►
From YouTube: Kubernetes SIG CLI 20210901 - kustomize bug scrub
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
Good
morning,
good
evening,
good
afternoon,
depending
on
where
you
are
today
is
the
first
of
september.
This
is
another
of
our
monthly
sexy
lie,
customized
blocker
box
crop.
Our
host
today
will
be
katrina,
so
katrina
take
it
away
from
here.
B
C
B
All
right
so
before
we
get
started,
I
just
wanted
to
highlight
that
for
those
interested
in
customized
we
actually
have
two
customized
related
kep
pr's
open
at
the
moment.
If
you're
interested,
please
go,
take
a
look,
this
one
is
for
a
proposal
called
catalog,
adding
the
initial
kep
artifacts
for
that.
So
the
initial
proposal
documents,
the
pr
number
is
2908
and
then
the
second
one
is
updating
a
different
cap
that
was
proposed
this
past
winter
to
implementable
and
and
refining
the
details
in
there.
B
B
A
After
the
call
as
well-
and
secondly,
there
is
a
123
cap,
spreadsheet
tracking
thing:
does
that
mean
we
should
be
adding
both
of
those
to
to
the
spreadsheet.
B
A
I'll
have
a
look
at
the
caps
after
this
call
and
I'll
ping
you
in
a
dm
about
whether
we
need
to
or
not
add
them
over
there.
Okay.
B
E
I'm
here
I
actually
did
not
see
this
so
I'm
reading
through
it,
considering
we
just
deprecated
all
the
config
commands.
I
am
tempted
to
just
tell
them
that
and
market
won't
fix.
B
Let's
see
it
probably
references
the
issue,
let
me
link
to
it
quickly
here
it's
described
in
and
this
is
our
big
discussion
right.
Yes,
this
should
be
the
right
one.
B
B
B
Seems
to
me
that
if,
if
we're
allowing
repos
to
be
cloned
like
this,
then
the
load
restrictor
should
correctly
allow
them
to
be
used
as
well,
because
the
load
restriction
is
none
sounds
like
we
shouldn't,
be,
throwing
an
error
about
the
location.
A
Didn't
we
drop
because
he's
mentioning
explicitly
that
this
is
working
until
3.8,
then
we
drop
the
remote
opening
of
files
and
customize
you
to
the
problems
of
bumping
customize
in
queue.
B
That's
a
good
point:
we
dropped
hashicorp
go-getter
support,
but
not
plain
get
url
support.
If
I
remember
correctly.
B
This
is
quite
the
error
message
to
read
here,
but
I
think
it
is
saying
that
it's
cloning,
the
resources,
see
it's,
I
see
it's
cloned
it
into
private,
var,
folders,
etc,
etc,
etc.
Here,
and
you
can
see,
it's
got
the
aml
file,
so
I
think
it
did
do
the
clone.
F
B
The
default
load
researcher
policy
is
that
when
you
invoke
customize
on
a
given
tree
of
customizations
you're
pointing
at
the
the
root
customize,
the
customization,
that
is
sort
of
like
the
point
of
origin
and
all
of
the
file,
references
have
to
be
at
or
below
that
that
directory
level.
But
you
can
get
around
that
by
saying
load,
restrictors,
none
and
then
you
can
point
to
any
file
that
you
want.
F
B
Okay,
I'm
gonna
accept
it.
It
was
probably
marked
by
already
yeah.
B
So
yeah
that
needs
to
be
looked
into
further
is
it?
Is
it
also
curious?
I
think
if
it's
since
customize
is
a
complete
control,
where
it's
putting
these
things?
It's
also,
I
think,
basically,
I'm
curious
that
it's
putting
it
somewhere
that
it
doesn't
consider
acceptable,
like
two
different
directories
here
mentioned
in
the
error
message.
B
A
B
Yeah,
I
should
also
paste
these
in
the
chat.
Sorry
about
that.
Here's!
What
I'm
looking
at
right
now.
E
It
looks
like
they've
resolved
it
in
themselves
in
like
their
first
few
sentences.
