►
From YouTube: Kubernetes SIG CLI 20210804 - 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
B
Hey,
thank
you,
everybody
for
coming
to
this
customized
bug
scrub.
Today
we
are
going
to
largely
use
the
lovely
triage
party
instance
that
eddie
has
set
up
to
go
through
the
issues
that
have
the
needs:
triage
label,
I'll
drop,
a
link
in
the
chat
to
that
board.
Oh
eddie's
already
dropped
it.
There
we
go,
but
before
we
get
started
on
that,
are
there
any
pressure,
pressing
customized
issues
that
anyone
wants
to
bring
up
for
us
to
look
at.
B
Am
I
able
to
let's
see
yes,
I
am
excellent.
Yes,.
A
Now
it
fails
the
entire
wide
screen
of
mine
and
before
it
was
half
of
that,
okay.
B
Great
all
right
so
does
anyone
have
an
issue
that
they
wanted
to
discuss
before
we
get
started
on
the
list.
B
B
They
have
data
inside
a
config
map
and
they
want
to
be
able
to
to
have
patches,
merge
that
data
that
it's
not
something
that
would
usually
want
to
do,
because
you
can't
assume
that
there's
structured
data
inside
just
a
string
field,
but
it's
it
is
true
that
it's
very
common
for
config
maps
to
contain
structured
data
like
this.
For
example,
I'm
aware
of
some
ingress
implementations,
some
popular
ingress
and
implementations
that
that
do
that.
B
D
Can
you
go
back
up
to
the
top
again?
Yes,
sorry.
D
My
initial
thought
is
yuck
in
the
sense
that,
but
maybe
that's
not
my
final
thought,
but
my
initial
thought
is
config.
Maps
are
kind
of
it's
healthy
to
think
of
them
as
being
just
holders
of
opaque
data,
but
I
don't
know
this
might
be,
it
might
be
doable.
I
mean.
Certainly
it's
doable.
B
And
12
on
the
new
one,
I'm
not
sure
if
there
were
only
these
two
issues.
I
think
it's
been
brought
up
before
and
do
you
know
what.
D
Their
what
the
use
case
is
for
this,
because
is
it
some
sort
of
hack
that
people
are
using
config
maps
to
store
ammo
inside
and
then
extract
it
for
some
other
purpose
or.
B
I
I
mean
you
could
characterize
it
as
a
hack,
but
I
think
it's
a
common
pattern.
It
would
be
a
more
generous
way
to
phrase
it.
B
A
I
can
speak
for
what
we
do
in
openshift
and
we
have
a
little
rely
on
the
fact
that
we
store
some
configuration
in
the
config
maps,
but
my
initial
response
probably
would
be
along
the
lines
of
jeff,
because
the
container
here
is
the
conflict
map.
Whatever
is
inside
of
the
config
map.
It
should
rather
be
orthogonal
to
the
conflict
map
itself,
so
I
would
probably
can
we
push
this
back
as
I
I'm
throwing
here
ideas.
A
Can
we
push
this
back
as
a
customized
plugin,
I'm
not
sure
if
customize,
how
far
cut
I'm
not
a
super
pro
customized
user.
So
maybe
we
could
approach
this.
If
people
are
interested
and
maybe
they
could
submit
a
a
plug-in
and
see
how
useful
that
will
be,
and
then
we
can
eventually
re-uh
revisit
that
decision
and
include
that
as
a
default.
B
D
I
think
if
somebody
wanted
to
explore
this,
they
could
my
my
just
to
reiterate
and
emphasize
what
montre
just
said.
What
I
said
earlier
is
that
the
data
is
supposed
to
be
it's
opaque
in
the
sense
that
you
can't
really
address
a
yaml
object
inside
configmap,
using,
for
example,
a
json
path.
Like
we
don't
say,
a
config
map
has
to
maintain.
It
could
be
anything
you
know
it
can
be
your
mom's,
a
letter,
your
mom
it.
It
could
be
anything
so
somebody
would
have
to.
