►
From YouTube: Node.js N-API Team meeting - Feb 11 2019
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
B
C
A
D
D
E
E
A
D
So
I
had
well
I
had
trouble
with
with
my
suppressed
destruct
PR,
because
it's
sec,
faulting
on
on
OS,
10
and
I
just
got
access
to
one
of
the
OS
10
machines
from
from
Raphael
and
so
I'm
gonna
debug
it
there
I
don't
know
what's
going
on.
It
seems
really
strange
it's
as
if
it's
as
if,
when
you
declare
a
static
variable
inside
a
class,
and
you
have
two
different
shared
objects
which
both
have
that
class.
Somehow
the
the
data
symbol
is
being
shared
between
the
two
shared
objects
and
then
I.
So
I
figured
okay.
D
If
that
happens,
I'll
just
stick
it
in
an
anonymous,
namespace
and
that'll,
be
like
you
know,
static
only
to
the
file
and
not
to
like
it
should
be
static
inside
the
file.
So
it
should
not
be
seen
the
symbol
shouldn't
be
seen
at
all,
but
then
it
just
like
faulted
and
I
have
no
idea
where
it's
seg
faults,
but
the
correct
solution
is
to
wrap
the
whole
class
in
an
anonymous
namespace
and
so
I.
D
Mean
like
two
instances
of
the
class
I,
don't
know
what
it
is
so
so
so
there's
a
there's
a
class
declaration
and
it
has
a
it,
has
a
static
variable
right,
right
of
type
pointer
and
so
so
for
debugging
I
did,
like
you
know,
f,
printf
and
%
the
variable
to
see
where
in
memory
the
variable
gets
stored
right
and
on
Linux.
You
know
we
have
we
compiled
the
we
compile
each
test
add-on
twice
right.
We
compile
with
exceptions
and
without
exceptions.
E
B
D
Tests-
and
so
so
so
you
know
this
printing
out
of
the
address
of
the
static
variable
prints
out
the
point
there
twice
right
and
in
Linux
those
are
two
different
pointers,
but
on
OS
10
they
are
the
same
pointer
which,
which
is
unheard
of,
because
why
like?
Why
do
they
overlap
in
memory?
It
does
does.
Does
OS
10
do
some
kind
of
really
weird
simple,
merging.
D
We're
talking
about
two
different
shared
objects
or
thing,
two
different
symbols
that
just
happen
to
have
the
same
name,
but
they
should
be
located,
yeah
sure
it's
actually
shared
in
anyway,
because
they're
two
different
libraries
I,
should
get
a
cymbal
clash
or
something
and
refused
to
load
the
second
one
or
or
some
sort
of
more
normal
resolution.
So
I
have
no
idea.
What's
going
on
over
there,
two.
A
A
D
Want
Linux
Linux
doesn't
have
any
trouble
with
that,
but
I
guess
does
scope
things
better.
Maybe
I
don't
know
honestly
and
I
suspect
that
some
something
else
might
also
be
fishy,
because
when
I
do
wrap
the
whole
thing
in
an
anonymous
namespace,
it
should
not
set
ball,
because
at
that
point
it's
all
local.
So
all
right!
That's
that's
why
I
need
access
to
the
Mac,
so
so
I
can
see
why
it's
egg
falls.
There
might
be
something
else
entirely
wrong
with
the
cold
right.
A
D
A
C
C
A
Not
normally
that's
okay
because
usually
like
if
non
NEPA
models
cannot
be
loaded
if
the
ABI
version
is
different
and
the
obeah
version
would
be
different
if
libuv
is
different.
So
it's
only
when
you
get
to
n
API
that
it
kind
of
disqualifies
you
from
the
from
the
the
benefits.
If
you're
trying
to
use
Livia
for
regular
modules,
you
know
v8
might
change.
Allah
BV
may
change,
but
it's
all
bundled
into
that
one
ABI
number
and
I.
Guess
it's
that's
where
our
other
discussion
around.
A
B
C
A
C
A
Think
I'd
ask
that
and
I
my
impression
that
it
wasn't
the
case,
but
who
knows
I,
probably
misunderstood,
or
things
changed,
but
okay,
but
no
that's
good.
It's
good
to,
like
you,
said,
learn
more
and
I.
Think
it'll
help
us
figure
out
what
the
path
forward
might
be
on
a
few
other
things
as
well:
I'm
looking
at
the
other
ones,
icon
so
I
know
Nicola
you'd
asked.
Maybe
you
could
take
that
one
over
in
the
issue
I
haven't
seen
if
there
was
in
the
answer
or
anything
like
that,
yeah.
E
E
A
A
C
So
I
didn't
quite
realize
when
I
was
doing
the
PR
for
this
guy,
quite
who
he
was
he's
done
a
law.
So
don't
hey
last
time,
I
said:
have
you
had
another
look?
He
has
reviewed
them
once
and
he
asked
me
to
make
changes.
I
made
two
changes
and
I
four
days
ago
said
you
have
you
had
another
look
but
soon
as
he
wrote,
no
chip
and
all
those
other
things
I
think
he's
a
really
busy
guy
and
he'll
get
round
to
it
when
he
gets
around
to
it.
Okay,.
A
G
A
We're
still
hopeful,
we
can
use
the
async
worker
that
we
already
have
for
that.
It's
just
not
as
straightforward
as
we'd
like
so
right.
Okay,
node
report,
we've
already
talked
about
really
it's
in
landed.
In
course.
Oh
yes
and
then
we've
0
and
Q
who've
already
got
a
next-generation
version
on
the
roadmap.
