00:02 colomon joined 00:50 dalek joined 01:25 ilbot3 joined 01:49 FROGGS joined 02:00 woosley joined 02:25 benabik joined 06:36 FROGGS joined 07:13 brrt joined
brrt hi #moarvm 07:29
nwc10 hi brrt
FROGGS morning brrt
<jnthn> .tell brrt I'm on target to have the initial code available on Tuesday 07:30
<jnthn> bah, no message bot here
<--- le message bot
brrt i've seen it Froggs, thanks :-) 07:31
hi nwc10
FROGGS :o)
brrt i read backlogs too ;-)
FROGGS okay, noted 07:32
brrt but thanks anyway
awesome, excited
FROGGS me too 07:33
quite a few nice thing are happening atm 07:34
things*
brrt uhuh
i'd like to help out with the whole lock-placement things,but i'm a bit strapped for time right now
07:45 cognominal joined 08:58 woolfy1 joined
cognominal FROGGS prend son pied à fait le bot. 10:02
FROGGS cognominal: google translates it as "takes his foot made ​​the bot", but somehow I doubt that this is correct :o) 10:04
cognominal that's a indirect pun about "pied bot" 10:05
jnthn mmm...pie bot 10:06
brrt :-) 10:07
brrt lunches
cognominal "il prend son pied" is translated litteraly by Google but really means to "he enjoys himself"
and pied bot is fr.wikipedia.org/wiki/Pied_bot 10:08
btw, s/fait/faire/ on my original sentence
masak "FROGGS enjoys himself creating the bot"? 10:45
jnthn playing the bot, maybe 10:46
or being
cognominal yup 10:50
polysemy makes google-translate a fool all the time. 10:51
masak oh, "acting the bot". 11:03
11:04 colomon joined
masak cognominal: polysemy is probably AI-complete ;) 11:04
cognominal nope "playing/being" the bot is/are correct 11:06
masak oki 11:07
cognominal il y du jeu dans les langages naturels. "Jeu" meaning either game or en.wikipedia.org/wiki/Backlash_(engineering) 11:09
masak marvellous. 11:12
11:40 zakharyas joined 12:09 colomon joined 12:35 JimmyZ joined 12:44 colomon joined 13:48 btyler joined 13:59 jnap joined 14:02 btyler joined 14:13 woolfy1 left 14:37 FROGGS[mobile] joined 14:51 zakharyas joined, FROGGS joined 15:32 brrt left 15:47 benabik joined 16:13 FROGGS_ joined 16:44 jnap1 joined
dalek arVM: 26e829f | gerd++ | Configure.pl:
add the both flags 'has-libuv and 'has-libatomic_ops'
17:17
jnthn gerd also notes, "I think there will no other flags for the Fedora build needed." 17:18
So looks like we're in decent shape for packagers.
FROGGS cool! 17:30
nwc10 that sounds like a challenge :-) 17:31
jnthn well, gerd++ is working on it for fedora. :) 17:42
18:11 colomon joined
timotimo is that gerd or not_gerd? 18:13
FROGGS gerd
timotimo OK
i still wonder what not_gerd is up to these days 18:14
FROGGS yeah, me too 18:15
I really hope he is well
18:29 brrt joined 18:35 jnap joined
dalek arVM: 57f735a | (Tobias Leich)++ | src/math/bigintops.c:
shortcut shr when we have a small lhs but rhs is >= 32

Otherwise we will run into issues where a >> 32 == a, at least on gcc.
18:57
arVM: 02c7ece | jnthn++ | src/io/syncstream.c:
Revert "First pass at locking for syncstream."

This reverts commit 1261936ddb75e91c78a91d830dad3a575d17ffe3. We'll take a safer, more general, approach.
19:15
arVM: eceea4c | jnthn++ | src/core/ (3 files):
Add infrastructure for mutex release on exception.
MoarVM: 81d57d5 | jnthn++ | src/io/io.c:
MoarVM: Protect all I/O operations using handle mutex.
MoarVM:
timotimo i like the sound of that 19:22
19:29 lue joined 19:36 jnap joined 19:37 colomon joined
brrt if i understand it correctly, the thread context is now provided with a mutex especially for when an exception is thrown 19:43
i'm not so sure that is a good idea
it maybe better - perhaps - to write macro's to create a try/catch/finally kind of idiom 19:44
but i'm not so sure about that, either
jnthn brrt: This basically *is* one.
brrt well, what happens when you forget to 'clear' the exception mutex?
jnthn brrt: Note the code path its on: exceptions thrown inside VM internals. Not general user-land ones.
brrt true 19:45
jnthn brrt: Same as happens if you forget to unlock one.
brrt but in that case, why not keep the unlocking local
jnthn Because exceptions are non-local.
We lngjmp back to the interpreter.
brrt i agree, its just - there are now two paths in which the mutex can be cleared
jnthn In the normal non-exceptional case, we never hit this.
brrt 'normally' and 'via the exception' 19:46
jnthn Right, the happy path and the sad path.
*nod*
brrt in other words, why don't we do the 'goto exception' cleanup dance that is typicall in c-land?
jnthn I think I'd rather not try and invent a try/catch/finally mechanism in C when we already have lngjmp going on.
brrt true
wouldn't work anyway, with what i'm imagining 19:47
jnthn Thing is that exceptions can be thrown fairly "deep" in IO stuff (such as, in the decoder).
brrt you could make an argument that the thread itself take a list of all mutexes that are held 19:48
threadcontext object
jnthn Well, if we ever get to a "taking multiple" situation we can make it a stack
But I really hope not.
19:48 colomon joined
jnthn Because as soon as you're taking multiple you're asking for deadlocks. 19:49
brrt it depends i guess
multiple locks are common in stuff like queues
jnthn You don't take both though. 19:50
brrt i'm just saying, i guess, that stuffing a single mutex into the thread context for cleanup when an exception is thrown seems like something that will be broken
.. no i agree
jnthn I agree it's fragile, but it was the second thing after the first thing was far more fragile.
C doesn't really provide us good ways to do this.
brrt maybe we'll come up with a better solution
no indeed 19:51
c++ has the compiler fix it for you, i believe
jnthn I'm fine with that, but whatever it is should be easy to reason about.
brrt longjmp isn't easy to reason about :-) 19:53
(for many other things, though, i think C is the knee's bees. not for exception handling, though) 19:54
jnthn Right, that's why I only want *one* longjmp'ing thing in the VM if possible :) 19:55
19:57 jnap joined
brrt as a manner of distraction, want to see mergesort-for-linked-lists-in-scheme? 19:57
some blogger i read said 'don't sort linked lists on a whiteboard during an interview' and i thought it'd be a nice challenge 19:58
jnthn People say "don't do live demos" but I always ignore that too :P 19:59
brrt and rightly so
jnthn could glance it :)
brrt github.com/bdw/lab/blob/master/sch...gesort.scm with pride :-)
the point to notice is that i only spend n time to 'divide' the list once 20:00
20:01 jnap1 joined
brrt in a more naive implementation i'd spend n time for each 'division' which doesn't affect asymptotic runtime but is not very efficient anyway 20:01
jnthn Cute :) 20:04
brrt thanks :-) 20:07
jnthn glances the rosettacode one out of curiosity
brrt i'm a closet lisp / scheme fan
d'oh, rosettacode
jnthn I didn't do enough of those, though I went to a "Build your own lisp for great justice" talk a few months ago and started hacking a lisp compiler in NQP in it :) 20:10
brrt nice 20:11
jnthn bbi10 20:15
21:09 jnap joined 22:28 colomon joined 23:45 woolfy joined 23:55 lizmat joined 23:58 jnap joined