►
From YouTube: Lang Team Meeting 2019.05.30
Description
Focus of the meeting: triage
A
D
D
B
A
B
B
B
So
there
was
a
there's,
a
bug
before,
where
a
mute
self,
if
it
appeared
with
other
lifetimes
in
the
signature,
behaved
badly
with
elision,
because
you
you
want
to
have
something:
that's
like
an
N,
mute
self
method
and
then
it
has
an
N
Muse,
something
you
know
parameter
and
then
there's
an
N
mu
output
right.
So
today
on
stable
that
works
and
that's
bad
because
it's
ambiguous
which
lifetime
it's
referring
to,
and
that
was
done
because
we
skipped
the
self-type
when
calculating
elisions
in
some
like
legacy
code
that
didn't
know
about
arbitrary
self-types.
A
B
B
B
D
B
D
A
B
A
A
B
B
So
if
you
have
it
in
a
lighted
lifetime
that
appears
in
the
output,
it
says:
hey,
there's
no
life
of
corresponding
lifetime,
an
input,
and
that
also
means
that,
like
if
you
have
a
lifetime,
an
input
and
and
another
argument
with
a
lifetime
in
the
input,
it
only
knows
about
the
second
or
lifetime
argument
in
the
input.
So
this
isn't
just
a
thing
for
for
pinpin
just
happens
to
be
the
only
place
where
anybody
actually
like
really
uses
this
today
right.
Oh
so,.
A
B
E
B
D
D
B
The
right
so
I
think
the
thing
that
was
going
on
here
right
if
I'm
understanding
this
the
the
why
the
initial
fix
was
broken.
It's
because
self,
like
it
looked
at
not
just
syntactically,
alighted
lifetimes,
but
any
alighted
lifetime,
which
included
lifetimes,
that
appeared
in
the
literal,
like
self
type
like
when
you
spells
electron
little
SEL,
F
and
so
I
think
it
just
needs
to
be
like
I
think
this
is
just
a
like.
D
D
B
B
A
D
A
A
B
D
B
E
D
A
I'm
just
doing
some
testing
as
an
aside,
our
current
behavior
matches,
Taylor's
intuition
I
think
it
also
matches
mine
I,
don't
know
the
answer
to
whether
we
have
a
good
I
would
like
an
outcome
of
this
discussion
to
be
a
good
rule.
Okay,
I'd,
like
us
to
write
it
in
the
reference.
I
think
I
agree
with
Taylor
that
it
should
be
syntactic
partly
for
implementation
reasons,
but
also
just
because
it's.
B
D
B
D
B
B
D
B
B
B
E
B
B
A
A
D
A
Mean
you
know,
I
think
there
should
be
in
it
personally.
I
think
this
should
have
been
an
error
and
the
rule
should
have
been
as
Taylor
is
saying
and
I.
Think
Central
Europe
is
also
saying
that
you
can
only
match
in
a
lighted
lifetime
in
the
output
with
an
allied
electrum
from
the
input
you
can't
give
a
name
in
one
place
and
not
the
other.
We
that
is
not
currently
the
rule
I
there.
My
self
types
are
weird
like
we
could.
We've
been
sort
of
steering
towards
that
direction
and
I
was
even
talk
of
deprecating.
A
A
E
E
B
E
A
E
A
B
D
D
A
A
D
D
Yeah,
it's
comes
unsafe,
unstable,
not
I'll,
say
good
seems
like
pretty
clear,
then
yep
so
I'm
taking
this
so.
A
D
D
D
D
But
mod
underscore
would
keep
that
in
for
for
driving,
macros
I
think
so
it's
probably
ones
between
which
one
do
you
think
is
better,
and
then
there
is
so
monos
what
probably
doesn't
have
an
engine
on
organizations
which
creative
thing
started.
So
it's
sort
of
finished
after
that
and
whereas
cons
underscore
could
be
extended
to
patterns,
others
then
there's
questions
of
what
do
we
do
in
with
those
and
when
they
are
associated
like
items
and
do
we
want
to
extend
them
to
hello
at
any
arbitrary
refutable
patterns
and
and
what
about
static.
A
A
D
B
D
D
D
A
D
D
It's
not
just
associated
items,
it's
feels
like
cattle
words,
but
on
the
other
hand
it
might
be
better
for,
like
it
works,
it
works
better
for
the
constant
sections.
So
that's
maybe
like
I'm
unsure
doesn't
mean
if
anyone
has
a
good
rationale.
I
can
use.
That
would
be
really
nice
because
I
reckon
so.
B
A
E
D
B
D
D
E
A
D
A
D
A
D
D
B
D
So,
to
summarize,
if
we
have
an
email
with
variant
and
the
trait
with
an
Associated
life,
also
named
variants,
if
you
then
project
with
so
variance,
then
at
the
moment
it
would
refer
to
the
Associated
type.
But
if
we
accept
the
site
there
sorry
the
enum
variant
types
RZ,
then
we
would
suddenly
had
the
variant
as
into
the
parking
space.
And
what
we
want
petrotech
implemented
is
that's,
because
environments
are
inherent,
think
things.
It
would
resolve
with
high
priorities
and
beam
invariance,
and
then
you
would
get
breakage.
D
D
E
E
You'll
get
a
you'll
get
a
duplicate
name
issue
if
it
happened,
because
the
two
things
in
the
same
namespace
that
we
wouldn't
be
able
to
disambiguate
in
anyway
yeah,
but
no
one
wants
to
do
that
anyway,
so,
okay,
but
for
now
people
can
do
as
trait
:
:
variant
to
get
their
associated
variant
type
if
they
really
want
it.
So
we
should
be
fine
yeah,
and
this
is
we
have
that
ability
warning,
but
it's
an
allowed
breakage,
so
we're
just
being
nice
about
it.
D
D
D
D
E
A
A
D
E
E
E
D
D
Ambiguity
it
might
be
become
in
the
future,
but
let's,
let's,
let's
up
nothing
to
Matthews
work.
D
D
A
I
think
there's
one
thing
I
wanted
to
discuss,
but
well
so
there
we
talked
about
2582
last
time,
but
it
seemed
like
we
didn't
quite
get
to
a
plan
and
I
was
noting
this
comment
by
Ralph.
There
seem
to
be
some
disagreement.
I
guess
abound
like
can
we
take
some
action
here
and
then
carve
out
the
rest
of
the
plan
ahead
of
time?
A
A
Yes
and
trying
to
sort
of
committing
to
that
as
a
syntax,
Bend,
entre
figuring
out
that
will
shape
the
rest
of
the
story
around
it
I.
It
does
feel
like
it's
a
potentially
large
change
in
some
sense
for
unsafe
code
authors
I
mean
it's
gonna
affect
a
lot
of
code.
I
think
the
rule
is
probably
gonna
wind
up
being.
You
should
just
use
this
like
by
default.
When
you
run
a
row
pointer,
even
though
there
are
only
a
limited
cases
set
of
cases
where
it
actually
matters.