►
From YouTube: 2021-09-10-Node.js Node-API Team meeting
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
So
welcome
to
the
node.js
node
api
team
meeting
for
september
10th
2021.
We
will
follow
our
standard
approach
of
looking
at
the
issues
tagged
for
milestone
11.
before
we
do
that.
Does
anybody
have
any
announcements
they'd
like
to
make.
B
Node
rp
headers
has
been
published
on
npm,
so
yeah.
If
you
won't
use
you
can
from
from
npm
now
and
yeah,
I'm
starting
to
work
at
a
building
tool
that
uses
cmajs
cmake
yeah.
That
used
not
being
noted
behind
us
so
yeah.
I
hope
to
give
you
good
news.
In
a
few
weeks.
A
My
purpose
yes
right,
and
so
if
there
are
other
implementations
of
of
the
node
api
you
could,
then
you
know
build
once
without
tied
to
a
runtime
and
then
deploy
run
them
with
any
of
the
ones
that
supported
that.
That's
that'll
be
really
neat
to
see.
A
A
Okay,
any
other
announcements.
A
So
the
first
one
was
the
review
on
node
serial
port.
I
know
a
few
people
were
looking
at
that
last
time.
It
was,
you
know,
comments
being
made
things
being
moved
forward.
I
don't
know
if
there's
any
updates
things
to
discuss
on
that
one.
This
week.
A
Okay,
so
I
guess
just
continuing
to
work
away
and,
as
we
find
time
to
comment
and
stuff
help
them
move
forward
on
that
sounds
good,
okay.
So,
let's
then
flip
over
to
the
stale
issues.
So
there's
this
one
9112,
which
was
right
so
I
know,
there's
an
update
from
jack,
at
least
on
that
front.
D
Yeah,
I
merged
the
pr
into
a
no
core
allowing
us
to
create
a
reference
on
symbol
types.
So
with
this,
I
think
we
can
probably
implement
some
like
check
handlings
from
our
node
api
side
out
on
epsilon,
if
possible,.
A
A
I
guess
we
could
enable
it
for
a
later
version
of
like
later
versions
of
node.
A
A
F
C
C
So
taking
gabriel's
thought
of
like
just
trying
it
and
if
you
get
invalid,
then
you
know
you
can't
do
it
so
the
solution
that
we
have
thought
of
now
that
we
can
reference
symbols
is
the
one
where,
if
there's
an
error
and
it's
a
primitive,
we
would
take
the
native
value
of
the
primitive
and
then
store
it
and
then
re-wrap
it
to
give
it
back.
Is
that
what
we
were
gonna
do.
C
Right
so
then,
if
we
take
the
approach
where
you
know
for
primitive
values,
we
just
do
the
wrapping
and
then
for
everything
else.
We
would
try
to
create
the
reference
which
would
be
functions,
objects
and
symbols
that
then
the
symbol
would
give
an
invalid
arg.
Then
how
would
we?
What
would
we
end
up
doing
in
that
case,.
F
A
A
C
C
C
C
E
C
F
Things
and
it
will
start
work
with
symbols
as
soon
as
this
change
periods
right,
so
you
know
we
can.
We
can
document
the
caveat
and
say
that
you
know
you
know
as
of
node.js
version,
whatever
the
next
release
will
be,
it
will
work,
but
until
then
it's
poised
to
start
working
at
points,
and
we
still
have
work
to
do
to
implement
change.
For
you
know
you,
you
store
primitive,
et,
cetera,
et
cetera,
so
that's
still
working
still
provides
a
lot
of
value
right.
A
Instead
of
a
crash,
does
it
like?
Is
there
any
way?
Is
there
anything
useful?
We
could
do
in
terms
of
like
taking
the
symbol
and
converting
it
to
something
else
we
could
throw,
or
is
that
going
to
mess
things
up
so
much
that
it's
not
worth
it
like?
You,
basically,
don't
get
the
exception.
You
through
you
get
something
else
which.
