►
From YouTube: 2020-10-19 Triage 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
meeting
people
can
put
their
names
on
adding
you
scott,
so.
A
Let's
see
where
are
we
let's
start
with
pending
proposals
I'll
go
through
these,
so
this
restrict
promotion
to
infallow
operations.
We
had
a
final
comment
period
it
has
completed.
I
don't
think.
A
Much
to
say
here
we
could
just
kind
of
close
it
as
approved
right
we're
basically
waiting
on
the
pr.
The
only
thing
is
that
I
was
encouraging
ralph
to
write
an
rfc.
I
don't
know
whether
we
want
to
track
that
somehow.
D
Having
a
tracking
issue
seems
like
a
good
idea,
but
otherwise
this
seems
fine.
E
I
think
at
minimum
it
would
be
good
to
copy
paste.
Essentially
the
motivation,
like
section
from
the
note
or
from
the
original
description
to
something
like
a
design,
note
or
similar
text.
If
we
don't
have
an
rfc
soon
just
to
have
it
tracked,.
C
A
A
Okay,
pub
macro,
so
we
had
a
move
to
close,
doesn't
sound
like
anything
box
objects.
Is
there
anything
else
anyone
wants
to
say
or
shall
I
close
the
issue.
D
A
A
Kind
of
collect
it
in
sort
of
a
wish
list
area
which
we
don't
yet
really
have
a
good
solution
for
that
might
be
something
worth
trying.
This
might
be
a
good
candidate.
A
B
B
Let's
do
that
right
now.
C
Concrete
this
head
to
some
discussion
on
zulip.
My
inclination
here
is
that
this
is
a
close
as
a
nice
conceptual
simplification,
but
not
helping
us
really
make
progress
on
periodic
generics.
C
So
this
is
saying
what,
if
we
make
types
and
lifetimes
have
const
only
types
so
that
they
can
be
passed
around
as
kant's
generics,
and
then
you
can
make
like
compile
time
only
lists
of
types
and
things
like
that.
C
D
A
C
Yeah,
it
seems
really
cool
to
have
the
you
know,
everything's,
just
a
constant
value.
It
says
lifetimes
or
even
constant
values.
You
can
make
a
like
list
of
lifetimes
and
things
it
sounds
cool.
I
don't
know
how
useful
it
is
and
the
bigger
problem
I
have
is.
I
don't
know
how
to
actually
get
those
types
out
of
the
const
values
and
use
them
to
do
things
based
on
its
proposal.
C
D
Right
either
you
need
a
giant
pile
of
new
syntax
or
you
need
to
effectively
punt
to
a
proc
macro
like
codec,
compile
time
system.
C
A
D
The
the
idea
of
reifying,
more
of
rust
syntax
into
the
type
system
is
an
interesting
one
or
at
least
into
the
compile-time
evaluation
system,
but
I
think
that
the
enum
is
a
good
example
scott,
because
if
we
were
going
to
go
that
route,
we
would
want
to
get
enough
of
the
ast
that
you
could
effectively
compute
at
compile
time
and
enum
definition
and
say:
hey
compiler,
please
process
this,
whether
that's
prop
macros,
2.0
or
just
an
extension
of
existing
proc
macros
is
another
question,
but
either
way
yeah
that
it
doesn't
look
like
this.
C
D
I
did
so
I
wrote
these
because
I
thought
it'd
be
good
to
take
a
little
bit
of
this
meeting
time
to
do.
A
combination
of
design,
meeting
planning
and
other
item
scheduling.
D
One
of
these
items
like
getting
consensus
around
language
team
membership,
language,
team
expectations
of
membership
etc
is
something
where
we
could
potentially
discuss
that
async
via
the
t,
lang,
zulip
streams,
but
playing
team
process
tweaks,
the
ones
that
we
were
just
talking
about
nico
seemed
like
something
that
would
be
easiest
to
start
out
talking
about
live
the
whole
seconding
thing.
D
A
Let
me
just
post
this
in
here
real
quickly,
while
we're
doing
that
we
talked
about
this
playing
team
expectations
once
before
and
there's
this
document
that
I
drew
up
and
which
has
been
edited
slightly
since
then.
