►
From YouTube: Quilkin Monthly Sync - May 2022
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
Not
throwing
stuff
down
yeah
we've
mostly
been
working
on,
adding
in
xds
to
qualcomm
itself,
we're
we've
been
trying
to
deploy
it
in
our
infrastructure
and
like
been
fixing
all
the
bugs
that
have
cropped
up
we're
nearly
there
with
it.
We
have
we're
just
adding
in
the
reflector
change
for
the
kubernetes
stuff.
I
believe
you
mentioned
that
earlier.
B
Once
that
is
fixed,
we're
going
to
try
to
deploy-
and
hopefully
hopefully,
there's
no
bugs
this
time
it
all
works
and
then,
like
all
the
xds
stuff
works
right
now,
there
just
seems
to
be
like
listing
that
service
just
regularly
seems
to
not
always
return
all
of
the
results,
so
we're
thinking-
maybe
it's
just
because
we
haven't
set
the
watch
so
we're
going
to
see
if
that
fixes
it
otherwise
investigate
whatever
the
other
bug
that's
causing.
That
is
yeah.
B
Also
I
meant
to
post
in
the
pr,
but
I'll
be
also
once
that's
done,
I'm
going
to
split
out
some
of
it
there's
a
another
partial
bit
that
can
be
taken
out
and
like
reviewed
on
its
own,
but
then
the
rest
of
it.
It
needs
to
all
be
reviewed
as
one
piece
because,
like
there's,
the
main
sort
of
changes
come
along
with
it
is
like
reducing
the
configuration
change
overhead.
B
B
You
know
currently
in
quilting,
we
have
a
lot
of
different
structures
for
like
updating,
updating
like
a
different
component
of
course,
kind
of
managing.
All
of
that.
This
pr
changes
that
so
that
essentially
the
config
struct
is
essentially
the
one
single
place
where
all
of
that
state
is
both
stored
and
like
where
you
can
get
like
a
snapshot
of
it
as
well.
So.
C
B
Not
they're
no
longer
like
two
different
things
that
happen
in
the
code
base.
It's
like.
There's
you
just
access,
config
and
like
you,
can
also
just
deserialize
and
serialize
your
own
conflicts.
B
And
that's
what
and
the
reason
why
I
can't
split
up
more
is
because,
like
yeah,
the
new
xds,
server
and
client
that
now
use
that
yeah,
I
would
have
to
like
pull
one
out
and
then
write
in
support
for
the
other
component
in
order
to
be
able
to
merge
them
separately
and
that's
just
like.
I
don't
think
that's
worth.
It
got.
A
B
Yeah,
like
I
think,
it's
a
really
nice
change
it
like
makes
a
lot
more
convenient.
Actually
like
it's.
The
main
reason
we
actually
lose
code
rather
than
actually
add
code
with
this
feature
is
just
a
reduction
in
like
having
to
centering
through
channels
and
a
whole
bunch
of
like
cue
stuff
that
wasn't
isn't
necessary
anymore,
sweet
actually.
A
That
sounds
great.
I'm
very
excited
about
this.
A
Pretty
awesome
one
of
the
ones
we'll
talk
about
my
my
words
of
port
stuff,
but
once
I
get,
I
think
I
think
I
have
an
idea,
maybe
on
how
to
get
through
that.
But
I'll
start
writing
the
harness
for
the
end-to-end
tests.
So
we
can
do
some
end-to-end
tests
on
the
system.
I've
got
some.
I
don't
think.
A
A
A
Sometimes
like,
if
I
have
like
a
socket
that
I'm
using
for
a
test
like
to
send
a
packet
and
then
see
if
one
comes
back
or
whatever,
when
I
send
the
packet,
nothing
happens
and
if
I,
if
I
retry,
I'm
like,
oh,
maybe
like
cooking,
hasn't
started
yet
I'll
retry
nope,
I
can
let
it
sit
there
for
a
minute.
Do
nothing.
But
if
I
have
another
system
like
I
like
fire,
netcat
or
something
and
then
fire
the
packet
into
the
coconut
on
the
port
that
I
was
running
on,
it's
received.
A
The
only
thing
I've
seen
that
seems
to
work
which
I
have
in
the
unit
tests
is,
I
like
recreate
a
whole
entire
socket
again
and
then
send
it
from
that
and
like
if
I
have
that
there
in
a
retry,
it
works,
even
though
it
never
actually
retries
there.
I'm
extremely
confused.
B
Yeah
I'd
have
to
look
into
it
more
but
like
yeah.
The
only
time
I
ran
into
something
kind
of
similar
right
in
the
exact
stuff
was,
like
I
accidentally
said,
like.
