►
From YouTube: Lang Team Triage Meeting 2019.09.12 -- with WG-Grammar!
Description
Focus area for this meeting was the Grammar Working Group (comes at the end). We talked about their goals, history, and what we can do to better integrate the Grammar WG into the team as a whole.
Other notable discussions:
- unwind, #63909
- the Ref type and noalias, #63787
A
A
A
So
we
have
still
happening.
We
have
a
couple
I
think
we
have
two
main
thing:
we'll
go
through
the
triage
chickens
and
then
we'll
figured
out
anything
else
on
this
list
that
didn't
happen,
conflict
on
function,
pointers,
I,
wrote
and
then
were
to
wait.
You
late!
That's!
Okay.
Did
you
write
a
comment?
Improving
sixty
nine,
oh
nine
I.
Guess
we're
going
to
talk
about
that
judge.
I.
A
Right
so,
let's
just
move
in
behind
issues
brief
update.
We
have
this
problem.
We've
discussed
in
the
past,
around
trade
objects,
I
kind
of
worked
out
a
series
of
mentoring
set
when
trying
to
find
someone
to
mentor
and
or
do
it
myself,
I
think
I
found
a
few
candidates,
that's
all
they're
released
to
say,
they're
progressing
slowly,
it's
been
a
bug
since
Ross
won't
point
out,
I,
guess
the
nominated
PRS,
so
there's
a
PR
Warren
if
include
macros.
A
Apparently
the
include
macro
reads
in
the
contents
of
the
file
and
parses
the
first
expression
and
then
discards
the
rest
of
the
file.
This
is
potentially
surprising
and
probably
not
the
behavior.
Anyone
actually
expects
and
right
now
it
doesn't
do
anything.
This
PR
is
written
as
as
described
gives
a
if
there
are
extra
tokens
in
the
file
as
actually
written
aborts
compilation.
If
there
are
extra
tokens
in
the
file
and
thus
meets
the
breaking
change,
and
the
question
is
what
behavior
do
we
expect
to
do?
A
I
wrote
down
what
I
think
are
probably
the
three
reasonable
choices
warning
and
preserving
the
current
behavior
for
better
or
worse,
probably,
worse,
erroring
and
maybe
through
a
future
compatibility
warning
for
nervous
or
do
a
more
extensive.
We
right
where
we
actually
parse
all
the
tokens
and
try
to
interpret
this
as
a
sequence
of
expressions
rather
than
individual
expression.
That
sounds
kind
of
hard
I
think
we
could
probably
scratch
that
off
the
list
and
not
worth
the
effort
is.
A
C
A
C
A
C
A
A
D
A
So
there's
an
include,
stir
macro:
it
does
not
normalize
new
lines.
This
proposal
is
to
make
it
normalize
new
lines
and
compatibility
with
screen
Constance.
Last
time
we
talked
about
this,
you
mentioned
RFC.
There
was
some
mild
pushback
that
this
like
overkill,
we
did
the
spring
constants
and
nobody
complained.
A
C
A
A
A
A
A
C
Thread
yeah,
so
the
consensus
in
our
previous
meeting.
It
was
that
the
state
that
we're
in
is
kind
of
unfortunate.
We
reverted
the
abort
on
unwind
out
of
an
extern
function
in
order
to
avoid
breaking
people's
current
usage.
But
that
then
means
there
is
a
real
possibility
of
Mis
compilation.
If
you
do
use
that
behavior,
because
we
still
tagged
functions
as
no
unlined
it.
So
the
consensus
from
the
previous
meeting
was
hey.
We
should
either
abort
and
tag
things
no
unwind
or
don't
abort
and
don't
tag
things
no
unwind
it.
C
That
said,
there
was
some
a
lot
of
further
discussion
and
thrash
on
discord
and
a
half
dozen
issue
thread
since
then,
and
I
would
like
to
double-check
and
make
sure
that
nobody
is
especially
considering
it
a
problem.