That
might
want
people
to
like
look
at
and
decide
if
we
think
this
is
right.
I
think
the
most.
A
We
don't
I'm
not
necessarily
want
to
do
it
right
now,
but
just
pointing
that
out
to
people
it's
here.
I
think
the
thing
I
would
call
attention
to
is
attending
both
how
we
feel
about
that
and
the
expectation
that
ideally
people
will
really
wanting
for
a
project,
even
if
you
don't
have
to
be
doing
that
all
the
time.
A
That
would
be
something
to
think
about
process
tweaks,
so
part
of
the
context
here
is
that
the
libs
team
has
published.
B
A
The
so
the
libs
team
has
adopted
this
or
as
proposed
this
charter.
It
has
a
lot
of
overlap
with
compiler
and
lane
and
it
would
be
cool
to
kind
of
unify
them.
That's
the
first
thing
so
that
we're
a
little
more
consistent
and,
secondly,
josh-
and
I
were
talking
about
like
some
specific
changes
like
using
seconding
to
close
rfcs
instead
of
having
everybody
have
to
check
their
box,
and
I
don't
know
what
else
we
were
talking
about
josh.
I
guess
using
seconding
in.
D
A
Anyway,
potential,
I'm
not
sure
if
I
agree
yet,
I'm
not
gonna
take
a
position,
but
we
could
talk
about
it.
Maybe
we
could
get
order
rouse
and
maybe
some
listing
folks
involved.
Even
I
think
it
might
be
good
to
do
a
smaller
discussion
about
this
too,
to
come
up
with
a
specific
proposal.
First.
A
Actually,
while
we're
doing
this,
I
would
add
scheduling
of
design
meetings,
how
to
do
it
better,
because
I
don't
like
the
way
we
do
it
right
now
and
then
there's
these
two
other
things.
This.
A
I
agree,
I
think,
probably
the
process
to
compile
the
right
one,
which
is
to
say
every
fourth
slot,
to
talk
about
the
next
three,
it's
kind
of
heavy,
but
it
works.
F
So
the
compiler
team,
though
we
have
a
few,
a
smaller
workload
of
meetings
to
go
through
now.
A
A
A
Maybe
even
just
the
first
one
and
then
the
main
goal
for
the
other
ones
that
we
actually
go.
A
Try
to
keep
meanings
effective.
If
so,
I
would
want
to
take
on
the
the
goal
of
doing
the
playing
team
process
tweaks
and
no
the
invited
discussion.
I'm
not
sure
exactly
what
document
is
needed
there,
because
there's
a
paper
everyone
should
read.
Maybe
that
is
the
document.
C
A
E
We
have
bonanza
left
too
right.
So,
like
that's
another
fourth
item
here
or
fifth
item.
A
F
Also,
I
don't
know
what
the
I'm
sorry,
if
you
already
mentioned
this
bullet,
I
didn't
catch.
It
then,
by
the
discussion
with
authors
of
russ
paper.
A
That
was
published
in,
I
think
popcorn
called
the
pld
I
forget
which
went
through
and
looked
at
a
bunch
of
like
breast
code
and
found
bugs
basically-
and
you
know,
categorized-
I
guess
pldi
categorized
what
categorize,
what
kinds
of
bugs
are
and
their
causes,
and
we
thought
it'd-
be
pretty
cool
to
bring
some
of
them,
bring
the
authors
in
and
we
can
kind
of
talk
about
it
because
they
also
make
some
proposals
ways
to
address
those
bugs
and
make
them
less.
Something
might
be,
might
receive
some
ideas,
gotcha,
okay,.
A
A
A
A
I
don't
think
so,
I'm
just
I
don't
know
if
I'm
gonna
do
I
see
like
things?
I
think
there
are
fresh
comments.
I
don't
know.
A
Genetics,
lc
nr,
left
a
comment
that
I
was
following
up
on,
where
it
seemed
like
the
main
constant
area
was
more
or
less
today,
it's
really
interesting
and
that
they
wanted
some
help
on
designing
the
suite.
