►
From YouTube: 2021-06-15 Lang Team Triage
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
B
All
right,
so
this
is
the
lang
team
triage
meeting
and
let's
get
started
with
the
upcoming
scheduled
meetings.
We
have
three
meetings
listed
here.
However,
nico,
do
you
want
to
talk
about
the
structural
equality
item
for
a
second.
C
Yeah,
I
was
a
little
overly
ambitious.
I
was
busy
all
last
week
and
I
didn't
have
time
to
do
any
write-ups,
so
I
think
we
should
just
defer
this
one
till
next
time
I'll
try
to
get
a
nice
write-up
done.
We
started
it
but
only
started
it
today
and
it
seemed
like
it
wasn't
going
to
be
good
enough.
C
I
don't
know
felix,
isn't
here,
I
doubt
as
far
as
I
know,
felix
hasn't
had
time
to
poke
at
it.
I
don't
think
we
can
move
it
up.
B
Okay,
then
it
sounds
like
we
should
cancel
june
16th
as
a
meeting
and
then
the
other
two
meetings
can
either
stay
as
they
are
or
get
swapped,
depending
on
which
one's
ready
by
the
23rd.
C
A
B
That
works
out
well,
and
there
is
an
item
here
for
the
meetings
that
we
already
had.
I
will
get
to
the
highlighted
item
and
justice
act,
but
there
is
a
there's,
an
entry
here
for
meetings
that
we
already
had
and
did
you
have
a
specific
item
for
wanting
to
push
minutes
and
notes?
Are
there
minutes
and
notes
from
these
meetings
that
we
should
be
uploading.
C
Usually
I
try
to
keep
them
in
the
hack,
md
docs.
I
can
probably
do
that
as
part
of
my
post
line
team
updates,
I
may
even
have
already
done
it,
in
which
case
they
just
need
to
be
closed.
I
think
I
was
trying
to.
I
was
hoping
esteban
would
write
a
blog
post
about
generators
stuff,
but
he
had
some
family
emergencies
arise.
So
he's
been
out
of
touch,
that's
okay.
That
can
still
happen
later.
We'll
close
the
issues.
B
Okay,
then,
for
status
updates,
we
need
to
do
a
status
conversation
at
some
point.
That
said,
it
occurs
to
me
that
we
didn't
really
have
a
status
meeting
last
week
and
I'm
wondering
if
it
might
make
sense
for
the
june
16th
meeting
to
have
a
status
conversation.
If
we
can
be
ready
to
have
that
conversation.
B
B
It
seems
that
I
think
that
makes
sense.
So
in
that
case
we
have
more
time,
but
nonetheless,
if
we
can
get
somebody
to
volunteer
to
this
is
volunteer
to
ping
groups
for
status
updates.
But
I
would
say
ultimately
if
we
can
get
somebody
to
volunteer,
to
update
some
of
the
automated
scripting,
that
we
have
to
actively
ping
group
zulip
streams,
then
that
would
be
ideal
yeah.
B
C
I
should
reach
follow
up
with
them,
maybe
because
I
was
working
with
someone
who
was
they
didn't
work
on
that
specifically
yet,
but
that
was
kind
of
what
I
was
leading
them
up
to.
Okay,.
B
B
Right,
I'm
just
acutely
aware
of
the
fact
that,
even
when
we
managed
to
find
somebody
to
get
the
task
done,
it
feels
like
nico
you're
still
on
the
hook
for
coordinating
it.
If
we
can
find
somebody
to
step
in
and
coordinate,
then
that
would
more
effectively
distribute
the
load
agreed.
C
B
C
A
good
idea
would
be
to
open
a
pr
just
writing
out
the
process
which
isn't
that
complicated,
but
where
to
look
for
the
list
and
all
that
kind
of
stuff
right
mark,
you
were
gonna,
say
something.
D
C
It
gives
at
least
some
amount
of
contact
from
the
laying
team
on
a
regular
basis.
So
that
was
something
I
had
thought
about
discussing
in
with
the
process
proposal
or
at
some
other
future
point.
But
I
think
that
would
be
a
better
system
overall,
and
maybe
it's
worth
talking
about
yeah.
D
That's
nothing
true.
In
any
case,
I
think
we
can
move
on.
I
don't
know
if
we
have
a
volunteer
for
this
cycle,
but
I
think
just
pinging,
I
don't
know
it
doesn't
seem
difficult.