It
just
looks
like
a
documentation
issue
that
they
should
be
using
the
labels
field,
not
the
common
labels
field
for
this
functionality
right.
B
G
F
B
Yeah,
probably
this
reminds
me
of
an
issue.
We
saw
last
time
where
the
number
in
question
was
in
the
image
tag
field.
B
B
B
So
it
actually
goes
through
a
few
transformations,
one
of
which
are
just
like
breed
and
martial
and
marshallings,
and
at
some
point
the
explicit
quotes
are
being
dropped
from
it
and
there's
probably
something
that
thinks
they're
not
necessary,
because
it
knows
it's
a
string
field,
but
then
something
further
on
well
once
it
gets
once
it
gets
rewritten
they're,
not
there
and.
B
B
The
other
one
is
that
that
cue
pedal
apply
should
be
able
to
successfully
apply
a
namespace.
That
has
you
know
that
this
field
is
a
string
and
that
shouldn't
cause
a
problem.
A
I
mean
the
problem
that
the
root
the
the
root
of
the
problem
is
similar
to
both
cube,
cuddle
and
customize,
and
it
lies
in
the
libraries
that
we
use
for
yum
and
json
parsing.
A
A
I'm
pretty
sure
that
this
will
be
properly
parsed
by
the
library.
But
since
this
is
a
number,
what
should
be
treated
as
a
strength
and
not
a
number
that
the
root
cause
is
sorry
inside
of
the
library.
So
it's
hard
to
fix
on
either
end.
The
shared
library
is
the
the
problem
that
we
need
to
solve.
I'm
not
100
sure.
If
that's
such
an
easy
and
obvious
fix.
B
Yeah,
I
I
guess
the
the
only
thing
that
we
could
do
on
the
customized
side
would
be
something
very
complicated
as
now
that
I
look
at
this
more
closely
this.
This
is
a
namespace
field
and
customization
that
and
the
place
where
they
want
the
quotes
or
they
need
the
quotes.
It's
just
all
over
all
of
the
objects
that
that
end
up
getting
created.
B
So
I
don't
think
currently
there's
any
way
that
we
could
easily
do
that
like
record.
The
fact
that
this
is
supposed
to
be
quoted
and
then
everywhere
that
it
effects,
which
could
be
actually
quite
a
large
number
of
places,
including
cross-subject
references
needed
to
end
up
getting
quoted.
A
I'm
not
a
customized
guru,
but
I'm
I'm
wondering
whether
they're,
whether
there
are
some
kind
of
a
filtering
that
we
could
use
to
explicitly
state
that
this
is
a
string
or
something
along
those
lines
that
would
potentially,
I
don't
know
a
give
a
suggestion,
and
this
way
we
could
pass
the
suggestions
over
to
the
parser
to
ensure
that
this
is
treated
as
a
strength,
because
the
problem
is
basically
you're
getting
a
bunch
of
characters
and
the
machine
just
can't
figure
out,
because
we
don't
have
a
way
to
state
whether
this
is
a
and
or
or
something
else
I'll.
A
Alternatively,
we
could
look
at
the
open,
api's
scheme,
enforce
that,
and
I
wonder
if
that
should
be
done
on
cuba
or
customize
size
or
both.
B
Something
that
customers
could
potentially
do
is
is
have
the
option
to
change
the
output
yaml
style,
because
some
some
styles
are
more
like
there's
a
quoted
style
that
might
do
the
trick
here,
like
default
to
quoting
strings
even
when
it
doesn't
seem
necessary.
But
I'd
have
to
look
into
it
in
more
detail.
It's
a
while
ago
that
I
looked
at
that,
and
I
I
don't
think
customize
itself
exposes
it
but
k,
yellow
might
which
could
give
us
some
options.
B
D
Sorry,
real
quick,
just
follow
from
the
last
one.
We're
saying
that
the
issue
is
that
customize
is
not
outputting.
The
quotes
for
that
namespace
name
right
is
that
the
issue.
