►
From YouTube: 2020-12-11-Node.js N-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
Welcome
to
the
node.js
napi
team
meeting
for
november
11th,
sorry
december
11th
2020.
we'll
go
through
our
regular
agenda,
which
is
to
look
at
the
things
we've
tagged
for
milestone
11.
before
we
do
that.
Does
anybody
have
any
announcements
or
other
things
they'd
like
to
share.
B
A
That
the
one
related
to
like
like
was
it
like
special
add-ons,
correct.
B
Yeah,
like
I
man,
maybe
as
we
discuss
I'll,
bring
up
the
link
and
if
we
have
time
later,
I
I
can
share
the
link.
A
B
A
Issues
that
were
tagged,
let
me
share
so
that
people
can
see.
Where
is
the
share
screen
there?
We
go.
A
A
So
the
first
one
that
was
tagged
is
use
stale
bot
for
node
add-on
api.
That's
one
that
I
added
recently
I
I
did
a
mini
cleanup.
A
We
had
like
100
old,
older
older
issues,
and
I
wondered
if
maybe
we
should
turn
on
stale
bot
which
is
being
used
in
some
of
the
other
projects
and
that
basically,
after
a
certain
amount
of
time,
says
hey
this.
This
looks
stale,
I'm
going
to
close
it
unless
somebody
comments.
A
B
Yeah,
I
think,
and
then
it
gives
the
it
gives
the
originator
the
option
to
reopen
it.
I
imagine
all.
A
A
Comments
it,
it
doesn't,
you
know,
become
stale
there's
also,
I
think
things
where
you
can
put
like
a
never
stale
tag
and
stuff,
so
it
kind
of
if
nothing
else
is
also
a
reminder
to
us
like
hey.
This
is
something
that's.
We
should
go
back
and
take
a
look
at
because
I
often
drive
based
on
like
recent
notifications.
B
A
A
Okay,
if
there's
no,
it
sounds
like
everybody
thinks
that's
a
reasonable
idea,
so
I'll
try
and
figure
out
how
to
do
that.
I
think
it.
A
Hard
yeah,
okay,
the
next
one,
is
why
the
source
node
api
and
node
api
types
files
were
moved
to
the
node
repository.
This
was
actually
opened
in
node
add-on
api,
but
I
moved
it
over
here,
so
we
could
put
it
on
the
agenda
and
I'll
move
it
back.
A
I
think
you
know,
as
as
everybody
will
remember,
but
for
those
who
may
be
watching
on
the
stream
we
used
to
have
those
files
in
node,
node
add-on.
Api
is
a
bridge
for
older
versions
that
didn't
support
an
api
at
all.
They
were
a
bit
of
a
challenge
to
keep
up
to
date
and
so
forth.
A
And
that
kind
of
makes
sense
to
me,
but
not
not
as
part
of
noted,
on
api
itself,
though,
but
maybe
somewhere
else.
So
I
don't
know
what
people
think
about
that.
That
would
take
us
some
amount
of
work
to
put
a
copy.
Like
I
don't
know
I
was
thinking
like
it
might
make
sense
to
have
one
which
matches
each
of
the
specific
versions
like
an
api
7
and
api
6
and
api
5..
A
We
don't
actually
maintain
anything
that
actually
looks
like
that.
We
have
one
cumulative
one
where
the
define
lets
you
choose,
which
one
you
want,
and
maybe
that
still
just
makes
sense
like
it's.
It
would
basically
be
a
copy
of
whatever
the
latest
is
in
master,
maybe
without
the
experimental
section.
D
That's
that's
that's
complicated
right
because
you
know
the.
If
you
want
a
snapshot
of
what
it
was
at
time
of
publication,
then
then,
okay,
but
whatever
it
was,
it
is
already
present
in
in
the
latest.
I
mean
they
are
cumulative
right,
so
the
only
the
only
benefit
would
be
that
that,
let's
say,
for
example,
at
the
time
an
api
four
was
published
right
yeah,
you
know
things
like
nappy.
Freeze
object
didn't
even
exist,
even
in
experimental.
D
So
so,
if
you
turned
on
like,
if
you
turned
on
nappy
experimental
with
the
latest
version,
as
is
present
on
master,
then
you
would
get
nappy
freeze
object.
But
if
you
turned
on
nappy
experimental
against
the
version
four
published,
then
you
wouldn't
so
things
would
be
absent
and
would
fail
to
compile.
A
D
Well,
you
don't
you
have
to
strip
it
out
right,
but
the
the
definition
of
what
is
experimental
was
different.