I
I.
D
Take
on
just
the
pinging,
but
not
chasing
people
right.
C
D
Okay,
I
I
I'll
for
a
follow-up
on
deal.
Then
yeah.
C
B
B
And
next
up
pending
language
team
project
proposals,
so
I
don't
think
we
need
further
conversation
on
lane
team
86,
the
write-up
is
started,
but
is
still
ongoing
and
no
progress
since
then,
and
then
the
negative
impulse
integrated
into
coherence
is
new.
The
current
status
of
this
is
that
somebody
already
seconded
it
anything
to
be
discussed
about
this.
B
C
Well,
I
don't
know
what
you
said
last
time,
but
the
idea
here
is
to
support
the
ability
to
write
negative
impulse,
which
we
now
have
but
allow
it
to.
I
guess,
like
it
says,
be
integrated
into
coherence.
It's
a
fairly
conservative
approach,
working
a
little
bit,
I'm
working
a
little
with
jane
on
implementing
it.
The
motivation
is
that
it's
it's
really
needed
to
overcome
some
of
the
like
limitations
around
the.
C
D
C
So
I
guess
I
would
probably
say,
like
I
feel
good
about
going
forward
with
it.
Obviously
there
would
be
an
rfc
for
before
it
could
be
stabilized
right
or
and
before
when
we
get
far
enough.
I
think
we're
like.
C
E
C
Like
I
don't
know
why,
maybe
this.
E
To
clarify,
I'm
not
saying
that,
like
I
think,
specialization
is
the
right
tool
for
this.
I'm
saying
that,
like
the
the
thing
that
you're
trying
to
write
in
the
downstream
crate
looks
like
a
thing
that
I
would
have
like
traditionally,
if
someone
told
me
they
wanted
to
write
that
and
rust,
I
would
have
told
them
that
the
tool
they
were
waiting
for
was
specialization,
where
you
have
like
a
specific
type
impul
and
a
a
blanket
impul
that
are
potentially
overlapping
right.
C
E
Yeah,
I
understand
that
this
is
sort
of
a
different
special
case,
because
you're
not
actually
specializing
the
blanket
impul.
This
is
giving
you
the
information
to
know
that
the
blanket
dimple
doesn't
apply,
which
is
a
little
bit
of
a
different
thing.
It's
sort
of
closer
to
what
we
have
with
like
fundamental
traits
and
fundamental
types.
Yeah,
that's
right,
but
but
also
like.
I
don't
know
the
it
that
this
looks
this
like
sort
of
shape
of
thing
where
you
have
a
blanket
impul
and
then
also
like
more.
E
You
want
to
write
some
more
specific
impulse
for
specific
types
like
it
would
be
kind
of
nice
for
that
to
just
like
work.
I
guess
I
don't
know,
I
don't
know
what
I'm
saying
necessarily
other
than
that
like
like.
There
are
a
lot
of
times
where
I
wanted
to
write
this
type
of
code
and
building
intuition
around
like
when
I
have
to
write
default
versus
not
and
when
I
want
a
negative
impul
versus
a
specializing
in
bowl.
Like
I
don't
know,.
C
C
It's
an
interesting
point
in
that
we
have
some
specialization
soundness
bugs
and
other
things
like.
There's
a
need
to
prioritize
progress
towards
specialization,
and
I'm
definitely
in
favor
of
making
thing
like
these
kind
of
impulse
figuring
out.
If
we
can
make
them
just
work
so
to
speak
so
that
you
can
write
them
and
it's
sort
of
like
well,
it
will
be
specialized
if
it
needs
to
be
and
not
otherwise
and
stuff
like
that.
But
and
maybe.
C
E
I
definitely
want
that
to
exist.
I
don't
know
whether
I
think
it's
right
to
prioritize
it
above
it,
and
this
is
the
case
I
I
haven't
like
I
said
I
just
looked
at
this
issue
for
the
first
time,
so
I'm
not
sure
that
I
understand
fully
the
motivation
for
this
unlocking
this
specific
case.
So
these
two.
A
A
Different
in
the
sense
that
specialization
is
about
implementation
like
having
a
better
implementation
in
some
cases,
whereas
the
negative
implementation
is
about
api,
where
you
promise
as
a
create
that
we
will
never
implement
this
trait
for
this
type
right,
I.