C
You
would
end
up
getting
something
else
right,
because
you
would
throw
a
symbol
and
then
we
could
like
stringify
it.
But
then
you
wouldn't
be
able
to
compare
it
to
what
you
threw
right,
where
you
could
be
able
to
do
it
with
everything
else
with
all
of
the
primitives
and
all
of
the
references
except
simple.
G
C
A
C
F
A
Know
it
may
not
be
comparable
and
then
once
it's
once
it
actually
works
the
you
know
the
same
across
all
of
them.
We
could
make
that
the
default.
A
C
So
the
the
behavior
would
be,
you
can
opt
in
for
some
stringification
of
the
symbol
so
that
it
would
continue
to
not
crash
or
you
don't
opt
in
and
you
get
a
crash.
A
C
Dave
is
asking
to
be
promoted.
Oh.
A
B
I
I
think
that
we
could
check
if
it's
possible,
to
obtain
no
so
to
to
add
the
options.
So
it's
not
a
developer
responsibility
check
the
version
node,
but
we
need
to
do
that
so
in
so
this
is
my
my
opinion.
So
we
we
we
check
if
the
feature
is
available,
okay
check
in
the
virtual
note,
and
then
we
check
if
the
option
is
enabled
or
not
and
yeah,
we
can
give
the
right
message
in
these
two
cases.
A
I
think
that's
what
I
was
thinking
that
yeah
like
we
would.
We
would
do
the
check,
but
only
let
symbol,
work
or
not
like
simple,
would
only
work
if
you've
opted
in
and
working
might
would
mean
it
would
only
work
if
we've
checked
and
it
would
like
if
we've
checked
and
it's
the
right
version
of
and
the
node
supports
it
it'll
just
work.
If
it's
a
version
of
node
where
it's
not
supported,
then
you
would
get
the
same
crash
as
before.
A
A
C
Instead
of
this,
an
api
create
reference
sure.
H
A
Yeah,
and
is
it
any,
is
it
breaking
at
all
like?
Would
anybody
I
guess
in
that
opt-in
we
would
want
to
make
it
clear,
but
don't
depend
on
the
like.
Don't
depend
on
getting
this
other
error
message,
because
at
some
point
that
may
go
away
and
you'll
actually
get
the
symbol.
You
expected,
I
think
that's
fair
enough.
A
A
Sure
yeah,
I
agree,
but
I
think
yeah
I
think
if,
in
that,
like
in
the
opt-in
flag,
it
can
be
very
clearly
like
hey.
This
is
you're
going
to
get
different
behavior
on
different
versions
of
node
and-
and
I
guess
like
already
that's
implicit-
that
going
forward
at
some
point-
yeah.
No,
it
would
be
we'd
have
to
say,
and
at
some
point
we're
going
to
flip
the
switch
so
that
you're
just
going
to
get
this
by
default.
C
G
A
G
C
G
But
not
12.2
to
handle
different
one
different
node.js
versions,
make
it
an
opt
in
make
it
into.
A
Well,
I
think
it
would
be
14
and
16,
possibly
oops.
I
guess
we
need
to
back
part
to
16.
A
A
A
C
A
F
Yeah,
maybe
yeah,
I
guess
what
one
dog
change
that
we
could
make
is
like.
It
says,
like
in
value
and
api
value,
representing
the
object
to
which
we
want
the
reference
we
could
say,
object
or
symbol
to
which
we
want
the
reference.
So
we
could
yeah.
This
api
creates
a
new
reference
with
the
specified
reference
count
to
the
object
passed
in.
We
could
say,
object
or
symbol
passed
in.
So
I
guess
there
is
room
for
improvement.
F
C
A
C
A
C
A
Okay,
so
on
that
one,
I
guess:
wait
a
sec,
so
this
one,
we
should
probably
still
leave
on
our
agenda
until
we
make
some
more
progress
on
like
fixing
the
node
api
side
right.