When
an
api
4
was
published,
then
then
it
is
when,
when
an
api
7
was
published
right,
they
were
all
published
with
some
experimental
bits
right,
but
when
an
api
4
was
published,
certain
experimental
bits
which
are
experimental
today
didn't
even
exist
back
then
right
right.
A
A
D
A
D
A
A
D
Basically,
sort
of
the
the
official
and
api
headers.
A
So
the
requirements
to
use
the
napi
part
only
and
not
don't
need
all
the
node
project
api
headers,
because
yeah
nick
pointed
out,
you
could
just
go,
get
the
headers
and
you
get
what
you
need
right,
yeah,
and
so
he
pointed
us
to
nodejs.org
api
npp.
It
takes
an
api,
is
part
of
the
engine,
abstraction.
D
Okay,
yeah,
I
mean,
I
guess
you
can
link
against
you.
You
can
use
the
the
headers
and
then
link
a
shared
object,
yeah
and
because
the
shared
object
doesn't
have
to
have
all
the
symbols
resolved
right.
So
so
you
could
just
link
sort
of
in
a
vacuum.
C
D
It
would
just
boom
work
with
node.
Whenever
you
linked
against
node,
I
mean
dynamically
linked
against
node.
A
A
D
You
know,
and
then
you
would
have
your
own
build
system
that
would
that
would
use
it
right
because,
like
you,
would
you
probably
wouldn't
use
jip,
because
that
would
require
node
right?
You
would
probably
have
like
cmake
or
something
and
then
cmake
doesn't
require
node
at
all
to
be
present.
A
E
A
D
A
D
I
they
they
are
intentionally
separated
yeah,
which
is
what
I
thought,
which.
D
Also,
I
just
I
just
had
a
thought
like
we.
I
don't
think
we
need
to
edit
them
in
any
way.
We
don't
need
to
strip
out
the
experimental
stuff
because
it
doesn't
matter
what
the
experimental
stuff
is
right
at
the
time
of
publication.
We
could
just
document
that,
yes,
it's
mentioned
there,
but
I
mean
it's
it's
up
to
the
it's
up
to
the
person
who
and
ultimately
ends
up
loading
the
ad
on
whether
that
stuff
is
available
or
not
right,
and
so
so
people
who
people
who
want
this.
D
Yeah,
that's
right,
you
have
to
you,
have
to
pre
precede
the
inclusion
of
the
header
with
number
define
not
be
experimental.
If
you
wish
to
use
experimental.
C
A
D
Yeah,
I
was
just
going
to
say
I
no,
because
we
don't
only
need
to
release
a
new
version
of
no
dead
on
api
when
we
have
a
new
version
of
an
api.
So,
yes,
okay,
we
are
kind
of
forcing
ourselves
to
make
another
release,
but
only
with
each
version
of
an
api
right,
because
if,
if
the,
if
any
api
doesn't
change
between
node
major
releases,
then
then
it
doesn't
matter.
The
header
stays
the
same
right.
Even
even
if
the
implementation
changes
these
folks,
they
don't
care
about
the
implementation.
D
C
Yeah
this
sounds
good,
but
so
were
we
having
different
tags
for
the
nappy
versions?
D
C
A
C
A
C
A
A
A
Have
the
mpem
package
contain
the
following
files,
which
are
just
those
ones
which
define
the
function
types
available
from
api
experiment?
Functions
may
be
document
files,
but
the
define
should
not
be
used
to
expose
them.
A
A
A
A
A
A
D
A
E
A
D
D
So
if
we
say
node
api
headers
people
might
get
the
wrong
idea.
Although
the
node
api
package
is
being
downloaded
32
times
a
week,
so
it
may
not
be
such
a
big
problem.
What
is
what
is
node
api
anyway?
It
says:
node.
Api
lazy
require
core
modules.
This
is
a
trivial
module
that
enables
you
access
core
node
modules
without
the
need
to
require
them.
It
exports
a
single
object,
with
properties
corresponding
to
the
core
module
names.
The
properties
per
form
are
required,
the
first
time
they
are
accessed.
A
Right,
I
don't
know,
I
don't
think
if
that's
that
little
use
there'll
be
that
much
confusion
necessarily
right.
B
Could
we
could
we
scope
it
like
at
node,
slash.
C
Scope
does
not
just
I
do
not
just
the
scope
is
limited
to
the
possible
future
internal
packages,
asking
discussions
on
the
logical
repo
that
yeah,
but
you
might
use
a
potential.
I
did
not
just
say.
A
A
D
A
Okay,
so
that's
good
review
external
examples.
I
haven't
done
anything
on
that,
so
I
don't
have
an
update.
I
haven't
enabled
debug
testing.
