►
From YouTube: Lang Team Triage Meeting 2020.04.02
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).
B
C
Within
this
meeting
we
have
discussed
that
we
believe
that
this
is
their
last
remaining
roadblock
to
stabilizing
or
I,
guess
changing
the
behavior
to
safe
by
default
behavior,
but
I.
Don't
think
that
we've
posted
a
summary
of
why
we
think
this
is
the
right
decision
to
make
and
so
forth,
and
certainly
to
my
knowledge,
there's
not
been
an
FCP
of
that
right.
D
A
Good
I
wrote
up,
some
will
come
to
this
point
there,
these
other
items
I,
don't
think
haven't
done
so
we
should
decide
some
design
meetings.
I,
don't
know
if
we
want
to
take
meeting
time
for
it
now
we
could,
let's,
let's
do
it
towards
the
end,
and
we
can
do
it
if
necessary,
on
on
Zula
I'll.
Come
back
to
that
we
don't.
B
A
E
E
So
I
just
came
back
from
vacation
as
of
literally
today,
and
one
of
the
first
items
on
my
rust
list
is
to
check
back
in
with
Ryan.
My
understanding
is
that
the
current
situation
is
still
that
they
are
putting
together
a
much
reduced
proposal
in
terms
of
let's
see
if
we
can
reduce
the
scope
enough
to
make
this
simple
and
beyond
that,
I
still
need
to
have
a
conversation
with
them
and
find
out
what
that
looks
like
so
that
I
don't
retread
existing
territory
by
way
of
being
missing
an
unhelpful.
A
A
A
A
All
right,
never
type
stabilization,
I
left.
The
comment
with
some
alternative
plans
to
experiment
with
I
can
briefly
I
won't
go
over
the
implementation
details,
but
something
that's
worth
pointing
out
is
that
you
know
this
is
a
lint.
There's
no
perfection
here.
We
can't
detect
in
particular
the
way
I
get
most
it,
at
least
for
example.
It
would
not
complain
about
a
case
like
this,
where
the
tea-
and
this
is
the
example
where
you
why
you
would
not
want
to
complain.
D
D
A
Not
sure
an
interesting
question
is
whether
this
is
debt
or
investment,
because
the
so
the
exactly
I'm
looking
at
is,
if
you
wind
up
with
some
live
node
whose
tight
is
a,
is
it
never,
those
type
winds
up
becoming?
Never
because
of
fall
egg?
You
get
a
warning
and
I
kind
of
think.
You
never
want
that.
Like.
D
A
A
A
A
D
David
Polly
is
proposing
that
we
accept
to
dot
some
field
in
X
dot.
Some
index
in
the
grammar
is
somehow
on.
This.
Pr
is
specifically
doing
it
through
some
extra
adjustments,
which
is
one
approach
and
another
approach,
and
so
so
what
currently
happens
is
that
we
have
int
int,
sorry,
I,
don't
stopped
float
as
the
list
of
tokens,
and
so
this
change,
in
fact,
to
produce
items,
dots,
integer,
dot,
Institute,
which
means
that
the
parts
are
just
handles
it
correctly.
D
Another
approach
would
be
to
tune
it
by
like
how
we
break
up
vo
tokens
like
ampersand
ampersand
is
broken
up
to
Amazon,
and
then
you
have
another
one
sentencing
later,
and
this
is
sort
of
what
magical.
This
does
not
really
do
instead,
of
course,
there's
the
question:
do
we
want
to
do
this,
though,
so.
F
I'm
trying
something
the
analogy
here:
I
would
think
that
the
upper
10%,
it's
a
it
significant,
there's
no
white
space,
putting
three
impressions
right.
The
whole
point
there
and
put
this
case
and
you
won't
be
able
to
add
white
space
between
the
periods
and
the
zeros
right.
I,
don't
understand
analogy:
I,
guess,
I'm,
just
saying.
E
So
I
think
the
issues
I
understand
correctly
is
that
we
have
to
handle
things
like
reference
to
reference
to
foo,
with
like
ampersand
ampersand
foo,
which
should
parse
as
two
single
unary
ampersand
operators
and
separate
from
that.
We
also
have
the
and
end
operator
for
a
module
and-
and
so
the
lexer
needs
to
how
you
know,
rather
than
teach
the
lexer
and
parser
to
fundamentally
interact.
My
understanding
from
this
is
that
we
lacks
as
to
individual
ampersands
and
then
later
on,
glue
that
together
and
say.
Oh,
that's,
actually,
the
binary
operator
no
as
I.
D
E
D
D
A
A
I
I
would
adjust
my
opinion
no
for
two
reasons.
First
of
all,
we
live
with
it
for
a
while
and
I
don't
know,
I've
certainly
encountered
cases
where
it's
moving,
it's
a
minor
annoyance,
but
also
that
we
have
sort
of
ideas
and
technology.
We
didn't
have
before,
like
these
approach,
that
we've
used
to
break
up
complex,
likes
tokens
into
simpler
ones
or
something
we
do
already.
So
if
we
can
find
an
elegant
waiting,
integrated
I
think
we
should
do
it.