E
I
actually,
I
think
I
don't
agree
with
that
in
this
case,
I
think
so
I
don't
know
I've
given
some
examples
before,
but
I
think
a
lot
of
my
personal
uses
of
specialization
fall
into
more
api,
like
things
where
I,
just
because
of
the
shape
of
my
types,
I'm
either
trying
to
write
like
a
recursive,
impul
or
something-
and
I
can't
prove
some
some
bit
of
negative
information,
so
I
need
either
some
kind
of
like
overlapping,
impul
or
like
well
like
we
were
talking
about
before
the
sort
of
that
things
like
the
default
for
this
for
zero
length.
E
Arrays
right
are
the
sort
of
things
where,
like
that
I
hit
when
I
and
I
want
specialization-
and
this
feels
like
along
a
kind
of
similar
line
to
those
things
where
I
have
like
a
thing
that
the
compiler
doesn't
know,
isn't
an
overlap,
and
so
I'm
happy
to
like
slap
a
default
on,
but
or
maybe
it
even
is
technically
an
overlap.
But
that's
like
not
interesting
to
me
from
an
api
perspective.
A
But
if
you
have
like
two
crates,
a
and
b,
where
b
needs
to
implement
something
which
isn't
possible
because
a
might
do
it
right,
then,
with
the
negative
they
implementation,
that's
something
you
apply
in
crate,
a
where
it
can
add
a
promise,
whereas
with
specialization,
that's
something
you
can
use
in
create
b
to
overcome
this
problem.
So
in
this
case
we
want
to
add
these
promises
to
the
standard
library.
B
I
I
think,
there's
something
worth
capturing
there.
The
example
you
just
gave
mara.
I
think
it
is
possible
in
some
cases,
that
you
would
be
doing
a
negative
impul
in
sp
crate
b,
to
say
this
thing
that
I'm
writing
that
the
compiler
might
otherwise
think
might
overlap.
B
B
I
appreciate
you,
taylor,
expelling
out
exactly
what
the
case
was,
because
I
have
run
into
that
case
where
I
thought
I
wanted
specialization,
because
really
I
needed
to
teach
the
compiler,
even
though
you
think
this
overlaps,
it
actually
doesn't-
and
I
can
imagine
this
being
frankly,
a
more
convenient
way
to
use
that
use
case.
If
what
I
really
want
is
hey,
no,
this
doesn't
actually
overlap.
I'd
rather
just
tell
the
compiler.
This
doesn't
overlap
rather
than
very
carefully
using
specialization
to
say
it's,
okay,
that
it
overlaps.
A
E
C
E
And
we
the
reason
we
can't
do
that
is
because
stir
can't
implement
error
on
it
on
its
own,
because
it
can't
allocate
correct.
So
if
we
had.
B
C
C
C
You
should
ask
jane,
I
forget,
I
think,
I'm
sure,
there's
a
reason
but
yeah.
I
will
point
out,
though,
that
that
I
mean
I
there's
definitely
a
difference
between.
I
think
we
should
move
on,
but
there's
definitely.
C
Between
specialization,
like
you,
don't
need
a
default
right
if
you
know
that
the
impul
doesn't
exist
and
that
will
that
could
be
important
in
some
cases.
It's
probably
not
it's
not
here.
I
don't
think
in
particular,
but
like
with
associated
types
or
something
else
you
might
want
to
be
able
to
not
yeah.
E
I
wanted
to
be
clear,
like
always
being
applied
or
not,
and
in
this
case
it's
helpful
to
have
that
extra
information
and
then
in
other
places
I
could
see
it
being
useful
that
you
have
default
as
an
explicit
flag
that
for
some
type
that
you
know,
implement
error.
This
isn't
actually
the
the
impul
that
will
be
applied
so.
F
F
But
the
example
I
thought
was
interesting
was
that
saying
no
stir
is
not
a
function.
F
Yeah,
I
I
think
there's
some
types
where
it
make
fundamental
makes
sense.
I'm
not
sure
that,
like
the
traits
that
are
fundamental,
that
makes
quite
so
much
sense
like
right
now.
The.
F
C
B
Two
things
regarding
fundamental:
has
there
been
ever
any
discussion
of
being
able
to
generalize
fundamental
to
something
that
could
be
used
outside
the
standard
library,
as
in
here
is
a
very.