A
I
don't
know
the
node
sas
api
port
if
that's
still
underway
or
unfortunately,.
A
Okay
is
that
something
we
should
still
leave
on
the
agenda
or
yeah?
And
I
I
don't
know
if
it's
worth
it:
okay,
let's
close
that
out
and.
A
Okay
modules
that
were
ported.
D
A
A
Seems
like
it's
on
an
upward
trend
still,
which
is
good.
There
was
a
bit
of
a
blip,
I
don't
know
what
happened,
but
I
I
looked
at
other
modules
and
they
seem
to
have
the
same
blip.
So
I
don't
know
if
it
was
in
tracking
or
it
was
a
quiet
time
or
what,
but
it
seems
to
be
back.
A
A
B
A
B
D
D
A
A
B
A
A
A
D
A
A
I
think
the
only
other
thing
so
we've
talked
about
all
the
other
ones.
I
think,
except
for
are
there
any
issues
in
look
at
that
any
ones
that
people
want
to
bring
up
issues
in
terms
of
issues
or
anything?
We
should
discuss.
D
A
C
Yeah
I'd
like
to
discuss
this,
maybe
okay,
the
fix
will
be
since
the
issue
is
not
caused
by
this
condition
is
not
happening
and
not
add-on
api
itself.
It's
on
the
node
internal
pre
implementation.
C
Every
time
the
risk
condition
happens
in
the
inaudible
that
there's
a
function
it
will
where
the
function
receives
and
send
on
the
user
functions
it,
it
will
trigger
a
uv
async
and
the
everything
will
be
called
the
quebec
of
the
thoracic
function
right,
the
third
function.
The
other
function,
then,
will
be
calling
initiating
a
uv
ito
handle,
and
maybe
I
am,
it
will
be
better
to
show
the
link
posting.
The
discussion.
A
C
I'm
I'm
posting
a
link.
Okay,
all
right,
I'm
positive!
Okay,
thanks!
So
in
the
the
implementation
per
the
splash
function
itself,
it
it
seems
okay
to
in
this
condition.
That's
the
the
function
is
differing
the
code
to
the
interface,
but
that
caused
a
gap
between
the
async
t,
callback
and
the
actual
user
function.
Callbacks.
C
A
C
There's
the
condition
happens
in
the
before
the
actual,
not
that
only
api
called
executed.
So
there
is
no
chance
that
the
know
that
only
appear
to
prevent
the
risk
energy
but
to
the
start
to
function
the
same
each
go
back
in
it
is,
they
are
expected
and
it's
not
unexpected
behavior.
So.
C
E
D
We
can
hear
you
now
the
next
progress
item
thread
eventually,
unless
the
there
should
never
be
a
situation
where
the
function
implementation
may
drop.
One
of
the
progress.
D
A
B
C
D
D
I
I
don't
know
honestly,
I
don't
remember
anymore
what
what
the
state
was
back
then,
but
I
I
don't
know,
did
you
did
you
get
my
question
because
I
got
disconnected
as
I
was
being?
Oh,
we
didn't
get
your
question
okay,
so
my
question
was:
if
the
if
the
async
progress
worker
calls
the
main
thread
right
and
every
time
it
calls
the
main
thread,
it
passes
a
parameter
to
that
call.
D
C
When
the
either
before
the
maybe
we
can
back
to
the
explanation
of
the
we
came
back
to
the
issue,
I
have
posted
a
graph
of
the
event.
C
A
C
So
when
first,
there
is
a
stand,
progress
in
the
single
progress
work
and
the
uvsync
key
has
triggered
the
async
callback
and
we
are
awaiting
the
either
callback
and
there
is
a
another
call
of
the
10th
progress.
So
there
is
another
chance
that
uva
sanctity
is
is
able
to
trigger.
But
when
the
same
progress
b
happens,
we
have
replaced
the
replaced
slot
with
with
the
latest
data
and
after
the
stand,
progress
b,
we
are
actually
calling
into
the
module.
So
in
this
first,
according
to
the
module,
we
have
no
idea
that
this
is.
C
Of
the
the
second
compact
of
the
uv,
synthetic
and
the
either
go
back
and
call
it
according
to
the
node
guy,
so
the
async
progress
worker.
There
is
nothing
since
we
have
already
sent
the
data
b.
A
So,
basically,
we,
while
the
call
to
read
something,
has
already
been
started.
We
we
overwrite
what
it's
going
to
read
and
it
reads
the
later
value,
but
then
there's
all
there's
already
another
thing
triggering
it
to
say:
hey.
There
was
some
more
progress
and
when
that
second
one
goes
it
it
finds.
