►
From YouTube: Rust and WebAssembly Working Group Meeting 2019-04-11
Description
B
A
C
B
A
Sort
by
newest
this
is
no
I
guess
nothing
is
new
yeah.
So
if
anyone
has
some
time
I
think
a
bunch
of
these
are
not
actively
in
flight.
So
there's
stuff
to
pick
up
here.
A
So,
as
far
as
the
glue
toolkit
goes,
I
don't
think
we've
landed
anything
this
week,
but
we've
spent
a
lot
of
time
talking
about
versioning,
which
we
kind
of
touched
on
a
little
bit
on
last
week
and
I
think
we've
resolved
that
now
and
so
kind
of
with
that
resolved.
The
next
question
is:
what
do
we
want
in
an
initial
release,
and
so,
if
you
have
opinions
about
what
should
be
an
initial
release
of
glue
or
not,
there
is
a
thread.
Let
me
pull
it
up
here.
D
A
D
E
Don't
think
there's
been
a
lot
of
stuff
for
public
consumption,
yet
just
in
terms
of
like
we're
still
assembling
and
trying
to
make
it
no
actual
usable
form
and
so
I
suspect.
Soon
we
might
have
some
interesting
news
about
debugging
y'see
code
through
West
and
time,
but
certainly
not
that
web
browser.
Yet
we're
still
working
towards
that.
A
Speaking
of
I've
been
meaning
to
start
implementing,
or
at
least
like
a
skeleton
of
the
debugging
debug
info
support
for
walrus,
but
clearly
that's
not
been
happening.
I've
been
fairly
busy,
so
I
don't
know
how
many
cycles
you
have
Alex,
but
it
seems
like
something
that
maybe
you
could
steal
from
me
if
you,
if
you
do
have
cycles
yeah.
E
A
A
B
So
there's
a
couple
things
I
was
hoping
that
my
like
open
source
day
would
be
a
little
bit
early
its
earlier
this
week
than
it
was,
but
it
wasn't
so
today
is
that
day
and
so
I
have
a
generate
PR
open,
there's
a
couple
of
things
mostly
Docs
and
tests
that
need
to
be
taken.
Care
of
the
good
thing
is
that
the
vast
majority
of
the
work
was
kind
of
generalizing.
Our
install
implementation
and
so
adding
new
stuff
that
we're
gonna,
like
bin
stall
and
run,
should
be
nice
and
easy.
B
B
So
if
you're
interested
in
that,
we
should
take
a
look
at
it
and
we
might
want
to
try
and
come
up
with
I.
Don't
know.
I'm
gonna
take
a
look
at
exactly
what
they're
doing
and
see
if
it's
something
that
we
might
want
to
do
like
an
RFC
to
try
and
support
or
not
I
know
that
using
seed
EPS
and
compiling
Russ
to
was
another
day
to
I'm
gonna
know
people
want
to
do
and
to
come
up
a
couple
times.
So
I
thought
it
was
interesting.
B
A
Hope
is
that
resi,
since
that's
kind
of
taking
there,
has
to
be
an
ABI
I,
don't
know
if
the
API
is
technically
part
of
huazi,
but
as
he
has
to
have
like
a
standard,
ABI
and
I'm,
hoping
that
that
is
what's
going
to
enable
kind
of
C
interoperation
for
us,
and
we
can
kind
of
just
not
have
to
worry
that
much
about
this
issue.
Yeah.
B
A
E
Actually,
everything
works
today,
it's
more
of
there's,
no
definition.
There's
no
standard
agreement
about
what
a
C
compiler
for
wasm
even
is
it's
like
Y
Z
is
going
to
have
a
standard
system.
Scripted
has
the
standard
sis
route.
What
was
it
doing
on
own
there's,
no
sister,
there's
no
C
compiler
and
so
you're,
just
kind
of
on
your
own
to
go,
find
stuff
and
like
there's
no,
and
that
wouldn't
with
it
more
quickly
means
as
there's
no
C
libraries.
So
there's
no,
like
you
have
to
bring
your
own.