Yeah.
D
A
E
A
E
Was
going
to
briefly
observe
that
the
C++
community
has
had
many
issues
regarding
the
greater
than
greater
than
being
interpreted
as
the
right
shift
operator
versus
to
ending
tokens
of
the
closed
angle
bracket
or
a
template
and
has
gone
through
similar
levels
of
turmoil
to
get
rid
of
that
a
deal
with
it?
So
I
was
simply
observing
that
it
seemed
of
similar
levels
of
importance
to
deal
with
something
like
this
early,
rather
than
letting
it
become
a
long-standing
wart
in
terms
of
language
or
Augen
allottee.
E
The
other
question
I
was
going
to
ask,
was:
is
there
or
has
anybody
done
any
grammar
analysis
to
figure
out?
Is
there
any
legitimate
parse
of
a
like
outside
of
a
macro
in
which
a
floating
point
number
follows
a
dot
in
a
legitimate
parse
of
rust
I'm?
Having
trouble
coming
up
with
an
actual
scenario
for
that.
E
Was
going
to
suggest
that,
if
that's
the
case,
we
might
theoretically
be
able
to
hint
to
the
lexer
that
if
it
happens
to
see
dot
followed
by
floating
point
number,
that
it
should
interpret
it
not
as
a
floating
point
number,
but
as
dot
integer
integer,
and
that
might
simplify
things.
If
we
can
simply
make
sure
that
floating
point
number
is
not
something
that
can
follow
dot
any
legitimate
parts.
D
A
D
A
F
A
From
time
to
time
is
people
doing
it
precisely
to
work
around
the
logic
feature
that
we're
now
into
many?
Yes,
I've
seen
that
a
million
times
these
that's
happened
with
macros
like
we
were
using
a
macro
to
emulate
the
syntax
that
you're
now
added
and
now
the
macro
is
broken
because,
of
course,
correct.
D
D
A
A
A
So
I
just
wanted
to
double
check
on
this
foreign
exceptions.
I
think
the
consensus
from
our
last
meeting
was
that
we
were
going
to
leave
that
we
wanted
to
do
would
prefer
to
abort
for
now,
but
leave
the
documentation
is
saying
the
behavior
is
unspecified
and
may
change
in
the
future,
its
undefined
behavior,
and
we
expect
to
change
it
in
the
future
and
point
to
the
after
firing
line
group.
For
example,
yeah.
E
A
B
B
It
just
no
one's
passed,
it.
A
A
B
E
A
Yeah
I
guess
that,
in
contrast
to
the
cases
where
we
saw
before,
where
we
saw
people
rely
on
Linux,
there
is
sort
of
no
correct
version
of
this
code.
And
then
you
like
the
catch
unwind.
Is
there
probably
to
do
something?
And
it's
not
that
whoever
wants
to
not
actively
I
guess
the
correct
version
idea
that
the
intermediate
library
propagates
or
something
that.
B
C
I
think
what
an
argument
in
favor
the
crater
run
is.
This
is
the
sort
of
thing
that
we're
sort
of
very
likely
to
miss
on
the
beta
run,
because
it's
likely
to
lead
to
something
like
an
abort
in
some
libraries
tests
and
generally
those
are
ignored,
or
we
look
at
it
and
we
say:
oh
you
know
this
library
is
doing
something
weird
it's
hard
to
track
down
with
her.
It's
like
actually
a
regression
given
that
we
usually
have
you
know
twenty
or
thirty
libraries
that
do
this
sort
of
sometimes
aborting
on.
B
Krejci
run
is
done.
We
have
three
route
prescience,
none
of
them,
so
they're
all
doing
heap
allocation
of
types
of
large
alignments,
so
well
I
mean
we're
back
to
two
to
two
options:
either
we
accept
the
breakage
for
these
three
crates
or
we
only
restrict
only
apply
the
alignment
restriction
on
statics
and
promote
it.
Constants.
B
A
B
B
B
A
A
B
D
A
B
D
B
F
A
D
E
E
I
would
actually
observe
alignment
by
itself
is
already
going
to
be
a
somewhat
unusual
feature
and
alignment,
having
like
a
specific
alignment
requirement,
that
is
large,
is
going
to
be
rarer.
Still.
The
example
that
was
given
at
the
bottom
of
this
list
saying
somebody
needed
like
a
32k
alignment,
for
example,
is
one
particular
embedded
project
with
specific
needs,
and,
leaving
aside
that
I
would
generally
expect
specific
target
embedded
projects
to
be
underrepresented
in
crater.
E
I
would
also
suggest
this
is
a
case
of
you
know
what
fraction
of
the
people
who
are
actually
using
the
feature
just
got
broken.
I,
don't
think
three
is
like
acceptable
collateral
damage
here.
I
think
three
is
a
noticeable
warning
sign
that
we
are
have
done
something
we
need
to
pay
more
attention
to.