D
I
would
want
to
hear
what
the
use
case
is
and
try
to
figure
out
if
there's
some
other
way
to
approach
it
like,
if
someone's
using
it
to
store
kubernetes,
yaml
or
something
and
then
extract
it
and
using
some
other
kubernetes
context,
that'd
be
pretty
weird,
so
yeah
off
the
top
of
my
head.
I
want
to
hear
somebody
come
and
present
a
a
use
case
for
this,
that
we
could
talk
about.
B
I
think
this
this
was
part
of
the
presentation
that
was
made
about
customized
at
a
big
seal
as
a
ccli
meeting
a
while
ago.
If
you
remember
this
individual
came
and
was
making
the
case
for
a
bunch
of
features,
for
I
think
it's
mentioned
in
the
issue.
Cube
flow,
yeah,
cube
flow
yeah
and
the
person
was
willing
to
to
contribute
for
these
features
that
they
felt
strongly
about.
I'm
pretty
sure
this
was
one
of
them.
B
I
I
agree
100
that
in
theory,
the
the
data
is
opaque,
but
in
practice
people
are
using
it
for
this
a
lot,
it's
very
easy
to
mount
a
config
map
into
your
environment
or
as
a
file,
and
that
makes
it
a
very
easy
way
for
applications
that
are
using
structured
configuration
to
get
that
configuration
into
the
application.
B
It's
just
I'm
not
saying
it's
a
great
idea,
but
I
think
it
is
fairly
common
and
we
definitely
should
not
assume
that
the
data
in
the
configmap
is
structured
100.
That
would
be
terrible
idea,
but
I
think
they're
not
necessarily
asking
for
that.
They
like
it
would
work
if
they
were
able
to
to
tell
us
that.
D
My
objection
is
based
on
purely
the
idea
that
I
want
to
hear
what
the
problem
is
a
little
bit
more
because
that
approaches
from
well.
Why
can't
we
use
a
generator
for
this
to
to
generate
something
specific
that
has
parameters
in
it
rather
than
try
to
patch
an
existing
config
map,
so
in
other
words,
they're
coming
with
here's,
an
implementation
of
how
I'd
like
to
solve
the
problem,
and
I'd
like
to
hear
more
about
the
problem.
First,
you
know
and
from
a
generated
point
of
view.
D
Maybe
we
can
like
in
in
the
language
of
components
you
could
edit
the
generator
arguments
before
the
thing
generates
the
config
map
and
maybe
achieve
the
same
effect
rather
than
trying
to
like
dig
into
config
maps
and
treat
their
data
as
yammer
that
we're
going
to
parse.
B
Right
they're
definitely
saying
that
they
need
some;
they
need.
I
guess
one
reason
why
they
might
be
reaching
for
a
patch
is:
is
that
conceptually
they
have
a
base
config
and
and
a
change
that
they
want
to
make
to
it
in
a
different
layer
right.
So
I
understand
why,
conceptually
that
maps
to
the
base
and
patch,
are
you
saying
that
if
they
they
built
a
generator,
they
could
get
that
same
effect
by
modifying
the
generator
config
right
and
the
overlay.
D
B
Okay,
I
can
assign
this
to
myself
and
leave
a
comment
asking
for
more
information
and
sort
of
explaining
what
we
just
discussed
thanks.
B
C
I'll
try
edge
this
one.
This
one
is
just
documentation,
don't
worry
about
it,
I'll!
Try
it
today
and
write
the
documentation.
C
C
C
B
Is
this
one
similar,
or
is
this
one
a
real
bug?
It's
about
replacing
this
again
do
not
work
with
fields.
B
B
There's
a
there's
a
lot
of
detail
here.
Have
you
already
sort
of
understood
the
problem
that
you
could
provide
a
quick
summary
natasha.
C
C
B
And
what
is
the
the
use
case
like?
Why
do
they?
Why
are
they
unable
to
achieve
the
same
result
without
breaking
encapsulation?
C
Let's
see
so,
if
you
read
the
second
to
last
comment,
the
guy
describes
a
more
specific
use
case,
they're,
trying
to
express
that
an
environmental
environment
variables
value
should
point
to
this
service
and
that
service
is
defined
in
a
parent
and
they
want
to
use
that
value
rather
than
a
value
in
base.