If
we
merge
this
and
in
particular
I,
don't
think
we
want
to
wait.
10
days
to
merge
this.
A
One
thing
I
know
essential
mentioned
product
that
they
had,
except
they
had
a
change
of
heart.
They
didn't
specify
about
what
exactly
I
know
that
I
personally,
along
with
it,
sounds
like
you're,
Josh
and
I-
think
central
as
well
feel
differently
about
what
the
RFC
should
contain
and
then
I'm
now
pre
persuaded
the
right
things
to
add
in
EDI
and
not
to
have
an
attribute
I,
don't
know
whether
they
meant
that
or
whether
they
meant
something
about
this.
C
C
Other
was,
and
I
will
acknowledge.
Here
I
am
conveying
a
point
of
view
that
I
don't
agree
with,
but
I'm
attempting
to
convey
it
as
accurately
as
possible.
Sentro
felt
that
he
would
rather
leave
the
undefined
behavior
in
place
for
a
while
longer,
because,
if
we
just
drop
no
unwind
so
that
we're
in
a
consistent
state,
it
removes
the
impetus
to
actually
do
something
and
we've
spent
years
not
doing
something.
So
he
wanted
to
leave
the
you
be
in
place
to
motivate
people
to
make
a
real
RFC
fix.
A
D
A
Me
be
more
clear
reasons
to
make
it
continue
to
be
you
be
to
unwind
through
this
new
boundary
in
which
is.
The
other
reason
would
just
be
that
it
sends
if
you're,
not
cleaning,
very
close
attention-
and
all
you
see
is
that
we
removed
we
made
it
non
UV.
You
might
get
the
wrong
message
like
our
goal.
Is
that
it
should
be.
A
You
me
eventually,
so
it's
sort
of
confusing
that
we
take
this
backwards
dub,
instead
of
saying
only
the
status
quo,
where
it's
been
since
1.0
of
its
technically
you
be,
as
it
happens,
most
people's.
There
seem
to
be
some
people
who
are
doing
it
and
getting
away
with
it,
because
you
bees
not
defined
to
do
anything
in
particular,
nothing
that
I
could
see
that
I,
don't
I,
don't
have
a
strong
opinion
about
it,
but
I
think
it's
a
reasonable.
A
A
C
I,
though,
if
nobody
else
has
I
wanted
to
give
a
minute
there,
but
if
nobody
else
has
a
strong
opinion
on
this,
I
personally
feel
slightly
the
opposite,
that
it
would
actually
add
more
impetus
to
get
an
RFC
done
if
we
do
an
630
909,
because
it
means
we
are
missing
out
on
an
optimization
opportunity
and
getting
that
optimization
opportunity
back
need
would
then
require
us
to
get
a
proper
solution
for
this.
You
know
I
see
when
we
land
and
beta
nominate,
6-3,
909.
E
C
D
A
A
while
till
we
check
this
base
sad
fact,
the
matters,
usually
the
guests
are
waiting.
I
feel
like
our
there's,
a
that's
just
a
general
problem
versus
by
the
time
the
boxes
get
checked.
That's
like
it's
usually
a
long
time
we
if
they
were
at
getting
checked
the
way
we
originally
imagined.
It's
like
very
fast.
C
D
A
C
A
D
I
was
a
timeline,
and
the
timeline
for
the
next
release
is
close
enough.
That
I
don't
know
how
much
people
care
about
that
in
terms
of
making
these
decisions,
but
I'm
personally
think
there's
enough
other
things
that
are
breathing
down
my
neck
for
the
next
release
that
I'd
that's
the
reason
I
would
say:
let's
not
back
port
right
now,
Thrones.
A
I'm
getting
it,
the
only
thing
I
would
also
add,
is
Josh
and
you're
right
on
I
think
making
it
clear
if
that's
a
good
write-up
and
you
want
to
post
in
the
release,
notes
or
something.
So
you
know
this
should
have
some
text
that
makes
clear
in
the
future
time
the
future
plans
to
make
this
movie
again,
I
kind
of
feel
like
whether
it's
backward
or
not.