B
I
missed
one
of
the
ports
for
the
clients,
because,
like
one
of
the
hard
difficult
things
very
difficult,
just
slightly
annoying,
it's
like
nothing
is
like
a
good
default
stage
for
qualcomm
like
doing
nothing
and
like
receiving
nothing,
because
udp
is
like
stateless,
so
it
can
be
very
difficult
like
if
you're
just
like
have
one
thing
misconfigured
like
everything,
just
sort
of
stops
and
it's
waiting
and
it's
like
that's
a
good.
A
A
A
It
gets
bound
early
like
by
the
time,
you've
gone
through
and
cooking,
like
it's
bound
on
that
part,
whereas
I
think
with
the
way
this
is
now
it's
not
necessarily
receiving
straight
away,
but
that
doesn't
make
sense
why
a
retry
doesn't
work.
The
only
other
thought
I
had
was
basically
setting
up
a
little
toolkit
that
was
like
just
make
sure
that,
like
packets
can
go
from
one
point
to
another.
C
A
Too,
with
like
some
of
the
test
stuff
like
like
a
little
tool
kit
in
there,
for
so,
you
don't
have
to
like
try
and
find
a
unique
port
to
run
an
integration
test
on
like
it.
Just
you
know,.
B
I
think
the
I
don't
know
how
you
do
it.
Mine
is
very
just
like
it
binds
a
random
portal.
A
random
address
gives
you
the
port
and
then
drops
that
socket
so
that
it
knows
okay,.
A
B
B
B
Just
yeah
you
just
even
just
bind
it,
and
then
you
just
use
the
default
like
ipv4
unspecified
and
zero
yeah.
B
A
The
socket
yeah
and
I'm
guessing
if
it
fails,
then
yeah.
A
A
A
B
Other
thing
that
was
implemented
in
the
xs
that
I
mentioned
was
just
a
configuration
change
from
a
flat
list
of
end
points
to
just
actually
specifying
the
cluster
configuration
in
the
thing,
because
we
were
just
doing
a
bunch
of
conversion
from
like
okay.
Well,
we
have
a
flat
list
of
endpoints,
then
we
convert
it
to
a
cluster
and
then
it's
okay.
We
receive
this
cluster
and
then
we
convert
it
to
a
flat
list
of
endpoints.
A
B
We
were
doing
as
far
as
like
doing
trying
to
deploy
qualcomm
or
we
were
like.
I
was
also
trying
to
do
more
like
dashboard
stuff,
and
it's
like
I
wonder
if
like
is
there
a
way,
we
can
package
like
a
good
dashboard
into
quilting
itself
and
like
collaborate
on
the
dashboards
because
currently
they're
all
like
internal
to
us,
but
I
was
wondering.
B
C
B
A
Is
this
is
what
we
have
now
we're
just
like
a
yama
file,
because
I
literally
just
copied
how
we
did
it
in
a
gunners.
We've
run
this
in
a
gun.
It's
over
multiple
grafina
versions
like
just
like,
like
a
you,
can't
read
that
because
that's
tiny,
where
it's
just
stored
in
like
the
json
file
and
we
just
have
multiple
yellow
files
and
stuff.
I
mean.
D
A
C
B
A
D
Simon
said,
we
can
just
build
the
json
that
one
and
the
json
output
of
that
in
in
the
thing
right
and
that'll
work.
Fine
and
that's
pretty
backwards
compatible,
because
graphing
actually
goes
out
of
its
way
to
be.
D
What
is
that?
What
are
you
using
a
thing
called
graphonnet?
It's
a
json
generator
for
fun
for,
for
that
json
that
you
just
trained
us
graph.
A
C
A
C
It
just
like
libsynet
functions
that
you
call
so
yeah
yeah.
C
A
D
There
are
solutions
to
that
right,
like
you
could
you
could
run
json
for
each
release
and
attach
them
to
the
release
or
a
bunch
of
other
things?
Definitely
yeah.
I
don't
know
I
mean
like
a
bunch
of
different
ways.
That's
cool
yeah,
like
I
I'm
fine
for
now,
just
just
pasting,
the
graffiti
that,
like
we
have
json
that
and
graphene
internally
right
now,
we
can
just
run
the
thing
and
get
the
json
output
instead
and
paste
that
okay,
that's
fine
initially,
and
we
can
worry
about
this
later.
B
We
could
probably
actually
like,
if
you
want
like
think
about
like
productizer
or
something
like
if
we
could
just
put
it
in
the
quilting
binary
or
something
so
that
you
could
run
like
quick
and
install
grafana
dashboards.
And
then
it
would
just
oh.
D
B
D
A
B
Yeah,
I
wanted
to
add
more
dashboards,
but
I
was
like
I
don't
know
where
to
put
these
and
like.