E
You
have
to
do
like
everything
like
printf
and
like
malloc
and
free
everything
like
that,
and
so
that's
not
what
any
existing
C
code
assumes.
All
existing
C
code
assumes
some
major
system,
header
files
and
all
that,
and
so
we
can
it'll
work.
Actually,
if
you
use
clang
and
just
produce
native
Wisel
files.
B
E
That'll
work,
but
it
probably
isn't
what
you
want
cuz.
You
probably
want
to
do
like
some
existing
C
project
and
you
just
want
to
compile
it
to
a
zoom
and
have
it
run,
but
the
chances
of
like
having
all
the
header
files
having
all
the
symbols,
getting
everything
wired
up.
It's
probably
pretty
small
I
feel
like
we
can
write
documentation
for
this
for
the
one
target.
It's
gonna
be
basically
full
of
caveats
of
like
by
the
way,
this
probably
doesn't
work,
and
it's
probably
it's
probably
not
what
you
want
the
Weslie
talking.
E
However,
weed
is
like
oh
there's,
also
the
broader
question
of
where
the
hell
do.
We
even
document
these
kinds
of
things
I
know
that
the
Y
Z
target
has
documentation
in
the
target
file
in
the
compiler,
which
no
one
ever
reads,
so
it's
a
terrible
place
for
documentation
and
so
having
a
official
location
for
like
I'm
wasm
see
what
are
my
options
and
because
also
describing
script
in
there
and
kind
of
layout,
so
I'm
figuring
that
out
might
be
good
to
do
and
we
could
just
dump
all
that
information.
There
I.
B
E
I
feel
like
we
want,
like
I
mean
rust
in
general,
doesn't
have
a
lot
of
target
specific
documentation,
but
we
want
that.
You
know
like
we
want
that.
We
want
some
like
not
even
and
in
the
sense
that
there's
not
already
a
venue
for
us
to
put
this
in
rust,
but
we
want
like,
if
you're
interested
in
wisdom,
we
just
want
a
general
reference
guide
style,
documentation
somewhere,
and
we
can
say
things
like
here's
the
Cindy
proposal
status,
here's
the
Atomics
process,
here's
the
C
proposal
status.
We
don't
we
have
no
precedent
for
that.
E
So
we
could
just
make
precedent.
We
could
figure
out
something
would
want,
but
it
would
be
good
to
have.
It
probably
would
be
good
to
have
like
if
you're,
just
interested
in
the
wasm
target.
Here's
a
reference
not
really
a
tutorial,
not
really
like,
wasn't
buying
generalize
and
PAC
oriented.
Just
here's
anything
about
it.
We.
B
Said
maybe
at
a
tracking
issue
for
this
I
think
I
don't
know
or
the
doctor
I
don't
know
the
right
location
necessarily
is
but
I
think
tracking.
This
sounds
like
a
good
idea,
because
this
is
often
stuff
that
we
say
to
people
in
conversation
like
over
and
over
again,
but
like
don't,
have
written
down
anywhere
yeah.
E
E
B
E
B
So
I
have
a
couple
of
other
things,
so
you
have
a
generate
PR.
It's
open
today,
I'm
also
going
to
do
the
NPM
depths
PR,
but
one
of
the
things
I
realized
is
we
had
this
like
feature
request
that
I
actually
can't
necessarily
find
an
issue
filed
for
us,
but
I.
Might
it
might
just
be
some
long-ago
I
forget
the
key
word
find
it,
but
we
had
this
idea
that
you
have
your
source
directory.
Where
you
write
all
your
stuff.
B
You
know
the
package
directory
with
all
the
generated
stuff
goes,
but
what,
if
you
wanted
to
like
edit
your
package,
JSON
or
add
some
sort
of
like
static
other
GIS
file,
or
something
like
that?
We
wanted
to
have
this
kind
of
like
assets,
directory
that
we're
gonna
generate
the
thing
into
the
package.
