►
From YouTube: Node.js N-API Team meeting - Oct 21 2019
Description
A
A
A
D
C
A
A
A
D
C
D
C
G
A
C
C
A
C
C
A
G
D
G
E
E
E
A
E
E
E
Okay,
if
you
change
not
just
Varsha,
you
will
get
that
mismatch,
another
model
Russia
and
that
invite
you
to
recompile
your
Maude
and
for
montanus
these
mean
these
means
that
I
do
with
the
issue
and
now
now,
if
you
search
from
the
table,
you
will
obtain
about
two
thousand
issue
about
this
problem.
Right.
A
E
E
C
E
A
C
A
C
E
C
E
E
A
Okay,
now,
the
only
other
thing
which
we
did
do,
the
last
time
which
we
haven't
talked
too
much
about,
is
I
try
well,
we
should
also
get
an
update
from
Jim,
because
I
know
he's
working
a
workshop,
but
we
also
did
look
for
modules
that
were
good
candidates.
I,
don't
know
if
we're
gonna
try
and
do
that
this
time
again.
E
A
D
D
E
E
H
E
H
E
D
A
A
E
H
E
D
D
A
E
A
E
C
E
C
A
E
G
D
A
A
D
C
Is
a
bit
old
come
to
think
of
it?
If,
if,
if
anyone
the
color
goes
through
that
spreadsheet,
he
may
have
more
things
to
post
here
right,
because
this
has
sort
of
become
a
collection
of
all
the
ones
that
have
been
posted,
yeah
an
API,
this
it's
kind
of
a
good
resource
that
way
I,
even
if
we
close
it
I
think
we
can
still
update
it
after
we
close
it
just
to
keep
it
around
as
a
resource
right
yeah.
It's
almost.
C
Because
the
thing
is
the
fundamental
problem
with
with
with
with
with
getting
a
handle
on
what
has
imported
and
has
not.
Is
that
figuring
out
what
has
been
ported
to
know
that
on
API
is
super
easy
because
it's
a
separate
package
and
you
can
check
the
reverse
dependencies
right,
but
the
things
that
use
an
API
directly
are
really
hard
to
find
because
they're,
they
don't
depend
on
anything
special
right.
H
A
C
A
B
C
B
A
We're
kind
of
well,
it's
been
it's
a
year
beyond
this
list,
being
the
active
list,
we've
pasted,
you
know
other
modules,
which
is
so
it's
kind
of
like.
This
is
a
good
idea
to
have
a
place
where
we
can
paste
and
sort
of
report
these
things,
but
is
continuing
to
just
adding
them
into
this
issue.
The
right
way
to
do
it
was
the
only
thing.
I
was
wondering:
okay,.
C
A
A
C
Having
something
that
you
can
just
edit
in
place
would
be
would
be
ideal,
I
think
and
then
every
once
in
a
while,
when,
when
somebody
when
somebody
bumps
into
a
new,
a
new
module
or
or
you
know,
says
an
issue
gets
filed
regarding
an
old
module,
then
we
can
say,
oh
by
the
way,
let's
update.
Let's
update
that
particular
record,
you
know
and
see
see
what
downloads
per
week
are.
Nowadays,
you
know
maybe
the
what
module
has
gone
out
of
favor.
C
Maybe
it
has
become
more
popular
who
knows
and
then
just
upload
the
or
update
the
downloads
per
week
and
and
just
leave
a
comment
saying
you
know
updated
as
of
and
then
today's
date,
and
that
kind
of
thing
so
basically
make
it
as
as
passive
as
possible
sort
of
a
server
dragnet.
If
you
will,
you
know,
whenever
we
bump
into
it,
will
remember
to
update
the
table
or
not,
and
then
it
yeah.
C
C
C
H
F
A
C
A
A
D
E
A
E
A
E
B
A
different
Sima
C++
function
that
we
pass
or
the
haulback
function,
the
one
that
gets
the
the
native
C++
data
and
transforms
it
into
a
JavaScript
value
and
the
reason
that
we
have
that
wrapper
or
whatever
is
so
that
in
blocking
call
and
non
blossom
all
we
passed
a
separate
callback
function
that
functionality
wasn't
there
and
underlying
math
and
thread-safe
function.
You
only
have
one
function
that
you
register
it
with,
and
it's
back,
but
this
wrapper.
We
pass
a
wrapping
of
the
data
and
your
callback
function
to
our
internal
callback
that
we
registered
with.