A
B
D
A
A
D
A
A
D
C
Do
tend
to
agree,
though,
that
like,
if
feels
like
this,
is
a
sort
of
a
thing
that
it's
fine
to
put
in
post
moralization,
especially
like
long
term
I,
would
imagine
that
it
might
even
be
the
case
that,
like
you,
called
a
linker
and
the
linker
errors
rights,
as
you
told
me
to
do
whatever
you
asked
and
I'm
going
to
refuse
now.
Linkers
today
might
not
do
that,
but
one
could
imagine
sort
of
a
nice
linker
saying
you
know
for
this
specific
target
on
this
device.
C
A
A
B
A
A
A
B
A
B
A
D
E
D
A
A
A
B
E
The
original
pitch
of
the
portability
line'
to
RFC
was
to
detect
code
that
was
compiling
on
your
platform
would
not
compile
on
other
platforms.
So,
for
instance,
you
build
on
linux,
but
you
have
no
CFG
is
that
are
saying
you
are
limited
to
linux
and
the
code
you've
written
will
not
compile
on
Windows
because
it's
importing
something
that
doesn't
exist
on
Windows.
So
in
the
context
of
these
type
conversions,
the
premise
is
well,
if
you're
writing
a
you
size
to.
E
If
you're
writing
a
you
64
to
you
size
conversion,
then
that
will
work
fine
on
a
64-bit
platform,
but
unless
you've
done
something
to
prevent
yourself
from
running
on
32-bit
platforms,
you
should
get
a
warning
to
help.
You
understand
the
flip
side
of
that
and
the
part
that
I
think
is
worth
a
little
special
consideration
here
is
I.
Think
the
domain
of
like
I,
only
run
on
64-bit
and
don't
care
about.
32-Bit
is
the
domain
of
the
portability
lens.
There
is
a
specific
case
that
comes
up
occasionally
that
I
think
is
worth
us
considering
carefully.
E
We
currently
don't.
We
don't
include
conversions,
safe
conversions
that
can't
fail
for
you,
sighs
that
make
the
assumption
that
you
size
is
at
least
32
bits.
We
allow
for
the
possibility
that
you
might
be
about
a
platform
where
you
size
is
16
bit
and
that
one
is
sufficiently
rare
compared
how
often
it
comes
up
that
it's
worth
considering
is
that
the
domain
of
affordability
lint
were?
Is
there
some
default
that
we
should
have
that
says?
E
Okay,
unless
you're
specifically
opting
into
extremely
tiny
address
spaces,
you
don't
have
to
deal
with
you
size
being
smaller
than
32
I'm,
not
cret
I'm,
not
trying
to
prejudge
an
answer.
There
I'm
trying
to
state
the
issue,
and
that
does
come
up
often
enough
and
when
people
run
into
it,
they're
often
mystified
by
the
fact
that
rust
theoretically
runs
on
16-bit
platforms.
A
E
A
E
It
needs
some
clarification,
but
I
think
there
is
value
in
evaluating
the
you
size
concerns
specifically
as
somewhat
separate
from
the
portability
Lentz.
In
that
we
should
ask,
should
you
have
to
opt
into
or
opt
out
of
you
size,
potentially
being
16-bit
I?
Think
it's
reasonable
to
expect
people
to
have
to
opt
out
of
you,
sighs
being
32-bit
if
they
want
to
do,
casts
that
assume
it's
64-bit,
but
I.
E
A
I'll
have
to
review
the
RFC.
I
was
hoping
to
see
it
really
really.
What
I
I
do
remember
this
being
an
explicit
concern
that
was
discussed
at
the
time
and
wanting
to
have
a
sort
of
notion
of
mainstream
like
mainstream
portability
versus
exotic
portability,
and
because
the
bar
to
maintain
the
wider
is
significantly
higher
and
right
and
significantly
less
important
to
most
people.
D
E
I
have
multiple
times
found
that
this
isn't
the
case.
One
notable
issue
is
that
only
you
size
can
be
used
as
an
array
index.
So
if
you
try
to
index
an
array
with
something
else,
you'll
find
yourself
doing
a
lot
of,
as
casts
and
I
would
love
to
use
safe
into
casts.
That
will
give
me
a
compile
error
if
I
do
something
that
might
break.
But
when
it
comes
to
you
sighs
a
lot
of
the
in
twos
that
you
would
expect
to
work,
don't.
E
A
D
A
D
E
Exactly
I
actually
genuinely
want
the
compiler
to
give
me
a
warning
if
I
assume
that
you
size
of
64-bit,
unless
I
explicitly
opt
into
saying
hey
I,
actually
know
for
this
project,
I
will
never
run
it
on
a
32-bit
address
space
but
I.
Don't
especially
want
the
compiler
to
say,
hey,
wait,
pointers
might
be
16
bits.
E
A
Some
reason
I'm
not
able
to
efficiently
into
Dropbox
paper,
so
I
can
take
any
notes
on
this,
but
we're
at
read
time
here:
I
would
like
I,
don't
know
I,
don't
I'm.
Somebody
want
to
leave
a
comment
saying
something.
I
still
feel
like
I
would
like
to
just
see
what
exactly
is
being
proposed.
That's
my
general
T
before
I
say
yes
or
no.
If
I
like.