Isn't
that
important
it's
been
around
since
one
knows,
I
said
it's
more
about
declaring
my
intentions:
I,
don't
care
I
wouldn't
be
opposed
if
we
can
make
it
work.
A
D
Worried
that
I'm
about
to
reopen
this
issue
that
we
just
tucked
into
the
clothes
board
board
are
finished
with,
but
I
want
to
make
sure
I
understood
what
you
just
suddenly
go
about
this.
This
is
undefined
behavior
again
in
the
future.
We
understood
Josh's
motivation
for
saying
why
we
could
land
Ralph's
thing
was
to
say:
oh,
it's
actually
coulda
still
like
his
impetus
for
fixing
things
properly,
because
landing
Ralph's
thing
will
be
a
pessimism.
Swords
it'll
cause
some
code
that
could
actually
be
optimized,
submiss,
that
optimization
and
so
they'll
be
they'll.
D
A
A
A
Basically,
it
continues
to
be
you
be
from
rusts
perspective,
however,
the
current
compilation
step,
like
from
all
of
us,
we're
removing
the
UV
from
LMS
perspective,
and
we
are
saying
that
in
the
future
it
will
be
defined
as
networked,
which
is
sort
of
a
subset
of
today's
possible
legal
behaviors,
and
there
will
be
an
alternative
way.
So
the
bottom
line
is,
we
are
planning
unless
something
changes
to
break
people
who
are
relying
on
unwinding
and
we're
playing
to
bake
them
in
a
hard,
visible
way.
C
A
A
D
I
just
got
this
impression
of
a
slicer
of
limping
along
being
like
we're,
defining
to
be
like
the
specs
of
this
undefined
behavior,
but
we're
gonna
keep
sort
of
supporting
this
use
case
and
doing
slight
somersaults
that
can
you
supporting
it
while
stating
our
intentions,
long
term
er,
to
stop
doing
such
some
ourselves
really?
This
is
we've
done
what
Central
is
saying
about
what
not
wanting
to
land.
A
D
A
A
A
We
currently
decorate
shared
references
with
the
no
alias
perimeter.
They
know
this
attribute
in
LOM
and
because
the
idea
is,
although
there
may
be
other
references
to
the
same
memory,
only
beads
are
allowed
and
they
don't
interfere
with
this
from
Elliott
Williams,
no
alias.
However,
this
is
not
necessarily
correct
around
ref
stalls
and
in
particular,
does
this
work
yeah.
Oh
sorry,
I
have
this
one,
and
this
actually
gets
back
to
one
of
the
first
problems
we
discovered
around
math
in
particular.
Arrests
currently
in
beds,
they've
shared
wetlands
inside
of
it.
A
Actually
that
reminds
me
that
we
can
do
a
very
so
there's
two
problems
here.
We
could
do
a
targeted
fix
on
ref.
We
could
change
ref
to
use
Garcons
instead
of
a
purse
and
a
and
everything's
fine.
Basically,
because
now
we
are
no
longer
promising
that
we
have
a
reference
with
lifetime
today,
which
is
a
lie,
but
more
generally
like
there's
this
question
of
what
about
other.
What?
A
If
someone
had
written
grass
all
themselves
and
would
it
be
allowed
which
I
guess
we
can
defer
a
little
bit,
one
I
think
one
other
option
would
be
that
we
could
just
be
more
conservative
in
how
we
use
no
alias
attribute
on
sort
of
say,
like
it's
still,
okay,
to
put
no
alias.
If
you
have
something
like
this
like.
A
If
the
ref
is
at
the
top
level,
then
it
can
be
considered
no
alias,
but
not
when
it
gets
and
Bennett
instructs
and
pings
where
it
becomes
less
clear,
then
we
have
to
decide
what
those
rules
are.
This
also
comes
back
to
Rossi
aliasing
rules,
because
no
Elias
hasn't
and
now
I
guess
thing.
