot too
07:07 JimmyZ FROGGS: it didn't change anything, just faster
07:07 FROGGS diakopter: that is weird
07:07 FROGGS JimmyZ: so why not do that on nqp@parrot too, to stay in sync?
07:07 FROGGS JimmyZ: I am testing this already fwiw
07:08 JimmyZ OK, personally, I don't want to maintain nqp@parrot
07:08 not_gerd FROGGS: I don't follow - HLL::Compiler expects a method called dump
07:09 not_gerd MAST nodes come with DUMP, which we could just rename
07:09 FROGGS not_gerd: I have seen DUMP in nqp repo too, that is my point
07:09 FROGGS ohh, hmm
07:09 FROGGS not_gerd: can't find it anymore...
07:10 not_gerd ;)
07:10 FROGGS nvm then :o)
07:23 krunen joined #moarvm
07:24 jimmy_ joined #moarvm
07:24 JimmyZ jimmy_: test
07:48 donaldh joined #moarvm
07:53 odc joined #moarvm
08:55 sorear joined #moarvm
08:56 JimmyZ joined #moarvm
09:01 dalek MoarVM: f073719 | (Gerhard R)++ | / (3 files):
09:01 dalek MoarVM: Rename DUMP to dump to make --target=mast work
09:01 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/f073719533
09:01 dalek MoarVM: 2f01fa9 | (Gerhard R)++ | lib/MAST/Nodes.nqp:
09:01 dalek MoarVM: Improve readability of --target=mast
09:02 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/2f01fa9d6f
09:11 FROGGS nice
09:31 not_gerd sadly, it din't help me figuring out why `try nqp::die(42)` returns a code object instead of NQPMu
09:32 FROGGS :/
09:33 FROGGS I failed to understand that one too
09:46 not_gerd note that we don't implement the handled op, which might or might not be related
09:50 not_gerd reading the docs, probably not
11:22 not_gerd I figured out why 56-role.t fails
11:22 not_gerd see https://gist.github.com/gerdr/babe2cbd5d85c9170dec
11:22 not_gerd not sure yet how to best fix it
11:37 FROGGS not_gerd: jnthn's plan is to refactor out the prior_invocation thing btw
11:44 not_gerd I don't really see how we can locate the correct outer if we haven't made the code object into a closure...
11:58 not_gerd ../moarvm nqp.moarvm -e 'role Foo { say(42)  }; class Bar does Foo {}'
11:59 not_gerd No lexical found with name '&say'
12:05 JimmyZ :/
12:13 FROGGS :/
13:02 jnap joined #moarvm
13:24 benabik joined #moarvm
13:39 not_gerd bye, #moarvm
13:39 not_gerd left #moarvm
13:50 tokuhirom joined #moarvm
13:51 yoleaux joined #moarvm
14:10 benabik joined #moarvm
14:40 colomon joined #moarvm
16:24 diakopter commutafarianism&
16:54 wsri joined #moarvm
17:42 tba joined #moarvm
18:14 flussence joined #moarvm
18:16 FROGGS joined #moarvm
19:09 benabik joined #moarvm
19:40 benabik joined #moarvm
21:10 cognominal joined #moarvm
21:50 foo_bar_baz joined #moarvm
21:51 colomon joined #moarvm

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