B
Json
then
copy
the
stuff
over
from
the
assets
directory
and
update
a
package
JSON,
and
so
this
came
to
mind
when
I
realized
that
with
NPM
deps
thing
I'm
going
to
be,
you
know
already
editing,
package.json
and
so
I
wanted
to
ask
the
group.
Do
you
think
that
we
need
this
tool
and
the
NPM
depth
thing
or
no
I.
C
E
Is
one
I
think
the
attempt
was
to
be
conservative
enough,
so
we're
forward
compatible
with
whatever
we
want
to
do
later,
and
so
right
now,
as
I
mentioned,
will
disallow
anything
other
than
the
dependencies
key
inside
of
a
package.json.
So
that
was
explicitly
so
we
can
source
or
something
how
rationalize
merging
these
later
on.
B
Know
about
poison
by
gen
only
in
doing
dependencies.
Keith,
that's
huge
and
makes
me
feel
a
lot
better,
because
I
can't
be
a
hundred
percent
sure,
but
I
feel
relatively
confident
that
we
have
a
fair
number
of
people
editing
in
a
package.json
in
their
package
directory,
which
is
like
a
kind
of
like
they
shouldn't
do
that,
but
also
we
haven't
given
them
another
option.
So
I.
B
That
if
we
start
editing
that
and
they've
already
edited
in
like
a
real
gross,
but
if
we're
really
conservative
on
the
wasn't
button
inside
I
mean
I
still
might
try
and
like
get
that
in.
But
that's
good
to
know
so.
I
was
worried
that
I'm
gonna
be
like
messing
with
some
people's
workflows
and
then
yeah.
That
was
all
I
had
there's.
Definitely
a
couple
things
on
that:
wasn't
pecks
print
that
could
use
people
to
pick
them
up.
So
if
people
really
want
to
do
that,
please
get
in
touch
that
would
be
awesome.
Oh.
D
E
It
was
we
so
on
Firefox,
it's
one
of
the
only
browsers
that
unstable
uses,
text
encoder
and
code
into,
or
has
support
for
that.
And
so
we
attempt
to
use
that
which
is
a
more
efficient
method
of
transferring
strings
from
Jas
to
rust.
And
but
it
was
a
bug
in
that
code
and
wasn't
very
well
tested,
because
we
didn't
actually
have
any
testing
because
we
didn't
have
any
implementations
of
tests.
And
so
it.
E
In
there
were
factoring
of
that
which
caused
entry
strings
to
only
get
passed
to,
and
so
it
was
just
tweaking
on
a
few
lines
and
fixing
that,
and
so
that
should
be
out
that
also,
this
is
the
first
release
that
has
support.
There
was
the
stood
web
folks
we're
doing,
supporting
and
getting
a
proof
of
concept
of
working
on
top
of
ways
of
mind.
You
know
I
think
they
had
like
I
have
a
world
staff
thing
working,
and
so
this
release
is
the
first
one
that
that
would
support
all
them.
E
E
You'd
get
a
terrible
error
messages
like
this
thing
we
ever
hear
isn't,
doesn't
implement
copy
and
it
doesn't
actually
what
you
would
rather
have
as
a
point
to
the
field
saying
this
field
doesn't
copy
this,
where
you
obviously
know
what's
going
on,
and
so
that
was
an
example
of
like
where
message
that
could
be
greatly
approving.
This
doesn't
take
too
much
effort
on
our
part
that
we
just
gotten
just
know
what
what
errors
to
be
improving.
E
B
If
you
see
a
bad
error
message
file
and
like
it's
really
easy
to
transfer
issues
so
like
wherever
they
file
is
fine
like
move
it
wherever
we
need
to
move
it,
but
I
do
find
that
a
lot
of
people,
especially
if
they're
trying
to
it
like
it's
around
like
a
new
kind
of
edgy
or
technology
or
a
little
more
reticent
to
like
feel
like
this
error
message,
sucks
so
I
think
explicitly
calling
it
I
wouldn't
be
like.
No.
No,
please
tell
us,
would
be
huge.