C
B
Right,
so
that's
that
much
seems
true,
I
think
I'm
more
getting
at
is
there
theoretically
a
way
we
could
define
a
trait
or
a
property
that
sufficiently
carefully
defines
what
properties
we
need
of
a
fundamental
type
that
it
then
becomes
possible
to
declare
your
type
as
fundamental.
That
is
some
container
around
the
t,
or
is
that
just
too
concerning
to
consider
extending
outside
of
the
standard
library?
Like
suppose,
for
example,
somebody
was
building
their
own
specialized
variant
of
box
and
they
wanted
the
same
property
box
has.
C
B
Okay,
one
other
thought
on
this.
I
recently
started
working
with
a
group
of
people
and
that
opportunity
may
provide
a
way
by
which
I
can
get
somebody
working
on
the
problem
of
merging
core
standard
and
alec
and
there's
some
interest
in
doing
that.
I
wanted
to
bring
that
up
in
the
context
of,
if
that
were
the
case,
would
that
remove
the
primary
motivation
for
this,
or
is
there
still
interest
in
this?
Even
if
we
didn't
have
the
problem
with
core
alex?
Did.
C
Yeah
there's
various,
I
don't
have
a
exhaustive
list
of
use
cases,
but
you
know
being
able
to
just
affirm
that
something
will
not
be
implemented
and
then
rely
on.
That
seems
useful,
however,
whether
it's
the
most
useful
thing
like
I
you
know,
I
see
taylor's
point
that
maybe
a
pushing
on
specialization
might
be
more
valuable
use
of
time.
E
Yeah,
I'm
sorry
I'm
sorry
to
take
us
on
that
weird
train.
I
think
I
was
just
trying
to
express
the
I
have
times
and
rest
where
I
want
to
write
an
impul
that,
like
is
clearly
sane
in
my
head,
but
like
can't
convince
the
compiler
and
it
seems
like
we're
like
developing,
like
at
least
three
different
tools
for
like
getting
around
that,
and
I
would.
C
E
C
E
B
Well,
this
concept
seems
fundamentally
reasonable.
I
think
I
was
purely
asking
a
prioritization
question
of:
is
this
still
high
priority
if
we.
E
It's
a
good
thing
to
work
on,
but,
like
I
I
don't
know
what
the
heck
I
designed
for
that
would
look
like
like.
I
have
a
bunch
of
really
specific
technical
questions
about
how
that
would
work
and
that
will
seems
like
it
would
take
a
lot
longer
to
resolve
than
like
hey
negative
impulse.
We
have
them
now
and
you
can
do
coherence
with
them.
F
Yeah
given
especially
given
that
we
already
have
negative
impulse
as
a
thing
and
a
bunch
of
the
a
bunch
of
the
rules
are
already
in
fact
implemented
for
them,
because
we
added
them
when
we
did
the
negative
impulse
for
ampersand,
mute
and
things
so
like
we've
had
other
related
things
to
this,
and
I
think
part
of
it
also
sort
of
to
what
taylor
was
saying
about
this
seems
easier
to
land
than
specialization.
B
B
Seems
appropriate
in
two
cases
that
would
be
nagging
yourself.
That's
right!
So
next
up
we
have
proposed
fcps,
so
we
have
first
of
all
calling
methods
on
generic
parameters
of
const
fins.
This
is
2632.
B
I
don't
want
to
spend
an
inordinate
amount
of
time
on
these,
since
they
are
in
fcp,
but
this
one
in
particular
looks
like
it
is
quite
an
old
fcp
and
may
need
to
be
revived
at
some
point.
D
I
think
ali
or
someone
yeah.
E
B
Well,
I
would
say,
regardless
of
that,
even
if
we
do
end
up
with
deciding
to
make
forward
progress
on
this
one
if
it
has
been
a
year
since
we
last
checked
boxes,
I
think
it
is
appropriate
to
review
and
can
confirm
what
it
was
we
signed
off
on
again.
But,
yes,
we
should
still
sort
that
out
and
make
sure
that
we
are
making
forward
progress
on
it.
B
I
found
it,
I
found
it:
it's
20,
21,
0601
niko,
to
suggest
a
new
threat
or
repropose
that
so
yes,.
B
B
C
Do
we
foresee
a
lot
of
like
negative
consequences
of
permitting
one
to
match
the
complete,
triple
say.
B
I
think
that
was
always.
