►
From YouTube: Async Foundations WG: post-stabilization triage
Description
In this meeting we go through the bugs that we deferred until after stabilization, discussing which of them to prioritize.
A
A
Are
sort
of
two
questions
that
we
could
be
talking
about,
but
one
of
them
is
bigger,
which
is
like
what
now
that
async/await
is
stable,
it's
a
good
time
to
sit
back
and
look
at
the
whole
a
single
effort
and
figure
out.
How
do
we
want
to
structure
it?
What
should
I
prepare
needs
me?
I?
Don't
really
want
to
do
that.
In
this
moment
we
had
some
thoughts
about
how
one.
A
A
A
September
26
is
like
when
the
beta
branches
that
carries
basically
so
and
it
seems
worth
talking
about
what
can
we
try
to
do
before
then,
which
will
affect
users
first
impression
of
async/await
like
when
it
hits
stable
and
then
maybe
there's
some
things
we
want
prioritised
before
that?
One
of
the
things
obviously
backwards-compatibility
bug
fixes
and
things,
but
hopefully
you
don't
have
many
of
those
and
then
afterwards,
I
guess.
A
And
I
don't
know
if
I
guess
what
I
we
can
do
is
just
start
walking
through
the
issues,
one
by
one,
it's
kind
of
what
we
did
play
back
when
we
started
this
up
and
then
see
what
we
think
and
maybe
some
categories
start
to
emerge.
My
one
proposal
is
that,
before
the
beta,
my
main
thought
is
obviously
back.
Compact
makes
sense.
B
A
A
A
B
A
B
A
B
A
A
A
B
A
A
Of
in
between
I
think
there's
some
variance
on
this.
That
probably
like
it's
I
know
why
you're
saying
it
you're
saying
it's
in
scope
and
I
think
that's
reasonable.
On
the
other
hand,
like
a
non
lexical
lifetime
certainly
considers
stuff
like
whether
you
have
drop,
and
in
other
words,
if
this
held
on
to
a
borrow,
the
borrow
would
be
you've
been
about
a
tomorrow
at
the
point
of
a
week.
So
there's
a
consistency,
argument
and
I
think
it.
Certainly
if
you
added
a
drop,
that's
just
in
some
sense:
further
alike.
A
B
There's
sort
of
two
layers
in
this
of
the
there's
things
that
over
approximate
and
don't
and
and
don't
correctly
talk
about
things
that
don't
wind
up
in
the
generator
state.
This
one
actually
does
wind
up
in
the
generator
state.
So
this
is
like
an
even
layer,
further
rate
of
like
taking
additional
things
out
of
the
generator
state.
B
B
A
Like
over
capture
and
a
particular
variation
of
those
over
approximation
where
the
static
or
the
like
premiere
now
assess
determine
yeah
that
we
will
capture,
but
we
do
not,
and
then
there's
we
are.
We
do
capture
I,
don't
know.
There's
over
capture
is
the
more
general
version,
I
guess
where
we
yeah
you're
actually
does
capture
what
does
that
needs
not
shouldn't
yeah,
okay.
B
And
that
one's
a
little
hard
in
terms
of
its
interaction
with
you
CG,
because
you
start
having
to
categorize
things
like
if
you
turn
something
into
a
Rob
pointer
like
a
reference
to
something
into
a
row
pointer.
How
long
can
you
expect
that
Rob
pointer
to
be
valid
its
access
that
members
of
that
I
mean
grow
it
turns
into
today?
What
we
do
is,
if
you
just
ever
take
a
reference
to
the
type
it's
assumed
to
leak.
B
A
And
I
were
discussing
that
so
there's
you
CG
interactions,
there's
like
the
of
course
number
six,
three
eight
one
eight,
but
then
there's
also
what
you're
saying
here
is.
It
brings
into
relief.
Let's
say
I
mean
the
same
question
arises
outside
of
of
a
few
key
on
futures,
but
if
you
make
a
raw
pointer
to
a
local
variable,
then
move
from
that.
A
B
A
Yeah
you're
right,
there's
also
the
question
of
what,
if
the
variable
is
dead
but
not
moved
from
right.
So
in
that
case,
what
you're
saying
here
is
now
I
start
to
understand
better
your
point,
you're
saying:
if,
if
I
had
taken
the
address
of
non,
send
and
coerced
it
to
a
raw
pointer,
it
might
appear
to
be
dead,
but
I
might
still
be
using
the
raw
pointer
and
correct.
Therefore,
if
I
were
eliminated,
the
stack
slot,
the
raw
pointer,
would
be
invalid
and
that's
interesting,
yeah.
A
B
A
B
A
Let's
call
it
late
Diagnostics
or
died.
There
are
a
point
of
use
to
the
ideas
like
when
a
future
is
not
send,
but
must
be.
Can
we
point
into
the
source
of
the
future,
not
just
the
site
of
use?
Yes,
okay,
I,
don't
think
I'm
kind
of
assuming
I'll
do
a
second
pass
and
try
to
fix
up
the
actual
issues,
and
once
we
figure
out
once
we
have
this
paper
document
working
good,
okay
in
line
attribute
on
async
FN
doesn't
work
properly,
I.