B
Strictly
speaking,
I
think
the
yaml
that
customizes
outputting
is
correct,
like
it's
perfectly
valid,
and
the
best
fix
would
be
on
the
cube
cuddle
side,
because
that
would
help
any
like
that
would
help
users
that
had
got
their
yellow
from
somewhere
else
as
well
and
as
much
as
that
is
probably
a
fix
to
the
underlying
yellow
library
that's
actually
used
by
both.
B
But
what
I
wanted
to
like
what
the
reason
I
assigned
it
to
myself
was
to
look
into
if
there
are
any
options
that
we
could
do
for
customized
users,
who
are
specifically
encountering
this
problem,
such
as
set
a
yamo
style
that,
like
quote
strings
by
default,
whether
or
not
it
seems
necessary
instead
of
what
happens
in
the
default
style,
which
is
they
only
get
quoted.
If,
for
some
reason,
it
thinks
they're
ambiguous.
D
D
A
B
I
know
that
kml,
which
uses
the
v3
goyama
library
like
it,
does
store
that
information
about
what
we
read
it
as,
but
because
this
field
is
coming
from
customization
customization
itself,
I'm
pretty
sure
is
read
with
the
six
yama
library,
which
is
the
cmlv2
which
in
that
case,
I'm
really
not
sure
it
preserves
that
information
and
would
be
inclined
to
think
it
doesn't.
G
B
E
B
Yeah,
but
if
I'm
reading
it
correctly,
that's
because
the
image
tag
isn't
included
in
that
root,
customization
the
root
one.
I
guess
we
could
open
up
the
sample
code,
but
I'm
guessing
the
root.
One
doesn't
include
an
image
tag
at
all.
It's
only
referenced
by
this,
the
well
children
of
customizations,
so
like
it's
kind
of
a
no-op
to
run
it
on
the
route.
It's
nothing
to
do.
E
B
Would
you
be
willing
to
dig
that
issue
up.
B
B
Yeah,
so
as
an
update
on
that,
there
is
an
anchor
expansion
issue
within
a
single
resource,
which
this
example
is
a
single
resource
that
would
apply
sorry,
there's
a
fix
that
is
actually
on
master
branch
right
now,
but
it's
not
hooked
up
yet
jeff
has
been
working
on
figuring
out
how
to
actually
integrate
the
feature.
That's
already
added
in
kml
to
the
actual
customization
build,
so
that
should
be
fixed
soon,
hopefully
by
the
next
release.
But
I
don't
think
this
is
a
noble
issue.
B
In
short,
I
think
this
is
a
duplicate
of
the
same
problem
where
it's
just
like.
There's
nothing
specific
about
this.
It's
just
anchors
do
not
work
in
general.
Right
now,.
B
Yeah
I
agree
it's.
It
sounds
like
we
changed
this
intentionally
to
avoid
problems,
and
sometimes
it's
best
to
be
opinionated
about
this
sort
of
thing,
because
it's
actually
more
con
confusion
confusing
to
be
able
to
do
it.
Multiple
ways
that
have
gotchas,
which
seems
to
be
what
we're
trying
to
avoid
here.
B
E
E
I
think
I
can
dig
deeper
into
this
one.
It
sounds
like
there's
a
bug
in
the
way
the
annotations
are
storing
previous
gbks
and
I
worked
on
that
code,
so
I
can
take
a
closer
look.
E
E
B
B
B
B
F
B
I
imagine
that
would
be
quite
a
lot
of
tagging.
I
don't
know,
but
possibly.
B
F
B
Yeah
the
release
process
was
totally
different
back
then,
so
I
really
am
not
sure.
Let's
just
accept
it
since
they're
proposing
a
fix.
G
B
E
Is
there
any
reason
we
can't
just
add
them
all
like?
Why
do
we
only
have
to
add
the
first
one.
B
That's
what
I'm
trying
to
understand
by
staring
at
this
this
code,
but
I
feel
like
I
need
to
spend
more
time
reading
it
like
it's.
Is
this
gonna
cause
a
problem
like
it's
expecting
there's
only
one:
it's
gonna
get
a
single
type
meta
out
of
it
and
then
like.
Maybe
this
could
be
a
loop
because
it
theoretically
is
going
to
return
different
type
meta
for
each
of
these
definitions
that
it
actually
got
back
here
if
there's
more
than
one,
so
that
should
work
fine
yeah,
but
I
think
I
think
that
should
work.