The
concern
is
that
if
people
can
match
the
whole
target
string
and
they're
used
to
target
strings,
that
may
be
okay
in
the
narrow
case,
where
they
really
do
only
wanna
run
on
one
or
two
targets,
but
in
the
broader
case
they
may
use
target
strings
when
they
should
have
matched
on
something
narrower
and
end
up
being
less
portable
than
intended.
B
For
example,
there
was
also
the
concern
which
came
up
last
week,
so
I'm
bringing
it
up
again
for
anybody
who
wasn't
here
last
week
that
it
seemed
like
a
lot
of
the
interest
in
this
was
heavily
about
wasm,
32,
unknown,
unknown
and
matching
it
because
it
is
incomplete
and
stubbed
out
first,
and
if
we,
if
that's
the,
only
concern
having
a
shorthand
to
make
that
easier,
doesn't
seem
like,
we
need
to
add
fully
fully
general
match
the
whole
target
triple
support
when
you
could
just
say
well,
if
I'm
target
family,
wasm
and
os
unknown.
A
I
want
to
say
something
about
the
the
portability
concern,
so
you're
saying
that
you're
concerned
about
people
being
too
specific
when
you
could
be
more
like
less
restrictive
in
their
configure
attributes.
But
I
think
the
opposite
is
also
a
concern.
If
someone
specifically
has
like
some
inline
assembly
with
with
platform-specific
functions
that
it
calls,
it
only
works
on
one
specific
target,
but
they
need
to
write
it
out
as
a
combination
of
like
target
os
and
target
whatever.
A
Then
they
might
accidentally
become
and
not
restrictive
enough
right
and
I
think
that's
a
problem
in
some
places
like
compiler
built-ins
or
some
other
things.
I've
worked
on
where
I
really
wanted
to
make
sure
that
it
only
compiles
on
one
very
specific
target,
because
for
everything
else,
I
didn't
even
try
what
it
does
and
that's
absolutely
possible,
and
I
couldn't
convince
myself
there
that
what
I
wrote
would
exactly
only
match
that
one
target
or
not,
maybe
accidentally
something
similar.
So
I
think
the
concern
also
holds
in
the
opposite
way.
B
B
B
Even
if
it's
not
the
ideal
solution
like
if
you're
used
to
target
strings,
because
you
come
from
the
see
and
configure
and
auto
conf
world,
then
you
may
use
a
target
string
because
that's
more
recognizable
than
the
names
for
the
components
and
the
more
verbosity
for
the
components,
and
then
you
end
up
matching
more
specifically
than
you.
Okay,
otherwise
could
have.
But
that's
always
that's
not
a
deal
breaker,
that's
just
that
is
the
one
of
the
potential
negatives.
The
question
is:
does
that
outweigh
the
positive
of
let's
give
people
the
option.
C
E
C
D
C
B
Actually,
I
was
the.
B
That
seems
to
work
the
one
other
thing
that
seems
worth
bringing
up
here
is.
I
don't
think
that
this
is
likely
to
be
an
issue,
for
the
common
case
of
here
is
a
target
where
two
of
the
components
actually
matter
like.
If
you
have
x86
unknown
linux
gnu,
nobody
cares
about
the
unknown
and
linux.
Gnu
is
straightforward
enough.
B
B
So
there
are
all
sorts
of
platforms
like
arm
or
risk
five
or
similar,
where
mini
details
have
been
wedged
into
them
like
arm
el
unknown
linux,
gnu
eabi,
where
the
abi
has
been
wedged
into
the
last
component.
B
B
We
have
logic
in
our
parsed
out
version
to
handle
that,
but
that's
not
something
we
can
expect
of
people
who
are
dealing
with
target
strings.
So
that's
my
concern
would
be
all
the
squirrely
little
flags
that
people
wedge
into
a
target
string
that
we
currently
know
how
to
handle
by
matching
them
inside
the
compiler.
A
B
B
C
A
Because
this
is
more
openly
only
more
advanced
feature
because
for
the
more
advanced
one
you
need
to
split
all
this
stuff
and
then
for
things
like
new
abi
like
how
do
you
split
it?
Exactly
that's
going
to
take
a
lot
of
like
shedding,
whereas
matching
on
the
full
target
is
pretty
trivial
to
just
accept
and
implement
right
now,.
C
G
B
So
we
do
have
the
split
out
version.
However,
we
don't
have
any
kind
of
shorthand
to
make
that
easier.
