dalek arVM: d3713c9 | jnthn++ | src/ (2 files):
Replace a #define with a static inline.

There was a note in the #moarvm backlog that we may be failing to pay attention to the "allocate in gen2" flag by using the macro too early. While I can't find any evidence of that, changing to a static inline should make such a bug less likely to be introduced.
10:09
nwc10 jnthn: I think that the problem is MVM_gc_allocate_object(), which is used in lots of 6model repr structs 12:36
looks like it allocates in the nursery
and it's called via the function pointer ->allocate() in various places
wait, huh 12:37
jnthn MVM_gc_allocate_object calls MVM_gc_allocate_zeroed which calls MVM_gc_allocate
nwc10 yes, hence "huh"
I'd just got there
jnthn Which is the thing that used to be a macro but is now the inline static.
I was kinda excited that we were gonna discover a stupid thinko and it'd explain so much, but it seems that no...
nwc10 anyway, I hit the problem a third time and didn't paste the stacktrace
but the failure (trying to lock a mutex again) happens down in SCRef.c each time 12:38
jnthn grumble
Of course I can never see these problems on Windows.
'cus mutexes *are* reentrant by default there. 12:39
timotimo o/ 13:48
FROGGS jnthn: (just a reminder) moarvm.org needs an update 19:16
jnthn FROGGS: I'm sleeping now 'cus 4:30am wake-up for flight (and after it I at least will have a week without any flights, *phew*...) Anyways, please poke me again in the next couple of days. :) 19:21
FROGGS jnthn: a week without flights is better than a week without sleep (which is what my brain constructed first) 19:22
k, will poke :o)
timotimo gnite jnthn :) 19:30
jnthn :) 19:31
zzz &
FROGGS gnight jnthn