►
From YouTube: sig testing commons
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
A
Yep
all
right
so
I
didn't
have
much
on
the
agenda
this
week
and
I
did
ask
George
to
do
a
triage
room
to
see
where
we're
at,
because
I
know
there
are
a
lot
of
PRS
there
like
I've
been
busy
and
other
people
have
been
busy.
Most
of
the
folks
working
in
this
effort
are
part-timers,
so
I
think
George
probably
wrote
this
initial
one
here.
Directors.
C
Yes,
so
if
you
look
in
the
triage
PRS,
that
is
actually
the
full
list
of
all
that
we
are
labeled
with
the
area
he
tweet
this
framework,
they
at
least
I,
add
in
the
dogs
are
the
ones
that
they
actually
need.
Some
attention,
some
of
the
other
PR
say
they
are
either
working
progress
or
have
something
else
from
that
we
can
just
hold
off
for
later,
but
if
you
only
don't
want
to
go
through
the
entire
list,
only
13
peers,
okay,.
A
A
C
That
one
is
actually
that
one
is
actually
mine,
so
this
is
I
want
to
propagate
I.
Can
I
want
to
pause
this,
so
there
was
an
issue
the
Android
created
to
start
off
a
to
start
cleaning.
Okay,
you
till
I
owe
you
finally
in
the
framework-
and
this
is
really
a
really
straightforward
clean
up-
there
are
a
bunch
of
scape.
A
escape
is
a
escape
related
functions,
so
I
just
moved
into
another
file
within
the
framework
I.
A
Don't
have
strong
opinions
here.
The
whole
purpose
of
the
federated
portion
is
the
primary
function
is
to
remove
the
depth
graph
of
the
transitive
closure
of
things
that
would
be
included
in
utils.
I
kind
of
think
skip
knows
is
my
opinion.
I
kind
of
think
skip
is
fundamental
to
are
pretty
close
to
the
chest
of
the
framer
itself.
I.
D
A
A
What's
the
size
of
this
file
200
lines,
I'm,
okay,
like
I'm
I'm,
super
old-school
when
it
comes
to
cone
structure,
anything
that
goes
beyond
like
300
lines
of
code,
I
usually
think
determined
that
it
has
to
be
its
own
file.
That's
pretty
pedantic,
but
it's
also
old-school
I.
Think
it
just
it's
easier
to
find
the
details
directly
on
initial
inspection,
so
I'm
totally.
Okay.
With
this,
the
way
it
is
okay,.
C
Awesome
and
the
other,
the
other
follow-up
to
this
is
to
just
keep
doing
this
type
of
refactoring,
not
exactly
the
same,
but
still
in
the
U
tilde
cofell's.
There's
a
ton
of
really
random
functions
that
shall
be
including
the
other
sub
packages.
Is
it
just
gonna?
Keep
me
a
quarry
just
going
to
keep
going
at
it
until
it's
being
as
clean
as
possible.
Yeah.
B
D
Even
for
functions
that
are
called
in
to
places
like
I'd
argue
like
small
duplication
is
better
than
like
kind
of
bloating
the
framework
and
then
supporting
it.
So,
like
I
feel
like.
Maybe
we
should
have
some
guiding
principle
that
says
like
if
it's
called
in
at
least
three
places,
and
we
can
have
it
in
that
framework.
Otherwise,
just
do
a
small
duplication
and
move
on.
C
C
This
one
actually
just
checking
up
it.
There
was
a
second
or
dependency
issue.
So
take
a
longer
conversation.
Last
week's
conversations
with
amici
and
one
of
the
issues
of
was
open
wall.
Well,
we
will
not
have
an
issue
system.
We
are
going
to
start
moving
all
the
tests
from
it.
We
log
to
the
intellectual
framework
and
I
guess
away
I,
guess
this
one
they
had
a
circular
and
then
not
able
to
import
the
framework.
A
A
C
D
C
A
A
It's
been
a
thorn
in
the
side
for
a
long
time.
Kiko
update,
plus
tech
fix
here
here.
Oh
this
is
the
this
is
Patrick's.
One
I
put
a
comment
on
it
that
we
updated
ginko,
but
he's
not
updating
Omega,
so
I
could
get
updated,
but
it's
unrelated
to
the
problem.
The
Pierce
on
the
processor
loosely
related,
so
Giga
one
time
should
be
valid
combination.
Well,
not
in
the
ven
during
I.
Look
is
even
on
this
call.
Not
the
keys.
I
looked
in
the
vendor
stack
for
ginkgo
and
Kinkos.
A
F
A
A
C
But
that's
actually
a
thing
that
intro
was
pondering.
