►
From YouTube: 2021-06-16 RFC 2229
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
Okay,
so
I
went
through
quite
some
of
the
places
where
there
was
an
I'll,
let
it
load
one
second,
so
I
looked
at
the
places
where
the
increase
was
greater
than
200
and
greater
than
400
percent.
Just
as
like
my
base,
heuristic
of
what's
some
degree
of
significant
change,
and
I
think
most
places
are
looked
at.
A
Oh,
this
is
just
like
span,
gets
split
into
multiple
of
them
because
it
contains
commas.
So
it's
got
it
got
it
yeah.
Let
me
see
so
basically
yeah.
What
I
was
seeing
was
a
lot
of
the
places
the
fields
are
being
individually
captured
and
you
have
basically
some
immutable
reference
to
the
thing
and
you
end
up
capturing
a
lot
more
than
you
can
probably
get
by
with.
A
A
So
there's
the
concern
around.
If
you're
in
a
move
closure,
you
can
end
up
actually
causing
like
you.
Can
this
optimization
can
cause
a
regression,
because,
let's
say
you
have
some
a
root
variable
x,
which
is
a
reference.
It
doesn't
matter
what
you
capture
x,
dot
y
and
now,
since
this
is
a
move
closure,
we
will
actually
copy
x,
dot
y
and,
if
y
is
let's
say,
a
one
byte
like
a
u8,
it's
cheaper
to
copy
the
ui
than
to
copy
the
reference,
because
the
reference
will
be.
B
B
I
don't
think
that's
a
good
assumption
like
we'll
go
back
to
the
example.
You
had
earlier
this
one
yeah.
If
you
change
string
to
like
none
sure,
okay,.
A
B
Yeah,
okay,
I
would
prefer
that
we
capture,
all
of
but
with
with
the
println
example,
but
yeah
okay.
Well,
let's
see
what
does
it
do
because
that
has
a
bug.
So
this
tells
us
more
information,
yeah.
B
A
copy
reads
as
an
immutable
borrow,
so
I
mean
I
think,
that's
okay,
debating
about
this
right
now.
So,
okay,
there's
two
things
going
on.
One
is:
when
you
have
a
move
closure.
B
B
But
I
don't
think
that's
a
good
decision.
Okay,
maybe
we
could
make
that
decision,
but
one
reason
I
don't
think
that's
a
good
decision
is
that
just
because
it's
copy
doesn't
make
it
cheap
to
copy,
it
could
be
like
a
ua
1024
and
now-
and
I
remember
us
going
on
back
and
forth
about
this
right
like
if
that
were
boxed
or
with
the
code,
as
is
you're
copying
a
pointer
to
a
huge
array.
A
B
So
I
think
it's
better
to
say
it's
that
way:
it's
always
a
subset
of
what
we
copied
before
it.
Never
it
never
copies
more
data
than
it.
A
A
B
B
B
A
Let
me
use
zoom
chat
all
right.
How
do
I
send
things
I'll
descend
it
on
zulu?
Okay,
perfect.
I
will
paste
my
link
right.
A
There,
okay
yeah,
you
should
have
access
to
edit.
I
suppose.
B
Okay,
sorry,
let
me
find
that
you
sent
it
in
the
channel.
I
guess
yeah,
I
see
okay.
B
B
Why
one
reason?
Why
is
let
xbox
u8
1024
or
like
even
if
the
full
path
leads
to
a
copy
type
and
then.
B
Whatever
hold
on,
I
guess
x,
dot,
lan
or
something
is
a
good
example.
B
Okay,
so
that's
rule
number
one,
but
the
other
thing.
B
B
B
That's
true
for
this
case,
because
then
I'll
show
you
what
I
mean
by
this
case,
but
yeah,
because
you're
you're,
not
gonna
m.c,
it
doesn't
like
capturing
m.c,
doesn't
make
any
difference.
However,
a
case
where
it
might
matter
would
be
something
like.
B
B
Mm-Hmm
so
here,
if
we
capture
m
our
closure
does
not
outlive
kick
static,
let's
actually
hold
on.
Let
me
give
these
names
because
actually
that's
not
a
good
example.