B
C
B
F
C
G
C
B
D
B
Yep,
that
was
one
way,
I
thought
of
it,
but
then
I
thought
of
another
way
just
now
and
just
implement
adding
a
new
method
that,
if
you
that
would
be
like
the
non
blocking
or
blocking
call,
but
it
would
be
sort
of
a
direct
one.
So
it
would
call
then
that'd
be
call
thread-safe
function
directly
instead
of
going
through
the
record.
But
this
is
one
way
that
I
have
here,
but
I
don't
know
if
it.
C
C
D
C
D
C
B
C
B
There
was
that
one
and
then
I
created
a
new
PR
Michael
I,
think
you
saw
it
for
adding
the
ref
and
ref
yep
yeah.
That
was
just
a
simple
one
because
they
were
missing.
However,
this
one
is
going
to
conflict
with
the
copy,
constructor
PR
right,
so
one
just
has
to
be
merged
in
first
and
then
I
can
update.
A
H
C
A
C
C
B
B
D
C
B
A
B
C
Yes,
yes,
but
I
mean
the
the
reason
like
it's
kind
of
like
the
test
passes
or
the
test
fails
if
it
exits
within
those
five
seconds
right.
So
in
that
sense,
we
can't
kind
of
do
what
time
test,
because
if
it
exits
after
five
seconds,
yes,
we
don't
know
whether
it
exited
whether
the
a
whether
it
exited
because
the
the
thing
was
unwrapped
or
whether
it
exited
because
it
was
reffed,
but
then
the
process
quit
anyway.
Why?
So?
So
we
don't
know
I.
C
A
C
A
F
C
A
A
B
B
C
B
A
B
G
G
A
C
And
additionally,
I
think
in
man
and
in
v8
these
things
have
no
copy
constructors.
So
the
fact
that
we
left
them
in
place
is
probably
an
oversight
on
our
part
is
that
they
are
supposed
to
be
used
exactly
just
you
know,
declared
and
then
let
them
go
out
of
scope
and
that's
it.
They
cannot
be
assigned.
They
cannot.
You
know
they.
They
don't
have
these
constructors
I,
don't
believe
in
v8
or
in
in
that
so
I
think
I.
G
A
Just
before
we
move
on
to
that,
one
is
what
I've
written
here.
What
we've
agreed
for
this
one
that,
like
the
team,
is
comfortable.
It's
extremely
unlikely.
This
will
break
an
existing
code,
as
the
code
would
already
be.
Breaking
would
always
be
crashing
and
vide
and
then
don't
have
the
copy
constructor.
So
they
don't
make
sense
consensus
is
we
should
remove
them?
Yeah
I
would
say
yeah,
yeah,
okay,
so
sorry,
I
just
wanted
to
make
sure
I'll
remove
my
comments
and
we
can
say
that
that's
ready
to
land.
A
G
A
So
it'll,
actually,
like
my
I,
think
my
question
was
around
in
the
past.
If
you
called
these,
nothing
would
have
liked
it.
Nothing
would
have
happened
right
yeah
we
keep
on
going,
even
though
you
did
the
wrong
thing
and
if
we
change
that
put
that
change
in
we're
probably
going
to
be
causing
it
to
stop.
So
is
there
any
chance
of
like
existing
code,
which
would
be
doing
the
wrong
thing,
but
getting
away
with
it.
C
It's
not
gonna,
it
doesn't
exist
exactly
so
I
think
I
think
the
same
argument
applies
here.
If
such
code
exists,
then
it
will.
It
will
break
silently
right
now
and
if
we
land
these,
we
would
be
much
more
pedantic
about
how
we
handle
it.
But
it
is
the
right
thing
to
do
when
it
is
the
way
Nana
VA
to
do
it.
G
G
C
A
F
C
E
C
Yeah
delete
okay,
yeah,
and
this
whole
thing
stems
from
the
fact
that
we
are.
We
are
heap
allocating
to
handle
scope
cause.
We
have
no
choice
right
because
it's
mm-hmm,
it
is
a
see
structure,
so
we
can't
we
can't
stack,
allocate
it
it
doesn't.
Have
you
know
stuff
that
it
does
on
going
out
of
scope?
It
doesn't
have
that
concept
in
C.
So
then
we
have
to
heap,
allocate
it,
and
then
we
have
to
worry
about
this
lifecycle,
and
this
is
part
of
that.
A
But
I'm
just
yeah
I'm,
just
thinking
now
like
in
the
previous
case.