E
I
don't
know
if
I
said
this
last
time,
but
along
y'see
things,
the
community
group
has
been
sort
of
formed
it's
in
the
process
of
being
form,
but
if
you
go
to
github.com,
slash
webassembly
/
y'see,
that
is
the
new
home
for
all
Y
Z
discussions
and
whatnot.
So
if
you're
curious
to
kind
of
be
in
a
standardization
effort,
well,
that's
a
good
place
to
go
and
then
also
on
the
Y
Z
target.
We
have
file
system
support
in
Russ
nightly
now.
A
A
A
They
have
like
kind
of
intrinsic
functions
that
you
use
in
your
in
the
LLVM
ir
to
like,
attach
to
bug
info
to
specific
things
and
there's
as
far
as
I
could
tell
there's
sort
of
like
two
kinds
of
those
calls
that
you
make
where
you
actually
are
attaching
to
like
local
SSA
values
or
memory
locations.
There's
there's
like
this
variable
or
this
SSA
value
is
holding
this
like
source
level
variable
or
something
like
that
right
and
then
there's
kind
of
above.
That
is
the
scope
actually
I
guess
no.
A
They
only
have
like
the
only
time
that
you
actually
directly
go
and
annotate.
The
d
like
real
things,
is
by
saying
that
this
SSA
value
holds
this
variable
or
fragment,
of
variable
or
something
and
or
memory
regions.
And
then
all
the
other
intrinsics
are
kind
of
away
from
that
and
are
like
referencing
other
kind
of
meta,
debug
info
things
and
aren't
referencing.
The
actual
LVM
ir.
A
And
I
think
it
makes
sense
to
kind
of
copy
their
like
data
representation
for
lack
of
a
better
word
of
like
having
like
the
same
kind
of
like
this
meta
world,
where
there's
lots
of
debug
stuff,
referencing,
other
debug
stuff
and
then
just
like
at
the
actual
interaction
with
our
atom
instructions,
have
just
the
things
saying:
oh
this
variable
that
has
an
associated
scope
and
all
this
stuff
or
an
Associated
subroutine.
Is
you
know
this
variable
through
these
regions
of
the
of
the
wesam,
but
I?
A
Don't
think
we
want
to
use
actual
like
intrinsic
functions,
the
way
lbm
ir
does,
I
think.
Instead,
we
want
to
kind
of
have
a
side
table,
because
we
don't.
We
can't
expect
other
people
to
annotate
there
like
wasm
right,
like
we
just
we're
reading
in
the
the
dwarf
for
an
existing
gasm.
If
it
exists
right,
is
this
making
sense
or
am
I
just
like
rambling?
No.
E
That
doesn't
make
sense
to
me
in
terms
of
like
that's
all
I
would
expect,
as
we
have
like
a
ton
of
side
tables
and
like
every
instruction,
basically
points
to
some
entry
in
that
side
table
or
what
I'm
also
curious
in
terms
of
like
what
are
dwarves
to
me
is
a
gargantuan
black
box.
That
just
does
everything,
and
it's
always
big
advice,
large
and
so,
like
the
general
purpose.
Is
that
I
know?
One
really
important
aspect
is
found
online
information.
So,
where,
from
what
file?
And
what
line
is
this
thing.
D
E
A
So
I
think
actually,
the
the
source,
location,
information
like
file
line,
column,
stuff,
I,
think
this
is
the
easiest
part,
the
so
the
dwarfs
actual
representation
is
like
like
like
in
the
abstract.
What
you're
representing
is
a
giant
table
as
you're
saying
of
like
where
each
row
in
the
table
is
one
instruction
in
your
program
and
then
there's
a
column
for
file,
there's
a
column
for
line.
That's
calling
for
column
and
and
I.
A
Think
there's
like
one
other
thing
too
and
then
like
that's,
obviously
a
huge
amount
of
data
right
and
then
most
of
it
is
not
changing
between
each
each
row.
Right
because
you'll
have
like
a
whole
you'll
have
like
five
instructions
that
represent
a
single
expression
in
this
first
language,
and
so
all
those
five
things
you
can
do
the
same,
and
so
what
they
do
is
instead
of
actually
including
this
giant
table.