The
stack
borrows
all
day.
I
forget
was
ralph
called
them,
but
these
function,
barriers,
I,
don't
think
we're
gonna
reach
this
hole.
A
A
C
D
A
A
A
A
E
A
A
D
Well,
it's
a
question
about
whether
we
like
sort
of
re-litigate,
our
overall
philosophy
on
each
one
and
let
people
sort
of
pet
interests
guide
them
right.
It
gets
pretty
tiring
weekend
or
are
we
gonna
try
to
have
like
a
you
know,
kind
of
kind
of
constitutional
articles
saying
this?
Is
our
general
approach
right?
Let
us
let
our
Supreme
Court
litigate
in
the
future,
I
think.
A
I'd
prefer
a
rule
of
thumb,
that's
like
unless
there
are
compelling
reasons.
Otherwise
we
will
generally
take
this
approach,
but
all
right,
I,
don't
think
we're.
If
four
quorum
to
debate
a
general
rule
again,
we'll
keep
going.
I
didn't
look
at
these
bugs.
Yet
that's
incorrect
undefined
behavior
and
Static
residences
created
from
the
pointers.
I,
don't
know
Taylor.
Do
you
remember?
What's
going
on
here,
I
see
you
coming.
A
A
A
C
A
F
That's
Taylor
I've
been
trying
to
talk
for
a
little
while
now
and
I
didn't
realize
you
couldn't
hear
me
the
whole
time,
but
that
makes
sense
know
why,
when
you
asked
me
about
the
undefined
behavior
on
static
references
craving
from
pointers,
thing
and
I
started
to
answer,
you
started
talking
over
me
like
I'm,
just
talking
to
myself.
Okay,
yes,.
F
F
C
A
A
A
A
C
A
C
A
F
F
Alright
but
yeah
I
mean
when
we
talk
or
like
basically
the
original
stabilization
included
customer
like
nested
receivers
and
then
when
I
went
to
actually
do
it,
they're,
like
oh
actually,
I've
removed,
custom
receivers
or
nested
receivers
and
nobody
had
asked
for
it
to
be
removed,
and
so
just
like
didn't
get
included
in
the
stabilization
because
it
was
more
stuff
than
was
necessary.
But
there's
no
like
problem
with
them.
Yesterday.
A
A
E
Thank
you
guys.
Well,
we
don't
have
a
lot
of
updates
on
our
front
like
we
had
the
summer,
which
was
as
low
as
usual,
but
essentially
we
have
a
D
movie,
Drona
generated
cogeneration
style
implementation
for
GL
and
going
with
grammar,
which
is
kind
of
like
interpreter
or
the
proposal
that
we
have
for
our
grammar.
You.
E
A
E
Stepping
a
lot
back,
we
started
this
grammar
group
to
try
to
specify
the
language
in
the
sense
that
we
have
a
grammar
where
the
link
team
or
whatever
other
person
that
wants
to
propose
change,
will
be
able
to
validate.
If
this
thing
is
going
to
break
or
not,
the
language
for
mark
I
also
design.
This
kind
of
extending
so
making
sure
that
we
have
ways
to
compare
with
whatever
restaurant
Eliezer
is
doing
or
even
worse,
he's
doing
like
making
sure
that
our
grammar
actually
can
check
reaching
the
others
and
make
sure
the
vehicle
systems.
E
E
E
E
E
Also
we
have.
He
has
working
on
like
okay,
so
let's
formalize
what
whatever
we
are
talking,
so
they
did
a
bunch
of
fishes
on
the
repo
that
you
probably
can
see
on
the
grammar
area
like
there's
a
bunch
of
concerns
and
how
we're
going
to
define
this,
how
we're
going
to
deal
with
that?
So
essentially,
we
have
like
a
bunch
of
questions
that
were
tackling
slowly.
A
So
I
think
this
is
like
awesome.
A
couple
of
questions.
I
have
I
think
when
sort
of
like
the
unsafe
good
guidelines,
we
were
lacking
like
a
good,
steady
feedback
process.