B
B
It's
not
true
because
of
implied
bounds.
Sorry,
let
me
think.
B
B
B
A
B
And
there's
this
other
field
of
that
has
like
data
with
tick
b.
So
maybe
that's
a
problem
to
now
have
access
to
that
as
well,
except
that
I
think
I'm
wrong,
because
the
basic
well-formedness
rules
of
rust
kind
of
require
that
you
know
if
if
some
of
the
things
are
accessing
fields
of
destruct
and
others
are
accessing,
like
references
that
are
found
in
fields,
those
references
have
to
outlive
the
fields
of
the
struct
itself.
So
it's
actually
not
a
problem.
It's
only
if
you
yeah,
so
I
think
it's
like.
B
A
B
B
That
seems
true
borrow.
This
is
borrow.
B
B
A
B
A
B
B
If
it
is
a
draft
of
a
shared
ref,
you
can
truncate
p,
you
can
truncate
p
or
if
it's
a
df
of
shared
ref
and
all
that
follows
are
field
accesses.
You
can
chunk
a
p
to
remove
the
field
x.
That
may
be,
I
think
that
might
be
like
that's
the
only
thing
they
could
follow,
because
it
would
either
be
a
dref,
in
which
case
it's
not
the
right
most
where
it
would
be
fields,
but
maybe
there's
a
third
case.
I'm
not
considering.
B
So,
in
that
case,
for
like
example,
one
my
great
keyboard
turned
on
example:
one
caps
lock
key
turns
on
my
grid
keyboard,
but
I
hit
it
by
accident,
sometimes
example
one
that's
this
one.
B
This
is
example
two
example:
one
you
access
right,
star,
dot
a
this
is
really
a
borrow.
B
Star
m.a
and
then
you
you
basically
truncate
to
borrowing
or
to
yeah
borrowing
m.
I
guess
you
don't
want
to
borrow
m.
You
want
to
capture
m
by
value.
A
No,
I
think
we
want
to
borrow
them.
It's
like
the
benefit
of
borrowing
m.
Is
we
also
get
it
on
the
fat
pointer
issue?
We
discovered
talking
about
last
time.
B
B
A
Because
like
if
we
also
think
about
what
would
have
happened,
if
assuming
m
was
just
a
root
variable
and
what
we
would
have
done,
without
this
precise
analysis,
we
would
have
borrowed
m.
So
we
would
have
always
borrowed
a
reference
to
the
reference.
B
A
B
A
B
B
It'll
just
make
our
it'll
just
make
less
work
later
on,
like
we're
not
losing
any
precision.
Yes,
that
way:
okay,
so
then
yeah
and
then
star.
So
basically
in
example,
one
we
would
both
of
these
would
would
wind
up
getting
truncated
to
the
same
path.
So
we
would
wind
up
with
like
as
if
we
only
accessed
one
thing
m
by
value
right.
A
B
A
B
A
B
B
B
So
that
means
it
can't
be
producing
any
borrow
check
errors
by
the
fact
that
we're
now
cop
like
moving
it
in.
So
I
think
I
think
this
is
a
safe,
optimization.
B
Do
we
have
a
plan
to
who
wrote
me
here?
Okay,
tie
this
to
the
edition.
A
Yes,
I
was
gonna
bring
that
up
as
well.
I
don't
have
any
problem
to
tie
this
with
the
edition.
I
want
to
discuss.
What
all
do
we
tie
to
the
edition
right
now?
Is
it
just
sorry?
Is
it
just
that
the
feature
is
enabled
by
default
or
do
we
also
want
the
link
to
be
start
triggering
on
edition
2018
as
a
future
and.
B
B
A
You
can
so
there's
a
lint
group
thing
that
is
industry
lynch,
so
you
can
just
like
list
it
as
part
of
another
lan.
A
B
B
It's
a
migration,
so
I
think
that
should
be
refactored.
However,
that's
an
orthogonal
problem,
as
far
as
I
know,
that's
used
that
metadata
is
used
for
for
all
edition
stuff
right.
A
A
So
and
we
set
the
filter
incompatible
and.