I
was
creating
a
couple,
a
couple
issues
that
were
explicitly
labeled
as
it
we
framework
it'll,
look
more
like
Joe's
general
test
fixes
in
the
all
part
of
the
conformance
group
with
people
still
aka
kept.
Taking
me
great
and
I'm
back
into
the
ìiî
ìiî
framework,
so
I'm
just
wanting
to
explicitly
mention
that
and
ask
if
that
is
the
way
to
way
yeah.
A
It's
not,
you
can
just
tell
right
away
like
II,
just
unlike
cursory
inspection,
you
can
tell
this
is
related
to
a
specific
set
of
tests,
and
you
can
see
that
you're
only
calling
out
the
test
files
themselves
right
so
that
they're,
not
even
in
the
framework
directory
at
all.
So
this
one
is
a
service
jig.
D
We
should
get
rid
of
service
jig,
because
I
think
it
was
just
added
way
back.
I
was
like
I,
don't
know
what
I
don't
know
what
it
does
like
exactly,
but
it's
like
a
sub
framework
for
service
tests,
which
I
think
like
the
core
framework.
There's
there's
nothing
in
the
core
framework
that
can't
do
what
it
does.
I
think
we
should
just
get
rid
of
it.
So.
A
C
F
A
F
A
C
B
A
C
A
Yeah,
some
of
these
things
are
really
hard
to
solve.
This
one,
in
particular,
is
super
thorny.
It's
because
we
don't
actually
do
component
config
yet
and
there's
there's
a
bunch
of
issues
that
are
related
to
this
they've
actually
been
pushed
up.
I've
been
fighting
Googlers
of
all
people,
which
I
thought
was
interesting
because
they're
they're
not
resourcing
any
efforts
with
regards
to
component
and
thing
they
started
down
this
path
a
long
time
ago,
but
there's
no
one
working
on
a
liked.
A
E
A
Usually
have
their
own
tests
set
up
for
them.
The
question
that
they've
tried
to
do
is
what
about
config
Flags,
because
there's
there's
feature
gates.
Those
are
pretty
broadly
covered
by
feature
:
whatever,
but
what
about
configuration
Flags,
where
we
we
test
the
defaults
that
we
want
to
be
able
to
test
these
other
incantations
and
the
answer:
is
we
don't?
Unless
you
have
automatic
detection,
you
can't
really
broadly
enable
this
and
the
automatic
detection.
A
The
the
grand
unified
field
theory
for
this
is
to
use
component
configuration,
which
is
to
basically
store
and
well-defined
types
very
similar
to
compute
machinery,
the
the
configuration
for
components
on
the
cluster
as
a
config
map,
in
version
those
appropriately.
The
problem
is
only
two
of
the
components:
have
well-defined
configuration
file
formats
so
far
and
it's
been
really
slow
to
move
the
community.
I
really
do
think
we
need
to
resource
this
effort.
I
wish
people
would
work
on
and
take
it
seriously,
but
I
don't
know
how
to
other
than
be
covering
about
it.
A
So
maybe
I
can
start
to
draft
a
blog
post
about
like
where,
where
things
need
help
in
the
community,
I
always
get
these.
Here's
a
broader
question
to
to
the
group.
I
always
get
these
statements.
That
kubernetes
is
hard
to
contribute
to
and
part
of
that's
totally
legit
and
part
of
it's
I
think
a
little
bit
self-serving
like
it.
If
you
fed
it,
if
you
ferret
around
enough,
you
can
probably
figure
out
your
way.
A
B
F
A
A
So
I
think,
like
generic
coverage
tools,
could
probably
simplify
the
number
of
this
stuff.
I
believe
in
run
like
code
of
coverage
and
the
framework
to
see
what
actually
Ian
I
guarantee
you,
there
has
to
be
code.
That's
dead
code
instead
of
the
framework,
if
not
signatures,
then
actual
code
paths
that
are
exercised
I
think.
A
Think
this
statement
needs
to
be
refined,
so
it's
actionable
like
that.
I
understand
the
broader
thing,
but
how
do
you
actually
interpret
this
in
a
concrete
action
might
be
beneficial,
but
it's
still
at
the
same
time
like
this
is
where
there
are
other
efforts
that
are
required
before
we
can
actually
get
this
thing
done,
I
think
it's
good
to
have
this
list
and
take
a
look
at
it.
Reevaluate
it
from
time
to
time,
but
this
one
in
particular
I
think
requires
a
little
bit
of
refinement.
F
C
The
only
thing
that
kind
of
comes
to
mind-
I,
don't
know
how
broadly
usefully
cookies
and
they're-
and
we
also
talked
about
this
previously-
it
was
about
if
we
factoring
the
test
context
for
the
framework
and
I
from
day
from
the
code
that
I
seen
within
the
framework.
