►
From YouTube: Node.js N-API Team meeting - March 4 2019
Description
A
A
A
B
A
A
A
Okay,
so
I
just
pasted
a
quick
question
to
Bethany
to
say
you
know
those
two
t,
FS
n
back
ports
for
10x
and
8,
that
X
will
they
land
as
staging
as
part
of
the
release
process,
or
should
they
be
happening
in
advance
and
because
she's
doing
the
12
dot,
X
release
and
and
is
in
the
release
working
group.
So
at
least
it
will
tell
us
whether
it's
just
the
way
it
should
be,
or
we
should
be
trying
to
remind
somebody
to
get
them
landed.
Okay,.
A
A
B
I
the
the
P
R
is
okay.
Now
it
runs
on
an
OS
10,
but
I'm
still
kind
of
I.
Don't
know
I'm
still
not
a
hundred
percent
sure
about
this
about
this
crash
and
I'm,
starting
to
run
out
of
ideas,
I
I
basically
tried
to
see
what
else
would
would
affect
whether
it
crashes
or
not,
and
the
only
other
thing
besides,
marking
the
functions
as
as
hidden
that
would
cause
the
crash
to
go
away
is
to
not
compile
one
of
the
add-ons
with
with
exceptions
and
the
other
without
so.
B
As
an
experiment,
then
again
it
doesn't
crash,
whereas
if
you
turn
on
exceptions
on
one
of
them,
it
does
crash
so
I
I,
don't
know.
What's
going
on
like
it's
completely
weird,
so
I
was
just
thinking
about
about
opening
a
second
PR
which
is
not
to
be
landed,
but
rather
just
to
illustrate
the
crash
and
and
there's
just
basically
just
I,
don't
know
pink
people
from
from
v8
and
stuff
because
it
does
crash
inside
v8
just
to
see
if
they
have
any
insights.
B
A
B
Oh
and
okay,
actually
I
do
have
one
idea
left,
which
is
that
we
should
try
to
put
the
implementations
of
functions
like
not
the
inline
ones,
but
but
things
which
we
actually
have.
There
pointers
taken
at
some
point
and
therefore
cannot
be
in
line
put
those
in
a
dot,
CC
file
and
and
so
and
then
compiled
it
as
a
static
library
and
link
it
against
whoever's,
using,
though
that
on
API,
just
just
like
we
do
with
with
the
implementation
of
an
API
when
an
API
is
not
built
into
nodejs
I.
Don't.
B
A
B
B
That's
not
normally
a
good
thing
right,
like
you
know,
normally
the
declarations
go
in
the
header
file
and
and
the
functions
go
into
a
dot
CC
file
right
now,
inline
stuff
can
go
into
a
header
file
right
because
that's
the
point
but
yeah,
but
this
is
not
inline.
This
is
an
actual
function
declaration
which
will
never
be
inline
because
you're
taking
the
pointer
of
the
function
at
some
point,
so
it
cannot
be
inline
because
of
that
right,
and
so
so
it's.
A
B
C
B
Yes,
but
if
it
were
templated,
then
then
then
you
know,
then
the
class
would
actually
be
declared
where
the
template
is
specified,
rather
than
immediately
where
it
appears
in
the
header,
and
so
so
the
symbols
would
be
would
be
different
for
every
specification
of
the
template.
So-
and
this
is
the
case
with
all
the
other
classes
right
because
we
have
run.
For
example,
we
have
the
we
have
the
object,
wrap
right.
Our
object
wrap
also
declares
a
bunch
of
static
functions
which.
A
B
A
A
B
A
B
A
B
A
B
B
A
C
A
A
A
C
C
Pushed
a
PR
on
the
first
developer
that
executed
the
porting,
but
I
received
any
any
response,
and
now
I
created
a
new,
a
new,
a
new
PR
trying
to
to
do
well
to
address
at
what
was
requested
in
the
original
para
cast.
And
now
we
have
a
new
para
cast
on
III,
going
to
call
so
I
I'm
waiting.
The
review.
E
Got
the
sqlite3
and
I
have
to
say
this
week.
I
did
not
make
any
progress
on
that,
but
my
immediate
objective
is
to
get
the
code
rolled
those
patches
rolled
forward
to
the
point
where
it's
passing
the
CI.
That
still
has
its
dependence
on
libuv
and
then
then
the
next
step
is
to
address
the
Libya
Libya
V
dependencies,
so
I
hope
to
get
that
first
phase
done
it
within
the
next
two
weeks.
E
A
B
E
B
There
was
only
one
thing
that
that
I
sort
of
realized
I
thought
about
it.
A
little
bit
I
never
actually
did
anything.
Is
that
so,
if
you
do
require
resolve
from
from
the
native
side,
what
is
it
that
you're
asking
to
require
resolve?
You
know
like
normally,
if
it's
a
required
resolved
package
name
and
then
you
would
get
your
path,
but.
C
C
B
Need
to
have
I
think
we
need
to
have
sort
of
a
fallback
sequence
like
try
to
package
name
it.
If
that
doesn't
return,
anything
then
and
then
try,
then
try
current
working
directory
or
something
like
that,
but
but
the
sequence
of
the
sequence
of
steps
and
and
the
yeah,
the
sequence
of
steps
needs
to
be
established,
and
we
need
to
agree
on
that.
You
know
like
how
do
you
best
establish
the
path
to
your
package
so
that
you
can
be
assured
that
that
your
add-on
is
located
in
a
directory
of
that
path?
Right.
A
A
A
B
A
Know
for
the
future,
like
a
short-term,
you
know
something
that
you
can
do
today
without
having
to
you
know,
get
an
updated
version,
it's
good!
So,
yes,
like
you,
said
it,
it
may
be
that
if
it's
like
you
know,
you
still
have
to
try
six
five,
five
different
things
or
something,
then
maybe
us
providing
a
path,
a
method
that
gets
you.
The
path
still
make
sense
right.
A
B
Yeah
well
yeah
or
some
something
of
a
similar
quality.
So
the
other
thing
is
in
terms
of
building.
This
way,
is
you
you
sort
of
have
to
I
mean?
Ideally,
you
know
you
can
always
hard
code
it,
but
ideally
you'd
want
to
establish
the
package
name.
It
builds
time
and
pass
it
into
the
source
code,
so
this
would
require
like
like
binding
chip
changes
as
well.
So
you
know
like
no
chip.
B
Module
name
will
give
you
the
name
of
the
add-on,
but
but
it
will
not
necessarily
give
you
the
name
of
the
NPM
package,
so
you
kind
of
want
to
like
parse
the
package
JSON
and
your
binding
chip
and
then
pass
in
the
package
name
into
the
build
system,
so
that
so
that
you
can
use
that
as
a
as
a
C,
preprocessor
constant
and
pass
that
in
to
require
at
runtime.
So
so
that's
another
thing
that
that
that
that
you
would
need
to
do
to
have
your
package
find
find
its
own.
C
B
B
In
the
end,
yeah
yeah,
so
yeah,
that's
yeah,
that's
that's
enough
sort
of
project.
Yep.
A
Okay,
tested
methods
think
still
on
our
list,
but
we
haven't
made
any
progress.
A
B
Thank
you
for
landing
the
those
those
macros
the
exposure
of
those
macros.
That's
that
that's
super
super
awesome
for
for
mixing,
plain,
C
and
and
know
that
on
API
calls
yeah
yeah.