A
Okay
yeah,
so
this
is
all
I
have,
I
think,
drove
didn't
get
time
to
work
on
the
was
it
called
where's,
the
the
migration
where
we
want
to
figure
out,
oh
yeah.
Actually,
this
is
also
a
thing
that
we
need
to
do,
but
figure
out
the
drop,
the
significant
and
insignificant
destructors.
Sorry,
I
was
forgetting
words,
and
then
we
also
have
this
issue
that
we
had
previously
talked
about
where
we
have
a
structure
that
contains
two
drop
types.
A
Now
the
drop
has
been
reordered,
because
now
we
will
drop
p
dot
y
before
we
drop
v
dot
x,
which
previously,
if
p
was
completely
captured,
would
drop
v
dot
x
before
it
would
drop
with
our
y.
So
you
basically
need
to
sort
of
sort
through
our
list
of
captured,
precise
paths
to
maintain
yes,.
A
A
Yeah,
so
it
shouldn't
be
that
bad
yeah.
I
think,
that's
mostly
what
what
we
have
remaining
I'll
file,
the
bug
for
the
boxing
right
now
and.
A
A
Yeah,
so
I
think
that
is
all
I
have
for
questions,
so
we
want
for
the
data
that's
generated
out
of
the
actually
closure
size.
Actually,
you
realize
I
have
a
couple
questions
on
that
now
that
I
forgot
one
when
we
it
will
probably
be
so.
A
Are
we
I'm
assuming
we're,
expecting
people
to
use
this
with
cargo
and
they
said
the
rust
flag
like
pass
cargo,
the
zed
profile
player
flag,
and,
in
that
case,
from
what
I
understand
it
will
do,
it
will
only
generate
it
for
the
current
project
and
not
for
any
of
the
dependencies
and
stuff,
which
means
we
will
lose.
A
A
B
B
Besides,
I
think
a
lot
of
crates
like
saturday
and
so
forth,
wind
up
generating
code
in
the
user's
crate
right
so.
B
A
All
right,
yeah,
okay,
that's
and
then
we
just
generate
a
csv
out
which
contains
okay.
Also,
I
wanted
to
discuss
was
do
we
want
to
give
the
user
just
an
option,
just
called
z
like
profile,
and
they
can
just
set
equals
no
collect
so
that
we
don't
collect
spam
information,
or
do
we
just
want
to
like
tell
them
to
not
share
like
like,
I
feel
like
them
go
having
to
go
remove.
The
span
is
going
to
be
painful,
just
because
there's
so
much
data
that
will
be
there.
B
Why
would
they
I
don't
want
them
to
delete
it
like?
We
might
want
to
look
at
that
data.
I
think
we
should
tell
them
that
it's
just
be
advised
it's
going
to
contain
spam
information
and
may
reveal
the
name
of
your
directory.
A
A
B
A
B
A
B
It's
not
so
hard
to
do
anybody.
A
Yeah,
okay,
yeah!
That's
all
I
have
do
you
have
any
questions
for
me.
B
My
only
question
is:
are
the
like,
what's
the
status
of
the
various
folks
in
the
group,
is
it
just
you
now?
Are
there
other
people
still
working.
A
So
drove
started
work
so
he's
probably
busy,
but
I
think
he
said
he
will
get
started
on
the
incident
destructor
soon.
I
think
right
now.
It's
just
me
and
I
think
roxanne
was
on
vacation
but
should
be
back
soon.
So
I'll
check
in
with
her.
B
I
just
want
to
make
sure
that
you're
like
if
it
happens
that
your
available
bandwidth
is
dropping
very
low.
You
know,
since
the
timeline
is
starting
to
matter,
do
let
me
know
and
I'll
try
to.
A
B
Is
there
any
other,
I
think
the
only
blocker
for
like
people
testing?
The
edition
is
the
tie
to
the
current
edition
issue.
B
In
this
bug,
you
just
yeah
just
discovered
but
yeah,
okay.
A
We
also
had
this
one
issue
that
we
wanted
to
do
as
a
polish
complete,
which
was
oops
so
one
second,
yes,
so,
basically
simple.
Example:
this
code,
which,
like
basically
when
you
course
a
closure
to
a
function
pointer,
it.