D
B
Yeah,
I
really
don't
want
to
reintroduce
that
global
like
that.
That
works
differently
than
all
the
other
fields
complicates
the
mental
model,
and
I
I
would
really
rather
not
it.
D
You
could
explain
you
could
explain
why,
and
you
know
here's
why
it
breaks
or
here's
what
the
drawbacks
are
and
leave
it
open.
As
a
okay
mine,
I
would
leave
it
open
and
think
about
if
there,
if
it'd
be
okay,
to
do
that.
I
I'm
not
I'm
not
fully
understanding
the
problem
at
this
point,
so
I
I
don't
want
to
close
it
right
off
the
bat.
Maybe.
C
D
I
was
looking
this
morning
into
the
the
anchor
issue,
so
that
was
my
focus
and
I
didn't
read
this
one
in
advance.
B
D
No,
I
I
I
looked
into
it
and
started
when
I
was
writing
a
unit
test
to
document
it
realize
that
there's
probably
a
decent
way
to
fix
it
on
the
ktml
side.
So
I
want
to
explore
that
a
little
bit
more.
D
I
think
we
can
fix
it,
but
there
are
some
edge
cases
in
here
where,
if
you
mix
replacements
with
with
anchors
you're,
probably
going
to
run
into
some
trouble,
but
I
think
I
can
send
a
pr
out
that
will
expand
the
anchors
I'm
just
debating
whether
or
not
I
want
to
expand
the
anchors
before
I
expand
everything
else,
because
that's
the
problem
is
the
use
case
of
mixing
anchor
replacements
with
customized
style
replacements
so
anyway,
I'll
send
a
pr
out.
D
But
I
just
wanted
to
comment
that
I
took
the
bug
I
signed
since
I
am
now
familiar
with
what
the
problem
is
and
I'll
work
on
it.
B
Great
yeah
for
context,
this
is
a
very
popular
problem
that
didn't
exist
in
the
older
version
of
customize.
It
came
with
the
conversion
to
kml
and
we
had
sort
of
solicited
feedback
from
the
users
about
how
important
this
was
to
them,
and
given
that
customize
has
a
lot
of
different
ways
of
of
well
creating
and
composing
configuration
and
a
lot
of
people
chimed
in
after
we
asked.
Did
you
see
the
scroll
here
with
lots
of
thumbs
up
on
the
comments
about
this
being
recorded?
B
One
thing
that
you
had
mentioned
jeff
was
that
you
thought
one
of
the
common
use
cases
which
was
patching
containers.
D
B
Of
resources,
you
were
saying
that
was
possible
using
a
customized
native
mechanism,
and
that
was
one
of
the
primary
objectives
on
here.
So
I
think
it
could
still
be
useful
to
comment
with
that.
If
it
was
proved
true
that
it
could
be
addressed
that
way,.
D
Sorry
did
that
make
sense
I'll
do
that
I
want
to.
I
want
it
to
make
sense.
Yes,
so
if
people
are
primarily
these
anchors
to
like,
say,
expand
either
container
instances
inside
a
pod
template
or
expand
pod
templates
inside
different
things
like
deployments
or
replica
sets
or
cron
jobs
or
whatnot,
I
think
that
using
customized
mechanisms
to
do
that
is
probably
better
than
trying
to
use
the
anchors,
because
anchors
are
just
you
know,
weirder
and
harder
to
maintain,
but
it's
it's.
D
I
think
there
is
a
way
to
get
anchors
to
work
before
customize
is
a
processing
begins,
just
basically
expand
all
the
anchors
and
see
if
that
solves
the
problem
that
might
not
solve
the
problem
for
some
edge
cases.
So
we
can
iterate
on
this
for
a
while.
It
may
take
a
couple
of
releases
to
get
to
a
place
where
we're
all
comfortable
with
mixing
anchors
and
and
customized
replacements
like
maybe
maybe
we
do
anchors
last.
D
You
know,
obviously,
if
somebody's
got
a
if
somebody
has
a
patch
that
has
an
anchor
in
it
doing
anchors.
First,
won't
work
so
I'll.