B
B
It's
kind
of
related
to
the
issue
that
we
just
fixed
for
the
execution
directory
of
exec
functions
in
the
system.
E
B
E
B
It's
that
that's
almost
a
bigger
question
of
distribution
of
these
these
functions,
but
within
one
customization
it
should
be
relatively
easy
to
make
it
work.
B
B
There's
also
an
issue
of
whether
or
not
we'll
support
starlark
in
the
long
term,
which
is
perhaps
a
different
question,
because
while
the
alpha
exists,
maybe
we
should
still
fix
it.
But
star
alert
conclusion
is
a
problem
for
parity
and
the
standalone
and
cube
cuddle
builds
of
customized.
B
Oh,
so
I
was
trying
to
read
this
blank
document
before
the
line,
presumably
instead
of
actual
customization.
B
If
it's
easy
enough
to
fix,
we
should
probably
fix
it.
If
we're
already
able
to
read
these
with
something
that
is
reading
a
stream
we
can
just
discard.
I
wonder
if
there's
somewhere
that
says
you
know
of
these
documents,
take
document
zero
discard
the
rest
and
we
could
we
could
see
if
length
is
two
and
the
first
one's
blank
then
and
take
the
second
instead,
I
don't
think
we
should
support
multiple
customizations
or
anything
crazy,
like
that
in
the
same
file
like
it
should
be
a
single
document
that
we
end
up
processing.
B
But
if
you
start
your
your
file
with
a
document
separator,
I
don't,
I
don't
think
that's
terrible,
I
could
handle
it.
I
can
assign
this
to
myself.
B
I
I'm
asking
like,
when
do
we.
B
B
Oh,
when
I
don't
think.
B
Particular
timeline
on
that,
because
there's
a
lot
of
different
pieces
that
have
to
come
together
to
get
ready
for
that.
So
I
think
it's
at
this
point
still
when
we're
ready
we'll
do
it.
H
B
In
general,
our
our
policy
is
to
make
them
backwards
compatible
or
possible,
and
also
to
even
provide
the
customization
fix
command
where
we
can
to
help
people
migrate
from
an
old
way
to
a
new
way
of
doing
things,
even
if
the
old
way
does
still
work
and
we're
going
to
continue
to
do
that.
Certainly,
when
we're
able
to
cut
the
v1,
then
we
can
deprecate
all
of
the
old
ways
of
doing
things
that
are.
We
can
have
v1,
not
support.
B
All
the
that
have
been
long
deprecated,
but
there's
probably
going
to
be
because
customization
is,
is
a
versioned
resource.
It's
worth.
B
B
So,
even
once
we
do
do
the
release
of
the
v1
resource,
there's
still
a
period
where
the
users
have
can
migrate
over
to
it
and
and
won't
won't
experience
a
breaking
change.
H
I
G
B
I
I
don't
think
it's
possible
to
provide
a
timeline
given
the
resourcing
that
the
project
has
right
now.
Okay,
and
I
do
think
that
the
v1
beta
1
customization
resource
will
be
around
for
some
time,
so
there's
definitely
value
in
in
updating
documentation
to
reflect
the
deprecations
and
to
actually
properly
show
all
the
fields
that
are
supported
today.
If
that's
something
you're
interested
in
doing
for
the
purposes
of
this
issue,
what
do
you?
What
would
you
want
us
to
do
with
it.
B
Okay,
we
agree
that
in
principle
it
would
be
okay
to
create
that
chart,
but
actually.
B
I
think
there
already
is
an
issue
talking
about
a
bunch
of,
like
just
lists,
a
whole
bunch
of
improvements
that
we
want
to
make
to
the
docs.
So
I
think
that's
a
better
place
to
explain
this
one
specifically
just
about
the
chart
we
were
discussing
in
the
deprecation
table,
thanks
well
we're
out
of
time.
Actually,
so
thanks
everyone
for
coming
got
through
a
lot
of
issues
there.
I
hope
this
was
valuable.