B
A
That's
what
I
do
too,
if
you
want
to
add
like
another
page
well,
so
if
you
basically
my
workflow
is
I
grab
the
yeah,
I
do
the
edits
and
then
I
export
the
json,
and
I
just
replace
it
in
here
for
that
particular
dash
and
then
just
have
a
separate
yaml
file
for
each
dash
like
actual
board,
and
you
can
just
copy
the
pattern.
A
B
I
think
the
only
thing
I've
is
like
I
kind
of
agree
with
what
you
said
where
it's
like.
I
need
to
look
into
this
more.
It's
like
yeah.
I
think
I
want
to
go,
read
like
how
rtp
works
or
whatever
and
like.
A
B
A
B
A
Like
that
makes
sense,
yeah.
B
Like
overflow,
or
something
like,
oh
well,
actually,
this
kind
of.
Actually
you
know
that
you
mentioned
this-
that
does
actually
bring
up
another
issue.
I
I
wrote
a
while
back,
because
when
I
was
doing
this
xcs
stuff
another
one
of
these
idiosyncrasies
of,
like
you
know,
endpoints
to
close
their
endpoints
was
that
in
the
xds,
how
that
works
is
like.
They
have
a
listener
discovery
service
for
discovering
listeners
and
like
in
invoice.
D
A
B
B
Yeah,
that's
always
yeah,
just
asking
it's
like
yeah.
Do
we
want
like
because
again
in
in
their
protocol,
so
we're
always
just
going
we're
always
going
to
get
like
a
multiple
list,
because
it's
going
to
be
sending
one
or
many
listeners,
so
it's
like.
Do
we
actually
want
to
fit
in
with
that
use
case
and
like
be
more
xds
compatible
or
do
we
just
want
to
continue
doing
our
own
thing
and
like
have
our
xds
be
like
okay?
Well,
you
can
cut.
A
B
Yeah
yeah,
the
thing
I
was
just
going
to
say
was
it
like:
it
comes
up
for
that
rtp
case
where,
like
you,
could
just
have
like
one
listener
per
one
session
per
listener,
and
then
you'd
have
multiple
listeners
in
a
single
process.
A
B
B
I
mean
with
your
changes
you
could
actually
just
have
you
have
multiple
listeners.
Some
say
it's
important.
Couldn't
you
like
with
the
changes.
A
A
B
Like
it
is
kind
of
like
it's
kind
of
how
we
do
it,
but
it's
kind
of
like
we
do
it
like
one
level
beneath
rather
than
like
having
multiple
listeners,
it's
like
we
have.
The
listener
has
multiple
like
subtasks
of
workers
that
are
just
yeah
since
you're
doing
the
same
job.
It's
all
defined
within
one
listener,.
A
Oh
one
thing
I
was
going
to
bring.
I
did.
I
did
a
preliminary
like
like
bottle
performance
test,
which
I
had
to
run
multiple
times,
because
sometimes
it
fails
because
of
race
conditions.
I've
had
didn't
see
a
huge
performance
increase,
it's
like
a
little
bit,
but
I
also
are
aware
of
the
fact
that
a
performance
tester
like
single
threaded
running
like
just.
B
It
looks
like
that
I
think
it
allows
me
to
remove
a
ton
of
code.
What's
it
called
yeah
thread
per
core
stuff,
like
I
think
that's,
where
I'd
be
interested
to
see
like
how
this
comparison
like
stuff,
like.
B
Runtimes
they're,
based
on
like
the
idea
of
like
threadbar,
core
and
stuff
for
this,
and
like
oh,
what's
that
new
io
interface,
not
that
new.
But
it's
the
async.
A
B
Like
I
believe
this
guy
came
from
was
I'd
be
interested
to
see
like
if
we
did
him
in
a
multi-throat
how
much
if
there
is
any
performance
gain
or
if
it
is
just
code,
reuse,
but.
A
D
B
Anyway,
it
makes
it
way
more
like
tcp,
where
you
can
just
cheaply,
clone
it
and
have.
A
No,
it
works.
It
works
that
way
too
so
yeah
something
I
can
get
it
all
working.
Probably
it
is,
I
think,
a
net
positive
anyway,
but
all
right
once
once
I
get
it
working
I'll
start
pulling
it
apart
too,
I
think
there's
some
there's
some
chunks
that
can
yoink
out
make
it
a
bit
easier
to
review
as
well.
It's
not
too
big
right
now,
but.
A
B
A
B
So
yeah,
but
like
on
top
of
like
new
tools
like
raphael
or
whatever,
something
a
tool
that
I've
been
using,
that
I've
just
been
reading
like
we're
talking
about
like
oh
doing
this
make
stuff
and
like
having
to
do
cross
for
all
this
cross,
compilation
and
stuff.