They
have
a
one
of
their
many
custom,
byte
code
languages.
A
A
It's
like
finish
a
row
or
something
right
so,
like
you
like
do
like
finish
row
and
then
that
like
kind
of
omits
one
row
in
the
abstract
table,
and
then
you
like
do
like
add
ten
to
the
instruction
pointer
like
add
one
to
the
line,
add
whatever
to
the
column
and
they
do
like
emit
again
and
that
fills
in
like
the
difference
from
the
last
row,
all
the
way
to
where
the
new
instruction
pointer
is
like
kind
of
fills.
All
of
those
rows
in
two.
E
A
Again,
Lee
has
a
it's
not
actually
an
iterator,
because
it's
it
does
like
borrows.
So
it's
like
a
streaming
iterator
thing,
so
the
API
is
like
not
super
Pleasant
but
like
it'll
interpret
it
for
you
and,
like
you
know,
we've
spent
some
time
making
it
relatively
fast
released
as
fast
as
a
dumb
you
know,
switch
interpreter
can
be
so.
E
D
A
E
A
So
Jim
Lee's
writer
API
is
actually
pretty
nice
and
sort
of
similar
to
how
loris
is
structured
and
that
it's
like
all
kind
of
arena
based
and
then
it
figures
out,
based
on,
like
the
records,
you've
kind
of
constructed,
how
to
actually
encode
it
and
like
yeah
and
and
assemble
everything
right.
You're,
not
referencing
things
by
like
offset
in
the
dwarf
section
like
it'll
fill
that
stuff
in
for
you,
like
it'll
map
like
IDs
to
other
dwarf
entities
into
like
the
proper
thing
is.
E
A
E
A
Does
have
offsets
into
the
Wesen
code
yeah,
so
I
think
the
the
hard
part
here
is
going
to
be
choosing
the
right
representation
for
like
for
ranges
of
of
code
and
and
like
how
to
kind
of
like
connect
those
two
things
like
connect,
this
kind
of
like
meta.
You
know
tree
of
information,
that's
referencing,
each
other
with
like
the
concrete,
like
here's
the
range
in
the
lesson.
Instructions.
A
But
yeah,
that's
that's,
basically
it
and
then,
as
far
as
like
the
user
facing
code.
So
there's
there's
like
some
principles
that
we
should
probably
just
like
steal
from
LVM
for
their
design
of
debug
info,
and
so
that's
like.
If
a
optimization
pass
or
rewriting
passed
or
something
doesn't
move,
some
instructions
then
like
their
debug
info,
should
be
exactly
as
good
as
it
was
when
it
came
in
to
that
pass.
A
Right,
like
it
shouldn't,
ruin
anything
and
like
like,
and
so
it's
like,
even
if,
like
some
things
got
moved
but
like
this
part,
doesn't
that
part
has
divine
foot
preserve
and
the
other
is
like.
If
you
want
to
write
a
pass
that
moves
things
you
shouldn't
have
to
think
about
debug
info
right
like
if,
if
everyone
writing
like
a
GC
pass
or
like
a
you
know,
lowering
64-bit
values
to
32-bit
or
64-bit
operations
to
32-bit
has
to
like
worry
extensively
about
maintaining
to
bug
info.
A
Then
we've
designed
the
api's
wrong
like
if
they
put
in
zero
effort,
then
they
should
get
ok,
debug
info
that
breaks
down
based
on
what
they
moved
right
and
like.
We
should
automatically
add,
like
essentially
like
cancelling
out
to
bug
info.
When
you
move
something-
and
you
don't
tell
us
how
to
fix
the
debug
info
and
then
like
the
second
layer
is
like
like
we
need
to
say
like
it
should
be
easy
to
say
like
given
that
I
want
to
care
about
to
bug
info.