I
was
gonna,
try
to
make
some
time
for
that
and
work
with
him,
and
that
might
also
then
be
a
nice
like
update
for.
A
A
Project,
safe
transmute,
one
thread
is
active,
you
haven't
read
it.
There
were
some
major
updates
and
I
guess
there
is
currently
a
certain
amount
of
controversy
about
or
like
controversies
may
be
too
strong
a
word,
but
there's
some
back
and
forth
about
how
minimal
this
should
be
and
what
exactly
the
use
cases
are
for
engines.
C
The
the
whole
here
thing
with
making
it
a
macro
and
the
implications
that
has
on
all
the
different
traits
is
quite
interesting.
A
Yeah
I
had
a
long
conversation
with
json
today
about
it
made
me
feel
like
I
understand
better
what
motivated
it
we'll
still
feel.
I
have
remixed
feelings
about
it
as
a
mechanism.
I
don't
know
how
much
we
want
to
talk
about
it
right
now,
though.
Okay.
C
D
C
C
A
A
D
I
was
just
going
to
suggest
kind
of
following
up
with
what
you
mentioned.
I
do
think
that
the
problem
that
here
is
trying
to
solve
makes
sense
in
terms
of
let's
let
you
do
things
on
the
basis
of.
I
would
have
permission
to
construct
this
by
hand.
If
I
wanted
to,
I
think
it's
more
a
question
of
is
this:
should
this
be
api
surface
area?
Should
this
be
exposed,
should
it
ever
be
stabilized
or
is
it
something
that
should
be
perma
unstable
as
part
of
safe
transmute,
internals.
D
A
little
bit
right,
I
mean
I
have
personally
no
objections
to
here
as
a
mechanism
for
like
internal
surface
area.
If
we
build
it
in
such
a
way
that
if
we
changed
it,
we
wouldn't
break
anything.
But
if
it's
something
that
we're
actually
exposing,
I'm
not
confident.
It's
the
right
point
in
the
design
space
versus
just
hiding
more
of
it
behind
the
compiler
curtain
and
saying
we're
just
here's
the
semantic
we're
trying
to
provide
of
constructability.
A
I
want
to
find
some
solution,
that's
more
like
you,
opt-in
and
we're
satisfied
with
that,
but
I'm
not
sure
what
are
the
we,
I
think,
part
of
it
and
in
talking
yes,
ryan,
I
got
the
sense
that
for
use
cases
beyond
what
I
originally
had
in
mind,
which
was
like
ticks
and
structs
and
convert
them
in
easy
cases
like
being
able
to.
B
Initialized,
I
don't
know
well
all
right.
A
A
Yep
no
major
comments
here:
implementation
proceeds
on
all
of
these
topics:
I've,
never
typed!
You
made
some
progress.
D
One
update
on
inline
assembly,
there
was
some
renewed
discussion
recently
on
whether
this
should
be
exposed
as
a
unqualified
macro
asm
or
whether
it
should
be
exposed
as
something
like
core
arch
target
name.
Asm.
D
There
was
discussion
about
like
which
one
makes
the
most
sense.
There
are
trade-offs
there.
Some
of
these
were
discussed
in
the
original
rfc,
but
not
nonetheless,
I
wanted
to
bring
up
that
it
was
being
discussed
again.
I
think
part
of
the
concern
there
is,
on
the
one
hand,
most
of
the
time
your
assembly
will
be
target
specific,
and
so
it
wouldn't
be
that
big
a
deal
except
that
it'd
be
an
extra
bit
of
friction
to
have
to
import
the
requisite
asm.
D
On
the
other
hand,
there
are
real
cases
where
you're
using
azim
in
a
surprisingly
portable
way,
whether
it's
I'm
writing
asm
for
both
32-bit
and
64-bit
and
don't
want
to
have
to
import
one
or
the
other,
because
that
might
be
my
only
32-bit
64-bit
qualifier
or
alternatively,
you're
primarily
using
assembly
directives.
D
D
D
Apart
from
that,
I
think
this
is
still
on
track
for
stabilization.