So
right
now,
if
you
want
to
match
two
or
three
components
of
the
split
out
version,
it
is
sufficiently
annoying
to
do
so
and
verbose
to
do
so
that
if
people
have
the
option
of
matching
the
target,
they
may
do
so
just
for
the
brevity
not
for
actually
needing
to
do
so.
B
G
C
B
B
Right
so
the
proposal
was
to
add
a
shorthand
that
amounts
to
target
with
parens
that
is
the
equivalent
of
all
and
inside
it.
You
can
write
all
of
the
target
underscore
things
without
the
target
underscore.
C
B
I
think
that's
there's
a
really
obvious
proposal
and
I
have
a
concern
registered
on
the
rfc.
That
amounts
to
please
add
that
proposal.
At
this
point,
I
personally
would
be
happy
with
going
forward
with.
Let's
allow
both
and
let's
specifically
request
that
the
rfc
also
specify
the
shorthand
version.
B
A
To
imagine
those
extra
flags
you
just
mentioned-
that's
currently
just
not
possible,
but
that's
something
for
a
later
season.
No,
I
think
it
would.
It
would
be
if.
D
C
B
Desired
so
yeah
there
is
target
abi
specified
in
an
rfc.
I
haven't
dug
through
the
tracking
issue
to
find
out
if
somebody's
implemented,
it.
A
B
But
it
seems
like
we
resolved
and
I
will
take
the
action
item
to
write
a
summary
of
the
consensus
and
propose
that
the
rfc
add
the
shorthand
syntax.
B
G
B
B
There
appears
to
continue
to
be
ongoing
progress
on
specifying
a
small
subset
of
config
accessible.
That
is
implementable.
There's
ongoing
discussion
in
this
regard,
so
I
don't
think
that
there's
a
blocker
here,
I
think
that
people
are
trying
to
make
this
work
all
right.
That
said,
I
should
point
out.
B
I
don't
know
how
much
implementation
energy
there
is
at
the
moment
and
if
anybody's
excited
about
seeing
that
subset
of
accessible
implemented,
they
might
want
to
poke
the
people
who
are
excited
about
them
and
see
if
any
of
those
folks
want
to
do
the
implementation
or
if
there's
just
a
bunch
of
people
saying
yeah.
That
sounds
good.
B
I
think
at
this
point
there
is
a
very
concrete
proposal
and
not
a
lot
of
remaining
bike
shedding
the
proposal
pretty
much
amounts
to
allow
config
accessible
on
things
where
the
path
starts,
with
one
of
the
standard,
crates,
standard,
alec
or
core
or
proc
macro
for
that
matter,
and
just
allow
that
as
a
subset
of
what
might
we
might
theoretically
allow
in
the
future,
I
don't
think
there's
any
further
bike
shedding
going
on,
so
it
really
is
just
a
matter
of
someone
implementing
this
okay.
B
C
C
We
added
track
caller.
You
can
use
it
almost
anywhere,
but
you
can't
get
it
through
dynamic
dispatch.
The
way
that
it
works
is,
it
adds
a
little
extra
information
to
the
abi,
an
extra
implicit
parameter
that
identifies
the
call
site
and
what
this
pr
does
is
say.
If
you've
put
that
attribute
in
the
trait,
we
will
add
that
extra
abi
information
into
the
v
table
so
that
you
get
it
also
through
dynamic
trade
object
access.
C
If
you
put
that
attribute
on
not
on
the
trait
but
just
on
an
impul
which
already
is
possible,
then
I
think
all
these
things
are
already
possible.
It's
just
that
what
happens
today
when
you
put
the
attribute
in
those
places,
is
that
you?
Don't
you?
Don't
you
don't
get
the
more
precise
caller
through
v
table
dispatch?
C
You
just
get
it
if
you
do
static
dispatch,
but
if
you
put
the,
if
you
don't
put
the
attribute
on
the
trait,
you
won't
get
it
through
virtual
dispatch,
even
if
the
impul
has
it,
because
we
wouldn't
have
added
it
into
the
v
table.
The
v
table
is
determined
by
the
trade
definition.
There's
really.
C
B
A
C
A
C
But
also
we've
already
that
all
stuff,
as
far
as
I
know,
all
that
stuff
is
already
sort
of
stable
and
out
there.