A
A
Yeah
I
think
my
expectation
would
be
that
it
in
lines
the
creation
of
the
future,
which
honestly
I
sort
of
expect
to
always
like
we
probably
should
mark
all
of
those
functions
in
my
nose,
because
then
they
just
make
a
struct
right.
That's
all
they
do
yeah
after
the
arguments
that
there's
really
no
reason.
A
A
You
could
I
feel
like
we
could
just
make
in
mining
engineer
and
we've
with
the
probably
yeah
I'm
for
now
until
there's
some
meaning.
We
intend
that
we
think
is
useful
I,
especially
in
line
always
wish.
I
mean
I
always
view
in
line
as
a
hint
anyway,
but
it
might
always
has
like
that,
a
feeling
of
something
you
can
rely
on.
A
B
A
They
say:
there's
a
lot
of
cases
where
we
currently
kind
of
leak,
the
D
sugaring,
maybe
more
than
we
have
to,
and
try
essentially
walking
through
some
initial
examples,
making
some
deliberate
mistakes
like
past
an
argument
of
the
wrong
type
or
forget
the
await.
Look
at
the
errors,
see
what
we
can
do
about
those
scenarios.
I
think
it's
a
great
idea.
If
someone
has
time
and
energy,
that's
the
question
like,
but
the.
B
A
B
A
I,
don't
know
how
to
trick
it
either,
but
okay,
we'll
leave
it
at
that.
Maybe
doesn't
belong
on
this
list
of
all
more
of
a
miscellaneous
edition.
The
problem
with
these
area
categories.
They're
often
there
are
things
like.
Yes,
it
is
basically,
but
maybe
this
just
isn't.
A
single
item
is
particularly
good.
It's
just
a
metric
on
the
full
tape
cannot
infer
type
when
using
question
mark
in
async
close
bag.
Diagnostic
yeah.
B
B
A
B
And
it's
just
I
mean
I.
Do
it
all
the
time
and
it's
annoying?
But
it's
say
men
agree.
It's
annoying
like
this
is
one
I
thought
we
were
gonna.
Have
this
fixed
with
one
of
the
things
I
thought
we
could
do
to
fix?
This
was
like
making
it
so
that
you
could
syntactically
specify
the
output
type,
but
that
was
sort
of
caught
up
in
like
what
is
happening
with
try
books,
yeah,
try
blocks
and
also
whatever
they're
called
Oh
type,
but
the
colon
thing.
Oh
my
gosh,
or
it's
not
description.
A
A
B
A
A
quick
question:
do
you
all
see
a
notification
than
the
thing
I'm
sharing
right
now
do
I
see
a
notification.
I
just
got
a
notification
and
I
was
curious.
If
you
could
see
it
was
there
for
a
part
of
the
recording
or
if
I
guessing.
That
answer
is
no.
You
would
have
seen
that's
good,
because
I
hate
having
to
scrub
yourself.
A
A
B
B
B
A
B
A
Really
an
entrance
failure
because,
or
it
isn't
necessarily
an
inference
failure.
The
question
mark
yields
up
the
original
code.
Snippet
is
somewhat
under
specified
in
that
the
question
mark
on.
In
that
the
return
type
return
type
of
the
async
block
can
be
any
error
type
II,
but
the
error
value
can
be
into
into.
A
B
A
B
Question
mark
in
a
sink
box
requires
explicit
annotation
of
air
tight
bond.
Okay
return,
really
I,
don't
know
like
I,
think
you
see
where
I'm
going
with
this
right,
like
you
give
something
that's
like
really
specialized
just
so
that
people
can
be
like.
Oh
I,
know
exactly
where
I'm
supposed
to
go
to
find
out
more
about
this
yeah.
A
B
B
A
B
A
A
A
B
B
B
A
A
B
A
B
B
Mean
that
is
actually
what
it
like.
It
is.
It's
the
literally
expected
opaque
type
found
a
different,
opaque
type.
I
feel
like
one
thing
that
we
could
do
is
we
could
fix
the
whole
diagnostic
to
where
it
says,
expected
type
impulse,
ooh
found
type
in
both
ooh.
We
could
just
get
rid
of
that
and
said,
like
I
found
two
different
imple
foods
or
something
yeah.
A
B
B
A
A
B
A
A
A
So
this
came
down
to
like.
If
you
look
at
the
constraints
there
aren't
really
enough.
We
currently
have
a
rule,
as
you
recall,
Taylor,
I'm
sure
that
we
sort
of
pick
in
some
cases.
We
do
arbitrarily
pick
things,
but
we
like
to
take
the
smallest
one
in
this
particular
case.
What
we
would
have
to
think
is
like
the
biggest
one.
Maybe
we
can
make
get
real
a
little
more
arbitrary
than
it
already
is,
but
that
would
be
it
may
be.
A
A
A
A
A
A
B
A
B
A
B
A
B
A
B
A
A
A
A
A
B
A
B
A
Well,
it's
not
really
an
estimate,
but
only
in
the
common
I
think
it'd,
be
okay
to
close
it
and
or
like
market
is
a
dupe
of
some
talk
about
closures.
I
mean
it
definitely
not
a
ideal
error,
but
it's
not
even
that
I
guess
I
guess
the
bug
might
be
saying
in
a
function
if
we
said
cannot
be
used
in
an
async
walk
or
can
only
be
used
in
an
async
buck
and
not
a
function.
That
would
be
less.