I
haven't
seen
any
other
issues
that
would
came
just
fine.
There
has
been
there's
still
some
ongoing
discussion
of
how
do
we
handle
the
llvm
version
checking
and
similar,
but
there
has
been
a
new
llvm
release
so
that
may.
A
A
A
A
Oh
right,
I
nominated
it
because
you
guys
were
gonna.
Have
some
you
ready
to
talk
about
it
or
not.
Yet.
C
Oh,
we
basically
I
looked
through
this
and
when
yep
yep
yep
it's
it's
doing
all
the
sort
of
things
we're
talking.
It
has
a
good
nice
long
discussion
about
the
fragile
base
class
problem
about
hey.
This
is
all
about.
If
people
update
the
class
and
the
trade
and
you're
being
more
specific
than
that,
the
name
conflicts
break
your
code
and
that's
not
really
what
you
wanted
seems
to
do
a
good
talk
through
of
all
of
the
sides
of
like
okay.
It
has
to
this
version.
C
It
here's
why
these
cases
still
need
to
error.
It's
very
it
has
it's
very,
very
similar
to
all
the
things
you'd
probably
expect
if
you
used
to
like
java
or
c-sharp
interface
rules
or
what's
considered
a
conflict.
C
The
most
interesting
thing
that
I'd
be
curious
to
get
some
people
who
know
more
about
the
implementation
side
here.
Is
that
if
you,
if
you
bound
your
generic
to
super
or
sub
or
super
plus
sub,
they
all
do
slightly
different
things.
C
Yes,
so
the
last
of
those
was
super
plus
sub
means
that
if
you
call
a
method
that
they
both
have
that's
an
ambiguity
error,
because
with
the
two
bounds,
even
though
one
is
a
super
of
the
other,
you
don't
know
which-
which
one
that
method
came
from
the
first
time.
So
you
can't
pick
it
has
to
be
an
error.
A
C
But
other
than
that,
it's
basically
doing
all
the
things
that
we
talked
about
in
bonanza
and
hypothesized
that
it
was
probably
doing
like
well
if
you're
in
super
it'll,
if
you're
in
sub
it'll
take
things
from
super
as
sort
of
a
fallback,
it'll
treat
the
things
in
its
own
trait
kind
of
like
three
parent
methods
and
then
they'll
fall
back
to
the
super
traits
if
necessary.
A
A
C
Yeah
my
my
core
feeling
on
the
whole
thing
is
that
what
it
said
it
was
going
to
do
and
what
it
said
that
was
still
had
to
be
ambiguous
seemed
well
motivated
to
me
that
there
wasn't
a
whole
lot
of
places.
I
said,
oh
shouldn't
it
be
this
other
way.
Instead,
basically,
whenever
I
was
thinking
that
they
came
along
later
and
said,
no,
it
has
to
be
that
way,
because
blah.
B
C
C
A
Okay
sounds
good,
we'll
revisit
it
next
weekend.
If
you
want
to
leave
a
comment
that
probably
would
be
appreciated
by
the
author,
but.
A
A
E
I
should
be
able
to
find
time
this
week
to
construct
the
clear
timeline
of
different
behaviors,
but
nothing
beyond
that.
B
F
A
Not
really
about
kaplan,
it's
about
the
fact
that
this
code
went
from
being
accepted.
I
think
it's
becoming
an
error
and
there
was
code
creator
like
there
are
existing
crates
that
relied
on
that
for
better
or
worse,
and
so
the
reason
the
dimension
of
cap
lens
is
just.
A
B
B
B
E
F
A
That
sounds
right.
I
guess
my
take.
I
don't
know
how.
I
think
I
would
be
happy
with
the
forbid
winning
and
you
get
a
warning.
That
would
be
a
good
outcome
for
me,
but
I
don't
know
if
I
can
live
with
the
current
behavior
also
probably,
but
it
is
a
regression
which
that's
another
design,
meeting
topic,
maybe
or
something
to
just
think
about
revisiting
our
regression
rules
and
how
much
we're
willing
to
tolerate.
E
It
seems
like
the
behavior
you
suggested,
where
we
sort
of
make
this
more
of
a
lint
makes
sense.