A
A
A
C
C
C
F
A
C
F
No,
I
don't
think
we
can
do
anything
yeah
I
mean
if,
if
people
are
stuck
with
c-style
arrays
of
an
api
values,
then
they'll
have
to
use
this
overload,
and
I
think
I
think
the
the
the
performance
implications
are
limited
to
this
overload,
and
you
know
people
will
come
to
face
them
when
they
do
right,
so
they
if
they
find
that
this
overload,
doesn't
work
too.
Well,
you
know
if
it
has
poor
performance,
then
you
know
they're
they're
gonna
have
to
work
backwards
and-
and
just
you
know,
make
the
make
the
arguments
fit.
C
C
If
you
scroll
down
just
a
little
bit
to
all
of
the
other
calls
here
this
yeah
this
one,
so
we
sorry.
E
C
F
F
A
F
F
Yep
yeah,
to
be
honest,
I
I
don't
know
what
what
an
napi
value
or
an
napi
colon
colon
value
looks
like
you
know,
in
memory,
you
know,
if
c
plus
plus
adds
any
any
overhead
to
it,
because
if
it
is
just
the
napi
value,
then
you
know
an
array
of
these
things
or
an
array
of
those
things
are
just
views
of
the
same
thing.
So
you
know
you
know
what
I
mean,
because
I
I
don't
know
if
I
don't
know
what
it
does.
F
I
don't
know
what
c
plus
plus
does
in
order
to
actually
render
an
instance
of
type
and
api
value.
Does
it
fit
inside
the
inside
a
core
napi
value
or
not.
F
Because
we
have,
we
have
like
an
implicit
casting
operator
and
so
forth,
but
you
know
in
terms
of
data:
that's
all
we
have.
On
the
other
hand,
you
can't
take
the
pointer
of
an
api
call
and
colon
value,
I
think,
in
which
case
it's
probably
stored
somewhere,
and
then
you
have
to
do
this
conversion,
so
I
don't
think
it
it's
c
plus
plus
is
magic
enough
to
actually,
you
know
just
consider
everything
related
to
an
n
api
call
and
call
and
value
to
be
met
with
that.
C
A
A
Saying
yeah
that
would
make
sense,
because
if
it
did,
it
should
already
work
in
fact
effectively
which
it
doesn't
yeah,
and
I
I
think
if
we
figured
out,
we
were
wrong
on
that
or
some
new
version
of
c
supported
that
then
it
would
be
maybe
a
matter
of
removing
this
and
it
would
still
work.
So
I
don't
think
we're
really
like
we're
not
getting
in
the
way
of
a
more
efficient
implementation
if
it
was
possible
in
the
future
either
so.
F
I'm
thinking
the
only
the
only
problem
with
that
is
that
you
know
it
may
not
be
able
to
deduce
the
type,
but
that
doesn't
mean
that
that
it's
not
a
view.
You
know
somebody
would
have
to
run
an
experiment
like
create
an
array
of
two
like
c
plus
plus
values
and
then
create
a
an
array
of
two
core
values
and
then
just
dump
the
data
and
see
if
it's
the
same
data,
you
know,
that's
you
know
you
can
you
can
do
a
quick
experiment
just
just
to
see
what
the
compiler
does
right
like?
F
Is
the
compiler
able
to
boil
down
an
array
of
c
plus
plus
values,
to
an
array
of
of
core
values,
or
does
the
compiler
you
know
create
like
actual
or
reserve
actual
memory
for
the
sequel?
Do.
C
A
Yeah,
they
can't
be
the
same
implementation
in
memory.
It's
just
if
there
was
some
way
that
the
compiler,
since
it
can
convert
a
value
to
a
and
not
you
know,
a
nappy
value
could
do
the
same
thing
for
an
array
right
and.
F
A
F
A
F
A
C
A
C
F
Yeah
exactly
yeah
yeah
yeah,
because
I
think
I
think,
if
you
use
like
an
initializer
list,
for
example,
because
I
was
just
thinking
like
okay.