A
A
A
While
doing
all
these
passes,
I
think
it
makes
sense
to
try
at
least
initially
to
just
use
gimli
x'
writer
api,
so
just
like
decode
everything
into
the
writer
api,
and
then
that
gives
you
IDs
and
everything
and
like
it,
has
all
the
entities
in
arenas
and
stuff
like
that,
and
then
we
can
try
and
just
use
that,
and
so
you
could,
you
could
say
like
when
I'm
replacing
or
adding
new
instructions.
Here's
like
the
ID
of
like
relevant
to
bug
info
or
something
yeah.
E
That's
kind
of
what
I
want
to
do
is,
and
I
want
to
add
a
lot
of
abstractions
about
debug
info
and
walrus.
I
can't
want
to
say
just
use,
Ginley
or
use
something
on
top
of
cumulative
ten
rekindling
or
something
like
that.
So
I
figured
we've.
We
probably
don't
want.
We
don't
necessarily
need
to
support
manufacturing,
debug
info,
so
so
much
cuz
like
the
only
passes
we
have
are
like
injecting
brand-new
functions,
rewriting
calls
or
replacing
instructions
with
other
instructions
and
deleting
functions.
So
all
the
agree
like
we
should
be
preserving
debugging
for
everywhere
else.
E
A
E
E
D
E
Didn't
probably
just
go
from
there
like
see,
do
we
critically
need
in
lining
and
what's
the
minimum
thing
you
need
to
get
in
lining
information,
then
what's
the
minimum
thing
needed
to
get
actual
like
variable
information?
This
is
still
like.
We
don't
have
consumers
for
this
information
yet
like,
even
though
all
the
wedgie
based
debuggers
are
gonna,
be
based
off
of
a
sometime
or
like
something
quasi.
E
E
E
A
A
E
B
B
D
A
B
Yeah
I
think
it
sounds
like
we
have
a
good
idea,
but
I
do
know
like
with
the
huazhi
debuggers
coming
out
and
like
the
browsers,
are
kind
of
on
hold
with
like
their
debugging
stuff.
We
should
just
make
sure
that
the
debugging
stuff
that
we're
doing
isn't
gonna
like
be
a
problem
later
with
the
browsers
I,
don't
think
it
will
be,
but
we
should
just
like
keep
in
mind
that
now
we're
gonna
have
several
sets
of
debuggers
that
might
want
to
use
this
okay,
so.
A
So
those
the
yzd
buggers
are
using
dwarf.
So
that's
good
like
the
same
thing.
We're
gonna
use
and
my
impression
from
kind
of
like
talking
to
folks
is
like
like
with
with
like
y'see.
We
can
just
like
make
a
debugger
and
it
doesn't
really
involve
like
standards
or
anything.
But
then
what
we
can
do
is,
we
can
say,
like
hey,
look
like
like.
Basically
the
duplicating
subgroup.
I,
don't
know
if
you
were
in
the
one
meeting
that
we
ever
had
no.
A
A
total
circus
of
just
like
I,
mean
like
it
was
like
30
people
who
all
like
had
their
pet
thing,
that
they
wanted
to
do
and
like
no
like
coherent
vision
together
or
anything
right
and
so
I
think,
like
what
will
be.
Nice
is
like
when
all
of
these,
like
quasi
debugging
demo,
start
coming
out
like
here's
like
a
concrete
thing
that
we
can
use
to
like
help
like
articulate
a
vision
that
everyone
can
like
start
moving
in
the
same
direction
with
right.
B
A
B
Totally
well
just
I
feel,
like
I've
said
this
multiple
times
now
so
I
think
I
must
be
on
the
hook
for
reading
at
least
one
of
them.
But
we
should
write
a
tracking
issue
for
what
you
just
said
and
then
probably
link
to
some
of
these
debuggers
as
being
like
potential
targets
that
we
anticipate
and
so
that
we
can
keep
an
eye
on
their
development.
E
D
E
Yeah
be
great
to
have,
but
no
one's
really
pursuing
it,
except
for,
like
the
I
think
URI
and
a
couple
of
Mozilla
folks
are
like
looking
to
have
any
debugging.