We
also
get
some
useful
information
just
actually
make
a
making.
It
is
here
to
actually
provide
some
information
for
users.
C
A
If
we
had
well-defined
component
config,
you
could
have
the
types
for
that
actually
outline
the
details
for
it
so
I
think
stage
one
would
be
breaking
apart.
The
larger
test,
config
stage
two
would
be
like
switching
the
test
config
to
be
well-defined,
well-defined
component,
config
style
types,
and
then,
if
they
are
well-defined
types,
we
can
basically
put
in
the
entire
godet
or
the
go
doc
block
that
would
outline
what
how
do
you
use
it?
A
The
details
there,
one
thing
that's
frustrating
is
all
hell
to
me-
is:
if
you
look
at
the
go
doc
documentation
for
a
lot
of
our
types,
you'll
find
that
they're
very
lacking.
With
regards
to
the
details
of
the
implications
of
switching
the
knobs
you
have
just
you
have
to
federate.
You
have
to
fare
through
the
code
to
understand.
H
A
Believe
he
actually
has
I
just
added
to
you
on
slack
I
believe
he
actually
has
documentation
on
getting
started
in
a
new
user
orientation.
So
he's
got
both
so
that
actually
I
would
go
through
the
orientation
stuff
and
then
I'll
take
a
look
at
the
other
details
that
are
there
and
he
can
probably
help
get
you
up
to
speed
pretty
fast.
But
that
is
that
is
the
overview.
A
F
A
F
You
for
the
reminder,
I
do
need
I
failed
to
create
the
second
issue,
for
that
I'll.
Do
that
and
then
to
clarify
for
other
people.
The
second
issue
is
what,
if
somebody
has
every
node
tainted,
and
they
want
to
make
the
test
workloads
actually
tolerate
that
taint.
So
the
first
problem
was
just
you
want
the
test
framework
to
start,
even
though
your
nodes
may
not
appear
to
be
ready
because
they're
tainted,
but
the
second
thing
is:
what
can
you
actually
make
the
test
mutate
every
test
workload?
F
A
F
And
we
can
discuss
on
the
issue
too,
when
I
make
it,
but
we
do
think
it's
gonna
be
similarly
simple.
Just
like
a
straight
list
and
all
the
test
workloads
will
be
mutated
to
tolerate
those
right,
there's
no
extra
knobs
that
they
need,
because
I
can
imagine
someone
wanting
like
awkwardly
fine-tuned
control
like
oh
well,
I
have
these
three
different
types
of
taints
and
I
want
to
sometimes
use
one
and
sometimes
use
another.
If
this
feature
is
used
or.
H
A
A
Changing
adding
a
whitelist
is
to
enable
users
who
default
they
setup
notates
as
part
of
their
setup.
This
is
super
common
for
people
who
use
things
like
cluster
API.
There's
a
ability
to
take
your
nodes
on
entry
and
people
use
this
all
the
time
with
provisioning
tools
because
they
want
to
have
a
default
policy
of
whatever
they're
doing
so,
the
it's
really
common,
but
it
wouldn't
affect
the
default
sneezer
scenario.
H
Yeah
yeah,
basically
I'm
kind
of
stuck
on
multiple
requests
because
of
this
issue.
I've
had
zero
progress
because
of
this,
but
basically
there
should
be.
We
should
have
some
staging
repo
for
the
test
images
that
we
have,
so
the
image
promotion
process
automatically
promote
those
images
to
the
official
repo.
E
H
One
one
seven
zero
one,
because
apparently
we
don't
manually
build
and
published
images
anymore,
so
we
basically
need
the
staging
repo.
We
need
the
at
least
two
extra
quick
quests,
one
on
queries,
kubernetes
one
and
another
one
in
the
test
here
for
our
repo,
which
defines
the
job
and
I
have
actually
added
a
couple
of
questions
and
concerns
to
the
papers
I've
mentioned,
because
we
actually
have
about
30
images
in
the
community
test
images
folder,
and
if
we
modify
one
of
the
images
with
that,
actually
pub
building,
publish
old
images.
H
A
A
So
I
think
the
there
is
a
Tuesday
meeting
were
Kate's
and
for
a
group
talks,
all
the
time.
I
think
that
is
a
good
place
to
go
to
to
get
your
answer
to
get
that
addressed.
So
there
I
don't
know
what
the
time
is
to
change
the
schedule,
so
no
longer
every
time
I
calendar.
But
if
you
look
at
the
community
repos,
you
look
like
a
club.
You
know.