It's
easy
like
it's
behind
the
scenes,
so
we
were
saying
how
that
gland
is
assembler
patch.
This
one
I
mean
we
could
potentially
bump
December
on
no
down
on
API,
even
though
we
don't
think
it's
gonna
make
a
difference,
but
then
that
would
mean
people
won't
get
the
new
version.
A
C
Yeah
I
mean
whatever
whatever
is
out
there
in
binary.
Will
art
will
continue
to
work?
Sure
yeah?
So
that's
the
good
thing
about
ABI
stability
like
whatever
is
out.
There
will
continue
to
work
no
matter
what
so
so
yeah
and
then
updates
updates
are
basically
updates
because
of
this
can
be
rolled
in
with
updates.
For,
for
you
know,
for
future
development
reasons
they
didn't
you
know.
Basically
we
are
avoiding
causing
an
update
because
we
broke
something
because
whatever
is
out
there
in
binary
will
continue
to
work
right.
A
C
So
then,
so
then,
the
the
the
up,
yes,
people
people
will
have
to
update
the
code,
whether
they
like
it
or
not,
because
you
know
if
we're
no
longer
compiled,
because
there
are
there's
no
longer
the
constructor
out
there,
but
they
can
roll
that
update
in
with
other
updates,
and
it's
not
gonna
hold
anybody
up.
So
right.
A
C
So
so
so
so
yeah,
we
have
several
reasons
right,
a
it's
unlikely
that
code
is
out
there,
especially
if
people
follow
the
the
NAND
model
and
the
very
model
which
is
to
declare
these
on
the
stack
right.
So
if
people
follow
that
model,
then
then
then
there's
no
reason
to
assume
that
it
will
be
broken.
So
that's
one
reason.
The
other
reason
is
that,
even
even
if
people
did
use
it
the
wrong
way
and
and
people
downloaded
the
result
and
installed
it,
it
will
continue
to
work
with
every
version
of
node
that
comes
out.
C
It'll
just
have
these
silent
failures,
so
so
it's
not
gonna
break
anybody,
and
and
for
the
purposes
of
updating
you
know
if
the
module,
if
the
module
is,
is
left
behind
and
and
nobody
makes
any
more
updates
to
it.
Then
then,
they're
not
going
to
update
the
package.json
to
depend
on
this
new
semver,
either
right
and
so
also
the
so
so
anybody
who
recompiles
it
in
the
future
will
continue
to
receive
these
silent
failures,
but
it
will
otherwise
be
okay,
and
if
they,
if
the
module
is
still
being
maintained,
then
there
will
have.
C
A
A
Because
it's
not
breaking
ABI,
it's
this
part
is
not.
Maybe
that's
why
we've
left
ourselves
a
flexibility.
This
piece
is
not
EBI,
it
does
not
does
not
depend
on
your
API
stability.
It
doesn't
affect
your
API
stability
if
you've
got
something
that's
already
compiled
continues
to
work.
If
you
do
upgrade,
you
may
have
to
make
some
code
changes,
but
we
don't
expect
it
to
happen
and
like
if
you
go
to
now,
I
guess
you'd
have
to
actually
update
your
version,
the
add-on
it
doesn't
matter.
If
you
updated
your
version
of
note
or
not
yeah,.
A
E
C
A
A
G
A
G
D
A
So
the
key
thing
is
like
it:
what
I'm,
thinking
and
I
haven't
had
time
to
look
as
potentially
it
would
cause
test
not
to
be
run
that
we
do
run
today
or
it
might
cause
that
that's
the
one
thing
I'm
I'm
wondering
like,
because
it's
we
end
up
saying
what
version
does
this
version
support?
We
then
set
the
experimental
version.
G
Yeah
I
think
I
have
proposed
up
our
time.
Footpath
I
think
we
should
distinguish
which,
which
picture
should
be
in
which
and
a
know
the
major
version,
some
just
a
period
before
that
the
into
and
enter
data.
They
are
both
not
experiment
about
they.
How
different
and
no
the
major
version
stages.
A
C
C
C
A
Six
right,
there's
I
guess:
if
there
was
a
real
demand,
we
could
do
it
now.
We
just
mean
that
nothing
more
could
be
back
ported
to
eight
yeah,
which
you
know
at
this
point
is
actually
probably
the
case
anyways,
true,
okay,
but
yeah
I
think
we
should
be.
We
should
resolve
those
two
in
this
release
as
well.
Just
sorry,
I
haven't
had
time
to
think
think
through
them
enough
to
be
comfortable.
Yet.