One
thing
that
tool
that
I
found
interesting
is
that
earthly
I
mentioned
I've
mentioned
it
before
to
like
the
residency.
One
is
like
it's
described.
B
Well,
I
I
got
it
from
c1
actually,
but
it's
described
as
like
a
make
file
plus
a
dockerfile
had
a
baby
which
probably
the
best
description
of
it
is
literally
just
like
make
commands,
but,
like
you,
each
command
is
like
a
docker
layer.
A
B
B
Yeah,
like
I
haven't,
mentioned
it
before,
because
other
previous
releases
are
bsl,
so
that's
not
allowed
on
google,
but
their
next
release
is
going
to
be
mpl,
so
that
will
actually
be
able
to
be
used
like
in
a
project
like
this.
Essentially,
this
is
like.
The
benefit
of
this
is
like
it
has.
B
C
Get
this
working
for
cooking
right.
B
We
didn't
get
working
for
the
deployment,
but
that
was
just
because
I
had
an
arm
machine,
so
I
built
an
arm
image.
Yeah.
A
B
Spend
I
just
didn't
want
to
spend.
A
C
I
mean
I've
been
looking
into
this
a
bit
and
it
seems
to
be
a
purely
across
issue.
There
is
this
upstream
issue,
I
don't
really
know.
We
only
see
it
on
wsl,
yeah
and
wsl.
We
already
in
a
weird
state
where
we,
where
we
run
the
docker
demon
in
windows
and
mount.
C
Wsl
and
then
you
run
docker
in
docker
in
wsl
right
basically
yeah
surprising
that
it
even
works
slightly.
No,
but
jokes
aside,
so
I
mean
I
actually
had
a
problem
building
on
windows
as
well,
but
that
seems
to
have
been
just
me
doing
doing
weird
stuff,
so
that
works
now,
so
we
have
a
way
forward
because
otherwise
yeah
other
than
that.
I
I'm
also
on
a
on
a
rm
mac.
So
I
can
build
on
that
and
definitely
sell,
doesn't
work
the
upstream
issue
in
cross
that
they
seem
to
only
mention
issues
with
nixos.
C
C
A
C
A
C
I
don't
like
I
don't
know,
I
don't
think
we
ever
had
it
working,
because
the
initial
builds
we
did
together
miner
and
I
think
we
just
did
with
cargo
instead
of
cross.
C
What
a
nightmare
so
yeah,
I'm
also
very
like
not
too
too
familiar
with
cross
compilation
for
so
I
I.
A
Don't
know
went
down
that
path
to
just
be
like
how
do
we
get
this
stuff
to
work,
especially
like
everywhere
else,
because
it's
super
easy.
I
did
find
a
circuit
down
here
there's
and
when
users
like
how
are
you
cross
compiling,
it
seems
like
that's
a
common
problem.
C
A
B
Have
a
big
tracking
issue
for
like
oh
well,
this
doesn't
work.
We
can't
like
build
this
entirely
natively
up,
because
this
one
tool
and
basil
doesn't
support
arm.
Yet.
C
C
Yeah
not
to
be
fair
at
this
point,
I
think
most
things
work
usually
at
least
if
you,
if
you
live
in
the
rm
world,
but
but
I'm
curious
like
do
you,
could
we
get
along
with
just
using
cargo
instead
of
cross
until
we
have
something
else,
because
right
now,
yeah.
B
B
B
A
B
Well,
I
think,
that's
handled
by
the
build
script.
I
think
the
open
sl
build
script
handles
that
for
you
as
long.
B
C
So
I
mean
as
long
as
that's
not
the
reason
why
it's
not
working
so
I
mean
we
can
get
a
build.
B
A
I
did
find
I
did
find
one
of
the
one
of
the
reasons
that
the
build
fails
on
right
now,
which
I'm
worried
about
on
the
the
port
reuse.
The
port
address
stuff
is
that
there
are
certain
socket
rpis
that
are
not
available
on
windows.
So
when
it's
trying
to
come.
B
That's
about
where
I
got,
but
I
think
I
was
kind
of
like
I'll
come
back
to
it
later.
A
C
A
A
Let's
work
on,
oh
actually,
I
do
have
sales
that
I
thought
of.
I
was
just
gonna,
not
super
important,
but
I'm
trying
to
remember
we,
we
output
like
a
debug
and
then
a
release
of
token.
A
B
B
B
B
Like
you
might
want,
for
example,
in
play
in
prcd
to
run
the
debug
build
and
then
in
like
push
cd
like
when
it's
a
release
then
run
the
release
bills,
because
that
would
reduce
ci
times
by
a
bit
and
then
for
the
actual
release.
We
would
only
release
the
release
version.
B
Okay,
okay,
see
you
next
month.