If
I
support,
if
we,
if
we
add
support
for
a
c
style
array
of
c
plus
plus
values,
then
why
not
add
support
for
a
standard
vector
of
c
plus
plus
values
right.
F
F
Yeah,
maybe
maybe
we
should-
maybe
maybe
we
should
ask
since
they
already
submitted
this
pr.
Maybe
we
should
ask
whether
it's
worth
adding
a
standard
vector
over
overload
as
well,
which
basically
just
takes
the
standard
vector
data
and
feeds
it
into
this
function
right
because
it
would
be
a
super.
Easy
would
be
a
super
easy
thing
to
do
you.
F
A
A
You
know,
as
opposed
to
saying,
is
so
I'd
be
like
we
talked
about
it
today.
We
don't
have
any
concerns
in
landing.
If
you
know,
if
you
can
rebase
and
one
suggestion
is
like
you
know,
maybe
to
add
the
you
know
a
standard
vector
override
as
well,
either
in
this
or
a
future
pr.
F
G
F
G
G
A
Okay,
that
one's
good-
I
think
that
takes
us
to
the
end
of
that
list
and
okay.
So
let's
go
back
to
tracking
issues.
I
don't
see
any
reports
of
anything
new
there,
but
we
could
look.
A
Terms
of
enabled
debug
testing
still
haven't
done
anything
on
that
in
terms
of
creating
the
matrix
of
tested
methods.
Jack
did
you
have
anything
we
should
discuss
this
week
on
that
front.
D
D
A
H
Yes,
michael
yeah,
let
me
share
my
screen.
Yeah.
H
I
have
a
working,
a
simple
like
a
end-to-end,
you
know
filter
condition
that
is
working.
You
know
you
can
give
we
can
give
any
like
it
could
be
object,
wrap
or
any
file
name
right
for
now.
I
consider
the
filter
condition
as
a
single
file
name.
We
can
group
add
additional,
find
like
the
capability
to
add
additional
files
could
be
added
later,
but
for
now
it's
just
one
single
file.
H
It
adds
a
test
for
object,
wrap
as
well,
so
if,
if
I
change
to
something
else
like
say
function,
so
we
have
the
function.cc
here,
yep
yeah-
and
I
just
mentioned
that
as
function
or
I
can.
We
can
change
that
to
exact
file
names
like
function.cc.
H
That
is
possible.
What
makes
the
build
for
function
instead
of
object
right?
It
makes
the
build
for
function.
The
hello
is
just
something
that
I
added
it
can
be
removed
later
and
then
it
right
now
is
about
to
run
the
test
for
function
and,
in
case
say
just
to
verify
that
the
build
so
the
build
is
also
is
the
unit
just
build
right,
a
separate
build
that
is
done
for
unity
under
the
unit
test
folder,
and
that
is
what
we
are
running.
The
tests
on.
H
We
are
not
running
the
test
towards
the
root
folder,
the
build
in
the
root
folder,
but
because
we
had
the
capability
capability
to
have
the
relative
path.
That's
a
there's
another
pr
for
that,
but
still
non-merged.
But
actually
cherry
picked
the
change
here
so
because
of
that
I
am
able
to
run
the
test
against
another
build
a
location.
That's
under
the
unit
test,
folder
in
case
you
know
that
we
just
to
verify.
H
If
this
is
the
exact
build
we
are
running
on,
I
can
uncomment
this
right
and
then
run
against
function
right.
What
I'm
saying
is
I'm
going
to
make
a
build,
only
the
form
for
the
function.cc,
but
here
I'm
trying
to
run
the
test
for
object,
wrap
as
well.
It
will
fail
because
the
export
symbols
for
object,
wrap
is
not
there,
so
the
build
will
do
the
build,
does
not
have
the
exponentiables
for
object
crops,
so
it
would
fail.
The
test
would
fail
so
yeah
just
to
verify.
H
Right,
yeah,
that's
one
more
thing:
yeah,
we
also
have
the
directories
I
did
not.