Take
it
I'll
I'll,
throw
out
a
fix
that
does
the
anchors
first
and
see
how
that
flies,
and
we
can
hash
out
where,
where
we
expect
the
anchor
to
expand
it
first
or
last.
B
D
So
if,
let's
say
a
patch
had
an
anchor
in
it,
there'd
be
no
way
we
wouldn't
be
able
to
expand
the
anchors
first
so
like
whenever,
when
we
read
a
document
right,
when
we
read
any
documents,
we
could
go
ahead
and
replace
any
anchors
right
off
the
bat.
D
There
might
be
dangling
pointers
effectively
because,
like
the
if
the
target
of
the
anchor,
if
the
reference
is
a
isn't
an
overlay,
it's
not
going
to
work.
This
is
why
I
actually
don't
like
this
mixed
this
mixing
of
these
two
concepts,
but
it
might
for
simple
cases
like
in
a
in
one
document
where
an
anchor
is
referring
to
some
reference.
That's
in
the
same
dock,
then
expanding
the
anchors
right
away
would
work.
Fine.
B
B
If
the
the
thing
that
wouldn't
work
easily,
is
that
that's
even
supported
by
the
yaml
spec,
because
if
you
don't
feed
it,
the
complete
document
that
has
the
reference
and
the
source
in
it,
then
how
would
that
how's
that
supposed
to
work?
It
seems
interesting
to
me
that
all
we
would
need
is
the
first
thing
that
will
work
easily.
D
That's
what
I'm
going
to
do?
First,
I'm
going
to
do
the
first
thing
that
would
work
easily,
which
is
expand:
the
anchors,
the
anchors
locally.
It's
just!
If
somebody,
if
somebody's,
expecting
this
to
work
in
a
patch
and
and
hit
work
across
patches,
then
it's
it's
not
going
to
fly
so
we'll
see
how
it
goes
I'll,
put
a
put
out
a
pr
we'll
do
a
release
and
see
what
kind
of
feedback
we
get
on
the
issue.
We'll
keep
the
issue
open
for
a
while.
B
Great,
I
will
accept
that
then,
if
that's
all
right.
B
C
Yeah,
I
just
read
this
issue
and
it
looks
like
they
have
their
referencing
the
same
base
from
two
different
overlays
without
changing
the
name.
So
I
think
that
this
error
is
expected.
I
can
assign
it
to
myself
and
respond
to
them
great.
Thank
you.
B
All
right,
so
we
now
have
some
older
ones
that
we
can
look
at
that
are
missing
the
label.
It's
possible
in
this
case,
because,
just
as
a
note,
this
repo
didn't
use
the
nice
triage
label
until
I
don't
know
a
month
ago.
So
it's
possible
that
some
of
these
have
been
triaged
and
will
be
just
need
to
need
the
label
to
be
applied.
So,
let's
see
just
going
to
scroll
down
to
the
bottom.
No,
this
one
is
genuinely
untried,
unable
to
run
customize,
build
on
remote
base
and
git.
D
That
shouldn't
happen.
I
was
thinking
this
might
be.
Oh,
we
don't
recognize
github,
urls
or
sorry
labels,
but.
C
There
was
another
issue
in
customized
for
index
out
of
range
and
someone
submitted
a
pr
to
fix
it,
so
it
might
be
fixed.
D
Looks
like
it's
in
rebus
back
and
yeah.
This
looks
just
like
a
horrible
kind
of
bug
in
the
parsing,
the
the
regardless.
The
error
message
should
be
better.
It
shouldn't
fail
on
index
out
of
range.
It
should
be
something
like
you
know.
You
could
start
off
with
well,
we
don't
recognize
this
gitlab
url
and
then
later
add
support
for
gitlab
urls.
D
B
D
I
don't
buy
that,
but
I
don't
see
how
they
could
work
with
couple
customize
and
fail
and
customize
native.
B
Okay,
like
it
could
be
a
super
old
version
right,
it.
D
Might
be
really
yeah,
it
might
be
really.
It
might
date
back
to
like
something
that
used
the.
What
was
that
dependence
that
brought
in
all
the
the
hash
of
corp
go-getter
thing
that
might
have
actually
worked?