What
and
whatsoever
is
sort
of
like
in
ad-hoc,
fashions
I'm,
basically
currently
pushing
forward
tour
so
like
getting
dwarf
in
wesam
files
and
then
having
that
hooked
up
to
some
debugger
somewhere,
which
works
great.
We're
like
the
initial
demo
is
very
relatively
easy
to
hit
like
LD,
be
working
on
baizen
time
but
Nick.
Are
you
aware
of
any
other
efforts
to
get
debugging
unwisdom
efforts?
No.
A
E
D
E
Uri
does
have
apparently
you're
he's
always
been
a
ton
of
patches
but
she's
kind
of
some
set
of
patches
which
integrates
Kirwan.
If
you
use
crane
lifts
and
SpiderMonkey
and
generate
wisdom
that
will
also
generate
dwarf
and
you
can
attach
LDB
of
firefox
directly
and
you
can
debug
thrillogy
be
wise,
encoded
in
fire
running
in
firefox.
Now,
obviously,
that's
not
how
it's
going
to
be
with
like
actual
web
debugging
because
know
what
you
can't
rely
on
attaching
LDV
to
firefox
in
general,
but.
E
But
there
is
a
there
is
one
proof
of
concept
of
a
video
actually
debugging
code
running
on
the
web
in
firefox,
using
a
debugger,
and
so
it's
got
a
long
way
to
go
from
there
to
actually
being
shipped,
but
so
the
I
guess
this
is
this?
Isn't
really
so
much
for
you
Nick,
as
opposed
to
just
like
if
video
and
everyone
else
in
general,
in
terms
of
like
how
far
out
this
still
kind
of
is
so
I
think
we're
how.
E
E
B
E
Just
by
having
said
all
of
what
I
just
said,
it
is
still
really
valuable
for
us
to
stay
ahead
of
the
curve
in
terms
of
we.
If
anyone
wants
to
like
where
we
have
very
strong
signs,
that
dwarf
is
the
way
to
go
and
wise.
The
Bunge
in
is
destroying
to
earth
like
the
actual
information
at
the
tactical
level,
and
so
as
long
as
you
preserve
that
and
give
people
something
to
test.
That
would
be
a
great
thing
to
trust,
to
always
be
ahead
of
the
curve
there
right.
E
That's
where
I
feel
like
if
we
get
file,
name
and
I,
never
information
and
kind
of
like
our
own
synthetic
tests,
all
working,
that's
probably
good
enough
for
now
in
terms
of
and
like
we
can
see
well
it's
up
to
like
the
general
debugging
community
and
browser
vendors
and
whether
one
types
at
large
to
figure
out
to
tell
us
like.
What's
the
next
step
of
level
of
deep
level
of
Dorf
that
we
need
to
be
producing.
B
E
Is
one
where
we
I
don't
think
Nick
that
the
dwarf
extensions
for
wisdom
are
official
in
any
way
right.
A
E
E
This
is
where
what
we
want
to
do
is
show
some
proof
of
concepts
with
the
door
for
working
out
and
like
LF.
We
do
have
some
patches
in
the
rest,
compiler
for
generating
wasn't
worth
information,
and
the
idea
is
to
like
prove
this
out,
show
it
working
on
y'see.
It
wasn't
time
and
everything,
and
then
we
can
kind
of
work
more
towards
community
consensus
around
your
phone
that
direction
we're
going.
A
Yeah
and
if
you
want
to
read
the
LLVM
debug
info
user
guide,
thing
I
think
like
the
first,
like
three
pages
are
really
useful
and
I
can
see
if
I
can
like
them
like
just
like
the
part
that
kind
of
like
describes
the
philosophy
of
like
guarantees
of
like
not
getting
in
the
way
and
like
not
breaking
the
bug
info.
If
you
don't
like
mess
with
something,
it's
just
nice
to
have
that
kind
of
perspective.
A
B
A
A
B
B
That
was
kind
of
why
I
was
interested
in
this
topic,
so
I
was
like
I
need
to
know
enough
to
be
able
to
explain
it
to
someone
else
to
get
them
to
do
it.
You.