There's
nothing.
D
A
D
This
is,
this
is
a
bit
of
a
problem,
because
the
whole
point
of
threat
say
function
was
that
it
does
not
drop.
It
doesn't
coagulate
right
because
uv
async,
t,
coagulates
right
and
the
point
of
threat
say
function.
Was
that
there'd
be
a
q
right.
So
if
it
drops
things,
then
that's
very
bad
for
first
threat
say
function
right
progress,
but
progress.
A
A
D
Okay-
okay,
all
right!
Yes,
no!
No!
Definitely
I
understand
that.
I
understand
that
the
semantics
of
the
of
the
async
progress
worker
are
sort
of
best
effort
and
you
get
as
many
progress
reports
as
will
fit.
But
what
what
I'm
concerned
about
is
whether
the
the
the
fact
that
that
one
of
the
progress
reports
gets
dropped
is
that
because
of
the
implementation
of
threat
safe
function,
because
if
it
is,
then
we
have
found
a
scenario
where
a
threat
say
function
drops
one
of
the
calls
and
that
it
should
never
drop
a
call
right.
C
D
Yeah,
yeah,
okay,
okay,
all
right,
okay,
yeah!
So
if
the
problem
is
confined
to
to
async
progress
worker,
then
then,
then
that
puts
my
mind
at
ease,
because
that
means
the
fix
will
also
be
here
and
and
and
thread.
Save
function
need
not
be
changed
because
that's
that's
always
a
big
hassle,
because
we
need
to
like
maintain
compatibility
and
stuff.
D
A
A
C
It's
nothing
it's
kind
of
similar
blue.
C
Yeah
there
is
a
test
for
the
file
for
the
case,
so
the
test
will
always
always
in
my
local
machine
that
will
fail
in
previous
implementation.
So
I
have
to
ensure
that
the
test
actually
succeeds
in
the
with
the
change.
B
C
The
test
will
fail
before
the
change
and
it's
succeeded
with
every
time
with
the
perfect.
C
C
I
have
another
one:
okay,
I
I
think
that
we
have.
We
have
travis
travis
ci,
not
running
for
a
long
time.
C
A
A
A
Okay,
okay,
thanks,
sorry,
yeah,
okay
and
the
the
github
actions,
one
will
will
move
over
any
other
ones.
People
want
to
call
out.
D
There's
there's
one
in
core
about
a
possible
new
api.
This
is
this.
This
would
be
december
minor,
okay
from
from
gus,
I
think
which
new
api
is
it.
Let's.
A
D
Caller,
that's
yes!
Yes,
okay,
yeah!
We
have
this.
We
have
this
sort
of
functionality
in
v8,
where,
where
you
can
we
have
this
watchdog
in
core
where,
where
you
can
like
tear
down
the
execution,
if
it's
taking
too
long
and-
and
so
you
know
this,
this
is
something
you
can
check
for
much
as
you
can
check
whether
there's
an
exception
pending.
A
Okay,
you
can
also
so
it's
related
to
cancellation.
The
execution.
D
A
A
D
D
D
It's
not
it's
not
as
strong
a
case
as
if
it
were
specified
like
an
ecma
script,
but
there
are
things
in
js
native
api
that
are
not
part
of
acma
ecmascript
like
external
muscles
and
references,
and
that
kind
of
thing
that
those
are
engine,
things
that
that
most
engines
have.
So
I
think
this
is
okay,
because
any
engine
can
have
this
and
just
always
false.
If
they
don't
support,
you
know
setting
a
flag
for
interrupted
execution.
A
A
E
A
A
D
D
Yeah,
it's
a
larger
pr
than
just
an
api,
and
so
so
there's
some.
You
know
there's
some
discussion
going
on
as
to
whether
this
should
just
be
part
of
vm,
because
vm
already
has
this
kind
of
functionality
right.
So
so
it,
but
that's,
that's
that's
out
of
our
scope.
A
A
A
A
A
A
A
A
If
not
thanks
for
everybody's
time
and
a
productive
meeting,
as
always-
and
I
guess
the
question
is,
is
anybody
gonna
be
around
two
weeks
from
now?
If
not,
we
might
just
choose
to
cancel
and
meet
back
in
january.
A
What
is
next
week,
so
I
meant
two
weeks
from
now
or
next
week.
You
actually
know
that's
true
we're
every
week.
So
maybe,
let's
quickly,
look
I'm
good
for
the
next
week.
A
D
A
B
And
the
week
after
that
is
new
year's
day.
A
A
B
A
Okay
sounds
good,
so
we'll
see
everybody
next
week
and
then
hopefully
everybody.