I
would
be
interested
in
finding
out
once
I
have
time
to
construct
the
timeline,
whether
that's
like
a
regression
or
how
much
of
a
regression.
That
is,
because
I'm
not
actually
sure
what
the
behavior
before
was
in
terms
of
whether
the
forbid
or
the
warner.
What
one
okay.
F
I
think
I
think
the
last
thing
that
was
pushed
was
what
one
that
was.
My
memory
is
that
that
was
the
bug,
as
in
the
most
recent
attribute
won
out,
which
was
itself
then
a
problem.
E
We
have
more
than
one
release
cycle
where
beta
is
148
right
now,
and
this
will
land
in
149..
So
we
have
time.
A
A
A
A
I
don't
know
what
I
would
consider
if
we
had
a
time
machine,
but
we
don't
all
right
switch
mutable.
Oh
yes,
this!
So
this
is
a
move.
If
you
remember,
I
left
a
comment.
This
was
this
was
deferred
a
couple
weeks
back
to
have
someone
leave
some
comments,
so
I
did
in
two
phase
borrowers.
We
had
this
pattern
when
we
first
added
it
where
a
pre-existing
borrow
like
let
s
equals
ampersand
v
was
allowed
to
overlap.
A
A
mutable
borrow
like
the
one
here
so
long
as
the
needle
bar
hadn't
been
activated
yet,
and
we
removed
it
because
that
introduces
a
kind
of
non-stack-like
behavior
and
that
didn't
fit
well
with
stack
borrows,
and
it
wasn't
really
the
original
goal
of
that
change,
which
was
to
accept
stack-like
case
where
you
have
a
mutable
borrow
that
has
not
yet
been
activated
a
shared
barrel
that
occurs
and
then
the
activation
of
the
needle
borrow
meaning
calling
the
function.
A
But
that
was
kind
of
controversial,
and
we,
if
I
recall
josh
you
and
I
settled
that
we
would
give
warnings
and
come
back
when
we
had
a.
We
could
take
more
time
to
really
explain
that.
A
What
we're
giving
up
by
making
this
decision
or
like
what
optimizations
were
enabling
that
kind
of
thing
and
this
this
vr
proposes
making
it
deny
by
default,
which
is
not
you
know,
finalizing
the
decision
ralph
mentions.
I
haven't
had
time
to
really
look
into
this.
That
stack
borrows
has
since
adopted
a
more
relaxed
one
that
actually
does
more
relaxed
model
that
actually
accepts
the
original
code,
but
that
will
come
at
the
cost
of
some
optimizations,
we're
not
sure
how
much
and
I'm
not
sure
exactly
why
this
was
done.
D
So,
given
that,
should
we
press
pause
on
this
one
until
ralph
gets
back
to
us
with
a
concrete
notion
of
what
needs
to
happen.
A
D
I
don't
mean
the
whole,
I
don't
mean
the
whole
stack
borrows
model,
I
just
mean
it
sounds
like
ralph
needs
a
little
bit
of
time
to
think
about
this
particular
lent,
and
we
should
follow
up
to
find
out.
Do
you
think
this
makes
sense
or
has
stacked
boroughs
changed
in
such
a
way
that
we
no
longer
need
to
block
this.
A
D
Can
we,
why
don't
I
post
a
comment
asking
if
that
means
that
a
like
concrete
decision
on
this
lint
would
need
to
wait
for
those
full
six
months
or
do
you
just
feel
that
you'd
need
six
months
to
finalize
the
whole
model.
A
That's
fine,
I
think
my
opinion
is
we.
One
thing
is
we
could
still
move
forward
with
like
we,
it
still
may
make
sense
to
be
conservative.
We
can
always
re-accept
the
code.
What
I
was
going
to
suggest
is.
We
also
could
kick
off
the
crater
run
to
get
some
idea
of
like
what
the
impact
of
this
change
would
be.
Anyway,
people
are
getting
warnings
now
they
would
be
moving
to
getting
errors
by
default,
but
we
don't
know
how
many
crates
would
be
affected.