E
G
Wait,
I
don't
I
don't
know
if
taylor's
argument
there
quite
makes
sense,
because
I
can
imagine
an
implementation
like
the
whole
point
here
is
yeah.
You
put
it
on
the
trade.
You
have
many
implementations,
one
of
them
doesn't
include
location
collar
and
isn't
using
location
collar
and
yet
they're
paying
the
cost
even
in
static
dispatch
of
having
track
collar
on
it
right.
G
B
That
actually
does
make
sense
in
that
we
should,
at
the
very
least,
allow
the
compiler
to
be
able
to
optimize
that
away
for
static
dispatch
at
the
very
whether
it
inlines
or
whether
it
simply
like
calls
into
a
later
piece
of
the
function,
it's
theoretically
possible
for
the
compiler
to
optimize
static
dispatch
to
not
care.
If
it
knows
that
the
caller,
the
callee
doesn't
care.
C
G
B
There
is
actually
some
evidence
that
there
are
performance
hits
in
that
we
have
had
prs
to
the
compiler
to
add,
or
rather
the
standard
library
to
add,
track
caller
to
more
methods
and
when
track
caller
is
added
to
especially
hot
methods
in
the
standard
library,
then
the
perf
test
suite
does
turn
up.
Some
hits
some
regressions,
so
it
is
not
free,
it
is
worth
it
in
many
cases,
but
it's
not
free.
That's.
G
B
C
B
C
So
last
time
I
I
put
this
oh,
this
has
been
here,
but
I
put
this
note.
We
have
this
edition
mara
pointed
out
that
this
is
like
blocked.
It
doesn't
work.
Well
with
the
edition,
I
propose
we
close
this
issue
and
request
a
new
like
we're
not
ready
to
stabilize.
I
think
we
should
take
it
off
our
bucket
and
ask
someone
to
follow
up
with
a
plan
that
of
how
to
manage
the
new.
B
B
C
Sorry
so
panicking
and
constants
is
supported
now,
and
there
was
a
proposal
to
stabilize
it.
However,
the
panic
macro
with
the
new
addition
it
de-sugars
to
it
always
uses
the
formatting
machinery,
which
is
not
const
constable
or
whatever
khan's
safe,
and
so
that
means
that
you
actually
could
only
do
this
in
older
editions,
which
seems
not
great
so
or.
C
B
C
A
We
promise
to
provide
automatic
migration
for
everything
that
works
in
the
third
condition
of
the
next.
So
oh.
B
Should
we
should,
we
add
some
manner
of
panic,
stir,
function
or
add
some
kind
of
fix
to
panic,
but
I
mean
I'm
not
trying
to
go
into
detail
on
already
exists
and
works.
It's
just
hidden
and
unstable.
B
I
see.
Is
it
possible
to
teach
panic
with
no
extra
arguments
to
do
that.
B
With
a
string
but
no
format,
arguments
is
there
any
reason
we
couldn't
have
that
translate
to
panic,
with
a
stir
as
a
function.
A
That's
using
core
because
panic
any
doesn't
exist
in
course,
so
the
core
panic
macro.
Currently,
if
you
only
give
the
literal
no
arguments
it
forwards
to
panics,
to
which
only
accepts
a
string,
but
it.
A
Any
formatting
arguments,
so,
if
you
put
the
brace
in
there
and
it
just
accepts
it
and
if
you
put
an
escape
brace
in
there,
then
it
just
shows
both
braces.
I
see
all
kinds
of
other
problems
but
yeah
that
function
exists,
so
we
could
expose
it,
but
I
rather
just
make
the
macro
work
for.
C
C
A
And
mentioned
last
time
there
were
a
lot
of
suggestions
for
just
making
it
work
with
panic
any
and
nothing
else.
The
pro
we
had
that
before
sort
of
the
problem
is
that
panic
any
only
exists
instead,
another
in
core
and
lots
of
people
writing
constant
stuff
are
using
notes
and
no
stud.
So.
A
B
So
I
think
I'm
convinced
at
this
point
that
that's
a
blocker
does
that
need
to
mean
that
we
close
this
or
can
we
simply
mark,
leave
the
concern
and
say
this
is
blocked
on
this
change.
We
will
stabilize
once
that,
once
somebody
comes
up
with
a
solution
for
that.
C
B
C
B
C
B
Okay,
let's
see
if
anybody
want
to
jump
ahead
before
we
talk
about
we're.