I
have
not
considered
that,
so
I
have
to
make
sure
that
you
know
this
is
addressed
as
well.
We
have,
there
are
other
options
to
come
as
well
like
we
do.
We
want
to
run
the
entire
directory
in
an
entire
directory,
but
that
is
a
separate
question
for
now.
H
If
we
think
about
only
files
we
have
to
have,
we
have
have
the
capability
to
mention
this
directly
to
global
object,
slash
you
know
or
something
like
dataview
slash,
dataview.cc
right.
F
F
To
the
it
would
be
up
to
the
to
the
list
generator
like
the
the
you
know,
the
the
the
bit
of
code
that
generates
the
list
of
sources
right
right
would
be
up
to
that
to
to
create
the
correct
list
of
sources
given,
given
that
that
is
a
directory
right,
so
that
logic
could
be
built
into
the
thing
that
produces
the
list
of
sources.
H
F
H
But
are
you
saying
that
we
need
the
extension
joe
to
be
mentioned
in
some
way.
A
H
H
Yeah,
so
I
that's
yeah,
so
that
is
something
I
have
to
add
on,
like
okay,
that.
H
Yeah
that
wouldn't
work,
so
I
I
I
would
have
it
like.
You
know
a
global
object.
Eventually,
this
is
how
I
have
to
have
it
right.
Object
right,
right.
A
H
G
A
H
All
right,
yeah,
there
is
some
other
kind
like
this
pr.
I
have
some
other
cleanup
to
do
as
well,
like
you
know,
removing
this
all
this
map
that
I
had
up
previously
so
and
make
make
use
of
the
fact
that
we
have
the
you
know
the
file.
H
The
file
can
be
used
to
derive
the
export
object.
Now
with
our
changes.
So
I
can.
I
can
utilize
that
all
the
file
names
I
have
in
another
pr
that
I
reached
last
week.
I
have
changed
all
the
names
to
with
a
snail.
This
is
a
this
is
a
snake
case.
I've
changed
all
the
file
names
to
the
snake
case,
so
it's.
If
I
go
into
any
of
these
files,
the
you
know
it's
from
the
snake
case.
I
will.
H
I
can
just
remove
the
dash
and
change
the
first
letter
to
the
capital
and
derive
this
so
yeah
yeah.
There's
no
need
of
this
maps
anymore.
I
have
one
more
pull
request.
Michael
has
uploaded
this.
This
is
the
one
I
cherry
picked
into
my
apr
so
that
I
I'll
be
able
to
run
the
tests.
H
So
once
this
is
there,
then
you
know
I
can
just
what
I
do
is
I
just
part
the
past
this,
the
capabilities
that,
when
I
am
doing
a
unit
test,
so
there
is
a
test.js
under
the
unit
test
folder,
where
I
am
spawning
a
test.
There
could
be
a
better
way.
So
what
I'm
doing
is
I,
when
I'm
doing
that,
I,
when
I'm
spawning
the
child
process
for
running
the
test
specific
test,
I'm
passing
the
you
know
the
relative
build
path
as
the
path
under
the
unit
test.
H
So
that's
how
it
gets
injected
the
env
variable
gets
injected
there,
and
that
goes
into
into
the
test
when
the
when
the
actual
test
file
is
invoked.
What
happens
is
that
the
every
test
actually
refers
to
the
common
common
run
test
right
and
there,
since
we
have
the
build
root
path,
it
picks
up
from
the
env
variable
automatically.
H
So
this
pi
is
something
that
I
you
know
I.
If
this
is
much,
then
I
can
just
start
working
out
the
rest
of
the
stuff
when
I
close,
probably
add
the
capability
for
the
folder
by
next
week
and
probably
by
next
year.
I
think,
by
our
next
meeting.
We
should
be
ready
to
be
reviewed
and
closed.
A
B
G
A
B
For
me,
it's
good
and
I
can
make
ci
to
be
sure
that
we
don't
land
something
bad
but
so
yeah.
For
me,
it's.