D
So
maybe
it
did
work
at
one
point
and
we
had
to
take
it
out
because
the
the
go-getter
couldn't
be
used
in
see
go-getter
can't
be
used
in
in
coop
cuddle.
So
I
don't
see
how
that
could
possibly
work
in
the
kubeco
version
of
customized,
but.
D
D
B
Eddie,
do
you
know
if
I
did
that
wrong,
because
the
bot
didn't
answer
really
quickly.
B
E
B
This
one
natasha,
you
said
kind
support,
so
you've
had
a
look
at
this
one.
Potentially.
B
B
B
Count
a
marshall
number
into
ghost
struct
field,
image,
dot,
image,
tags,
new
tag
of
type
string,
so
they
have
an
integer
like
tag
or
a
well,
not
integer,
but
a
number
like
tag
and
yaml
is
thinking
that
that's
a
number
not
not
a
string.
I
think
this
is
a
pretty
classic
problem.
B
Does
anyone
have
any
insight
like
I
I'm
pretty
sure
putting
quotes
around?
That?
Would
work
just
fine,
but
do
we
consider
this
a
bug
that
is?
Is
this
at
all
fixable
on
our
side.
D
Well,
we
could
always
special
case
it,
but
I
would,
I
would
suggest
putting
in
quotes
yeah.
Obviously.
B
B
D
Yeah,
it's
just
that
if
we
had
the
unit
test,
we
could
say
we
could
point
to
it.
If
this
gets
filed
again
and
again
and
again
and
say
this
is
the
reason
why
and
see
and
and
also
make
sure
the
quotes
solve
the
problem.
B
B
And
so
work
around
putting
the
tag
or
yeah
should
get
around
the
issue.
Yeah,
please
let
us
not.
The
case
sounds
good.
D
D
D
B
D
D
Value:
okay,
some
new
value
right;
okay
and
what's
the
actual
output,
a
lot
of
stuff
gone,
it's
not
being
merged,
so
it's
replacing
the
entire
list
of
environment
variables.
It
looks
like.
A
I
would
probably
just
start
with
checking
what
the
because
that's
part
of
the
pot
spec
specifically,
I
will
check
what's
the
merge
strategy
on
the
containers
field
in
queue,
because
if
I
remember
correctly,
you
have
the
entire
open
api
embedded
and
customized
that's
right,
yeah,
and
that
contains
the
information
about
how
you
are
working
with
those.
B
D
Yeah
write
that
in
a
comment-
that's
that's!
That's
the
thing
to
investigate.
D
A
But
the
key
is
name
strategy
is
merged,
so
theoretically
it
should
work.
It
might
be
the
the
the
the
problem
of
v1
beta
1
versus
v1
and
cron
minecraft,
because
that
was
promoted
in
in
the
last
release.
121
yeah,
it's
a
it's
a
brand
new
api
that
I
managed
to
promote.
Finally,
after
years,
thank
you
veda.
B
I'll
sign
this
to
myself
as
well
and
yeah
I'll
accept
it.
D
C
This
might
be
related
to
like
there
was
a
kml
panics
on
duplicate
keys
issue,
so
we
changed
it
to
throw
an
error
instead.
So
I
think
that's
the
issue.
There's
a
duplicate
key
and
we're
throwing
an
error.
C
C
But
yeah
looking
at
their
very
ammo
there,
there
is
a
duplicate
key.
So
if
this
is
invalid,
then
we
can
just
tell
them
that.
A
According
to
this
pack,
the
first
one
should
be
taken
into
account
and
follow
up.
C
So
ignored,
even
if
this
is
valid
yaml.
This
is
a
problem
with
the
go
yaml
library
like
the
goyama
library,
is
the
thing
panicking
when
there's
duplicate
keys.
So
I'm
not
sure
what
we
can
do
about
that.
B
So
I
guess
what
are
our
options
here?
We
can.
We
can
tell
what
we
should
tell
them.
The
problem
is,
the
duplicate
has
nothing
to
do
with
the
home
generator
in
particular,
but
then
we
can
direct
them
to
open
an
issue
on
goyamo
and
get
offered
to
open
it
ourselves.