So
I
think
we're
good
on
note
that
one,
unless
anybody
has
something
else
to
bring.
D
E
D
D
D
Modules
to
port
issue,
oh
and
it
being
it
being
our
socket
it
being
a
socket,
related
package.
I
guess
it
also
makes
heavy
use
of
Libya,
be
precisely
because
of
that,
because
it
does
things
like
ping
and
stuff.
So
for
that
you
need
pretty
low
level
access
wealth
either
you
use
libuv
or
you
use
the
OS
level
stuff
yeah.
You
know,
I
I,
think
it's
I,
think
it's
libuv
because,
like
even
if
you
use
the
OS
level
stuff
for
creating
the
socket,
you
know
integrating
it
with
the
event
loop
and
passing.
A
A
A
D
A
D
A
D
C
Sorry,
yeah
I
did
have
a
quick
look
at
that
gay,
broad,
so
hi
Gabe
rise
following
what
you
said.
It
looks
very
much
like
the
same
thing
as
the
node
serial
port.
Basically,
people
once
again
indication
you
know
an
app
or
a
file
description
when
it's
opened
and
changed
and
it
looked
like
those
you
know
they
were
doing
the
same
thing
and
I
wonder
how
many
more
people
are
doing.
This
same
thing.
C
A
It
gets
back
to
the
the
discussion
of
we
don't
want
to
wrap
all
of
the
Lib
Evie
right
if
it's
some
small
portion
like
if
it
was
one
or
two
calls
that
was
required
in
both
of
these
cases.
That
might
make
sense.
But
if
it's
like
you
know
all
of
the
socket
API
is
well,
then
that
doesn't
make
a
lot
of
sense.
I,
don't
think
what.
C
A
So
they're
using
the
native
tools
and
stuff
to
open
up
the
sockets
like
the
native
calls
to
open
the
sockets,
but
then
they're
polling
them
with
libuv
yeah,
okay
right,
so
that
could
make
some
sense
if
it
was
like.
You
know
couple
some
small
subset.
We
could
consider
if
we
should
add
that
to
an
API
itself.
A
Okay,
no
yeah
that'd
be
very
useful
to
come
back
with
that
list
and
I
think
also
do
we
have
an
issue
open
Gabriel,
where
you're
pushing
on
I
know
you
do
in
the
node
repo?
Maybe
that's
enough,
but
like
we
should
probably
you
know
see
if
we
can
push
to
what
we
think
would
make
sense
in
terms
of
that
I
ABR
checking
yeah.
D
A
D
Is
to
get
that
going,
yeah
I,
it
doesn't
involve
like
like
some
of
the
refactoring
that
I've
done
in
that
in
the
existing
pyaari
is
gonna
have
to
stay,
because
we
have
a
fundamental
problem
where,
where
the
the
you
know,
some
of
the
some
of
the
constants
that
we
need
are
are
declared
in
like
no
dot
H
and
no
that
H
can
only
be
included
from
C++.
And
so
now
you
know
those
parts
of
no
dot
H
which
are
really
C++.
D
Agnostic
should
not
be
inside
no
H,
because
there's
no
reason
so
you
know
that
that's
part
of
the
refactor
is
to
is
to
pull
that
stuff
out
of
no
H
and-
and
that's
that's
a
pretty
big
reshuffle
so
so
that
you
know.
Raphael
has
already
remarked
that
that
maybe
cember
major
though
I,
don't
see.
Why?
Because
you
know
we
don't
support
copying
files
out
of
out
of
tree
right,
because
you
know
that's,
that's
not
part
of
December
commitment
right
well,.
A
E
E
D
G
D
A
A
D
A
small
one
about
about
a
week
week
call
back
I.
Somebody
reported
that
we
should
not
allow
calls
into
the
engine
from
from
the
week
callback
because
the
the
engine
is
in
like
or
the
heap
is
in
an
inconsistent
state
during
during
a
garbage
collection,
which
is
when
these
weak
callbacks
get
called.
However,
v8
provides
sort
of
a
second
pass
callback
which,
which
provides
the
opportunity
to
attach
a
second
pass
week,
callback,
which
is
called
from
the
first
pass.
D
We
call
back
and,
and
so
so
then
so
then
the
U
and
the
second
pass
by
the
time.
The
second
pass
we
call
back,
it's
called
the
heap,
is
in
a
consistent
state,
and
so
so
basically
we
sort
of
do
this
ping-pong
where
you
know
the
first
line
week,
callback
resets
the
reference
and
attaches
the
second,
the
second
pass.
We
call
back
and
then
that
one
runs
the
N
API
we
call
back,
which
is
then
free.
D
It's
just
never
happened
before,
but
that,
but
the
possibility
is
open,
and
so
so
the
the
PR
for
for
for
doing
this
ping-pong
Inge
is
now
pretty
much
on
Rails
I
have
to
check
on
the
status
of
the
CI
and
then
land
it
and
then
comes
the
back
porting
to
version
10
and
version
8
of
that
er
them
that's
about
it.
From
from
the
napi
bugs
side,
yeah.
D
A
D
A
A
A
B
B
A
B
Just
exposing
through
the
API
access
to
those
objects-
okay,
okay,
great
you
were
looking
for
the
is
methods
essentially
really
in
that
one
apartment
I
think
they
worked
around
it
using
something
into
the
C++
wrappers.
They
were
able
to
use
to
work
around
not
having
a
real
native
one
in
the
API,
but
if
they're
gonna
be
something
we
should
ask.
Okay,.