Where
look
I
would
imagine
that
what
happened
is
you
would
be
working
through
a
lot
of
stuff?
Then
you
would
encounter
some
questionable
stuff
like
here's.
Some
yeah
here
are
two
possible
parses
rusty.
Currently
does
this,
it
seems
a
little
weird.
Even
not
we
actually
want
grant.
Does
that?
Maybe
we
want
to
surface
that
to
a
broader
discussion
and
get
more
feedback.
E
Or
not
so,
essentially,
we're
like
on
that.
You
know
the
bootstrapping
problem
we
kind
of
have
to
have
the
very
minimum
working
to
a
point
where
we
so
we're
making
progress
towards
the
goal,
but
we
are
not
near
a
point
where
we
can
actually
come
and
start
showing
it
more
in
a
more
practical
way
sure
it's
always
good
to
give
you
it
back
to
the
to
the
link
team,
but-
and
this
is
something
that
we
should
start
doing
now,
which
is
actually
great
likely
discuss
these
things
with
y'all
I
I
met
with
Taylor
at
Rusk.
E
A
E
Yet
but
yeah,
that's
that's
clearly
part
of
people.
We
want
to
be
able
to
actually
generate
the
reference
and
for
something
like
this.
It's
kind
of
having
getting
that
corpus
of
information
that
we
have
today
that
people
actually
working
and
discuss
at
the
edge
case
is
likely
being
able
to
but
teasing
to
take
remar.
A
E
E
D
E
That's
my
point,
but
essentially
we're
just
chatting
randomly
oh
no
discord
channel
and
is
one
of
the
leads
on
that
team
fails
to
post
frequent
updates
to
internals
because
of
course
you're
just
there
we
were
talking
in
centuries.
There
edges
there,
but
it's
not
like.
We
never
came
to
a
meeting
and
actually
had
this
as
part
of
the
agenda.
You
know
so
it's
kinda
making
increasing
the
transparency.
Let's
see,
I
think
posting
to
internals.
A
E
C
E
A
E
E
Want
to
see
something
parsing
in
things,
you
probably
can
take
a
look
on
the
tree
API,
which
is
another
discussion,
because
essentially
we
have
we're
like
parsing,
the
entire
source
resources
and
it's
taking
like
some
time
and
we
had
like
an
evolution
on
time.
Another
sub
project
that
was
there
but
again
way.
E
So
many
things
in
flights
to
Tyco
is
like
okay,
let's
try
how
long
we
take
to
parse
this
and
how
reasonable
is
like,
because
if
the
parser
is
way
way
too
slow,
what's
the
point,
but
then
we're
like
no
we're
not
using
a
parser
we're
building
the
language
specification.
So,
let's
step
back
a
little
bit
and
focus
on
what
is
more
important.
So.
A
E
E
E
E
A
E
A
A
E
A
A
E
D
E
E
E
It
there's
a
tone
like
a
bunch
of
papers
and
on
the
GOL
repo.
There
is
like
one
issue
that
it's
specifically
documentation
where
it's
like:
they're
all
links
to
research
papers
on
the
topic
like
how
to
parse
what
can
do
better,
etcetera,
yeah.
We
look
at
over
them
and
kind
of
like
ended
up,
because
one
of
the
candidates
at
the
time
was
to
just
use
a
lollipop
and
potentially
Paciello
gll
as
a
back-end
for
a
lollipop
which
is
went
nowhere
and
we
kinda
ended
up.
A
Yeah
I
mean
the
day.
The
plus
side
would
be
that
the
answer
of
this
question
would
be
clear.
Just
the
dot
side
would
be
that
the
answer
would
be
use.
Lr1,
which
is
like
honest
Odin,
is
human
friendly
form
of
disambiguation
right.
So
all
right,
so
I
guess
you
could
say.
Yes
if
there
were
people,
especially
like
you
know,
grammar
nerds
or
someone
who
enjoys
yeah.