A
B
Documents:
okay,
the
issue
is
you
know
what
I'll
just
I'll
type
this
up
clear
and
click
it,
and
I
will
submit
the
comment
later,
so
you
don't
have
to
watch
me
type.
B
D
B
D
D
And
they're
trying
to
do
this
at
head,
no
310,
it
looks.
B
D
A
I'm
guessing
the
problem
might
be
that
I
can't
remember
how
customized
handles
this,
but
I
think
in
some
depths
we
are
uploading.
Them
live
instead
of
using
the
vendor
directories,
which
might
be
causing
problems
because
you
might
be
picking
some
newer
version
of
some
directories.
D
Yeah
and
exactly-
and
we
don't
keep,
we
don't
maintain
tests
to
make
sure
this
sort
of
thing
works
right.
A
Yeah,
so
a
temporary
solution
that
I
could
probably
suggest
is
replacing
some
of
the
the
libraries,
maybe
with
a
comment
id
and
try
to
re-vendor
those
and
then
build
locally
yeah.
It
might
be
tricky,
but
I'm
I'm
pretty
confident
that
we
can
nicely
respond,
that
this
is
not
maintained
by
the
community
and
and
just
close
it
as
such.
D
A
It
seems
a
little
bit
excessive
initially,
but
in
the
long
run
and
if
you're
thinking
about
re-repetitive
or
whatever
the
name
is
basically
that
you
have
a
stable
output
from
from
given
input,
which
is
basically
one
of
the
security
concerns
the
vendor.
I
mean.
A
B
Goal
yeah
older,
so
releases.
B
D
A
A
And
then
you
can
basically
identify
how
many
additional
changes
you're
bringing
in
whether
there
are
some
conflicting
changes,
and
on
top
of
that,
you
can
also
verify
the
licenses
licenses
of
the
dependencies.
So
basically
all
of
that,
what
cube
is
already
doing
on
every
dependency
and
I'm
pretty
sure
that
eddie
can
speak
about
it
because
he
he
went
through
the
hassle
of
updating
at
least
several
dependencies
in
the
past.
E
B
B
B
D
Oh,
and
is
this
all
right
so
I've
remember
the
last
time
I
looked
at
this.
The
test
was
failing
and
the
I
asked
the
persons
here
to
create
a
passing
test
that
shows
the
incorrect
behavior
to
prove
that
it
actually
exists.
D
A
B
Yeah,
we
need
more
information
about
what
they
mean,
I'm
at
a
guess.
They
want
this
to
be
injectable
at
in
the
customized,
build
go
code.
B
E
B
Using
this
example
at
the
entry
point
for
this
one,
we
do
for
a
customized
build
I'm
pretty.
B
B
I
think
there
could
be.
Maybe
this
is
something
we'd
be
willing
to
do,
but
since
the
camel
well
the
v3
is
implementing
1.2
spec.
There
are
some
minor
differences,
such
as
boolean
interpretation.
B
B
D
Well,
it
doesn't
take
long
to
read
the
customization
file.
We
could
read
it
both
ways
and
see.
If
that
see,
if
there
was
that
the
marshalled
object,
sorry,
the
unmarshalled
object.
D
Has
you
know,
discrepancies
and
is
is
equal
from
a
reflection
point
of
view.
That
would
tell
us
if
bools
are
being
interpreted
incorrectly
or
something
okay
and
then
we
can
do
that
and
then
we
can
just
fail
and
then
we
we
then
we
can
report.
We
could
just
fail
and
say:
hey
fix
your
one
way
to
deal
with
it.
B
D
B
D
D
B
B
Yeah
I'll
assign
this
to
myself
and
like
find
the
cross
reference
issues,
because
the
the
current
release
doesn't
have
the
configuration
parameters,
but
it
does
have
the
revert
to
the
the
more
customized
preferred
style.
So
I'll
go
track
down
that
information.
D
B
All
right,
yeah,
that's
that's
a
good
point.
Yeah.
I
think
that's
enough.
We
got
through
a
lot
of
issues.
I
think
this
is
really
great.
This
is
my
first
time
ever
running
a
bug,
scrub
or
a
meeting
in
general.
So
if
you
have
any
feedback,
I'd
be
eager
to
hear
it.