F
Nico,
do
you
know
offhand
how
tightly
this
this
lint
is
with
the
rest
of
the
nll,
as
it's
called
how
tightly
integrated
it
is
like
the
person
who
posted
this.
I
got
the
impression
when
I
was
reading
over
the
whole
history
of
what
happened
here.
That
or
in
terms
of
this
particular
pr
is
that
you
have
this
contra,
this
contributor
who's,
looking
to
stabilize
full
nll
and,
as
part
of
that
work,
they
figured.
They
would
turn
this
to
a
deny
by
default
lens.
A
I
I'm
somewhat
motivated
because
I
feel
like
if
we're
gonna,
if
we're
gonna,
try
to
defend
against
this
pattern,
we
probably
do
want
to
deny
by
default,
which
is
why
I
would
be
somewhat
motivated
to
do
this.
F
A
F
A
Just
changing
into
a
hard
error
would
would
break
people
I'd
like
to
do
a
crater
run
still,
but
all
right,
let's,
let's
try
to
summarize
the
game
greater
run
to
see
impact
might
be
useful.
The
mountain
by
default
would
not
immediately
create.
So
if
we
want
to
disperse
folks
from
this
pattern,
the
logical
step.
D
That's
roughly
what
I'm
writing?
Yes,
I'm
trying
to
write
it
in
a
manner
that
does
not
put
pressure
on
ralph
to
change
his
timeline
expectations.
I
just
want
to
find
out
what
the
timeline
expectations
are.
A
All
right
I'll,
just
I'll
just
put
ask
just
to
leave
a
comment.
Clarifying
more
with
more
questions,
sounds
good.
A
Where
you're
like
using
your
you're,
exporting
a
name
that
you
then
resolve
against.
B
B
A
E
We
should
yeah,
I'm
not
sure
if
this
is
I'm
not
sure
what
which
nightly
this
was
potentially
either
in
a
few
hours
or
in
a
cycle.
So,
like
four
weeks.
A
Okay,
josh
asked
in
chat:
do
we
know
what
changed
to
break
this?
The
answer
josh
is
that
petrochenkov
landed
apr,
which
was
as
they
wrote,
not
strictly
a
bug
fix,
but
a
concert
more
conservative
language
rule.
I
don't
know
around
avoiding
yeah
what
the
pr
title
here
says:
self-confirming,
import
declarations.
E
A
A
A
What
is
this?
Oh,
no,
that's
the
same.
One
wrapper
transparent
on
generic
type
skips
exactly
one
non-zero
size.
A
C
It's
a
little
weird
that
if
you
do
it
with
a
generic,
you
could
put
that
st
in
a
transparent.
But
if
you
do
it
without
a
generic
you're
not
allowed
to.
C
D
It
does
seem
vacuously,
true,
that
passing
this
wrapper
around
a
zst
is
exactly
the
same
as
passing
the
zst,
which
is
to
say
you
can't
really
do
either
one
so
sure.
That's
fine.
D
The
other
alternative
would
be
to
complain
in
the
generic
case
as
well,
but
that
would
be
a
monomorphization
time
restriction.
D
A
I'd
be
happy
with
this,
but
I
feel
like
we
have
to
sort
of
define
what
the
behavior
is
when
you
do
pass
it
like.
I
guess
the
reason
for
the
rule
was
that
we
said
it
was
by
the
abi
the
same
as
that
single
field.
So
if
there
is
no
field
that
is
non-zero
size,
then
you
have
to
say
it's
treated
like
something
right.
B
A
D
Yeah
we
shouldn't
change
the
compiler
to
make
it
exactly
one
field
here,
I'm
sorry,
it's
not
exactly
one
field
and
that
field
may
be
zero
sized.
It's
there
should
be
zero
or
one
non-zero
sized
fields
and
allowing
the
zero
case
is
not
incredibly
unreasonable.
D
I
mean
along
the
same
lines.
You
could
also
have
a
repre,
transparent
type.
By
these
same
rules,
you
could
have
a
repre,
transparent
type
that
contains
both
an
actual
type
and
a
phantom
data,
for
example,.