►
From YouTube: Carvel Community Meeting - Oct 28, 2020
Description
Carvel's community meeting is held fortnightly on Wednesdays at 12PM ET / 9AM PT via Zoom (http://community.klt.rip)
Join our Google Group (https://groups.google.com/g/carvel-dev) to get updates on the project and invites to community meetings.
---
Agenda:
Announcements 📣
1. kbld 0.27.0 - release notes https://github.com/k14s/kbld/releases...
- can now tag images pushed by kbld, thanks to @StarpTech (https://github.com/StarpTech)
- now does additional digest check when unpacking relocated images
2. vendir 0.11.0 - release notes https://github.com/k14s/vendir/releas...
- you can now specify versions using SemVer! 😲
3. VMware Tanzu Kubernetes Grid 1.2 was released - release notes https://docs.vmware.com/en/VMware-Tan...
- using Carvel tools for templating and other stuff. Example on Twitter ➡️ https://twitter.com/lamw/status/13185...
Carvel Tech Forum 💻
1. How can we make contributing to `kbld` easier?
- GitHub issue by @StarpTech: k14s/kbld#54 (https://github.com/k14s/kbld/issues/54)
A
C
C
Here
are
some
of
the
points
about
what
made
contributing
hard
working
with
somebody
that
is
not
on
the
team.
The
other
day
also,
I
was
able
to
see
like
firsthand
some
of
these
small
nuances
that
people
that
aren't
working
in
the
code
base
every
day
might
hit
when
trying
to
contribute
a
specific
example
is
like
the
k
build.
If
you
have
k,
build
brew
installed
and
you
want
to
contribute
running,
the
test
will
use
your
brew
installed,
k,
build
and
not
the
one
that
you've
built
locally.
C
So
just
little
things
like
that.
That
could
cause
frustration
and
confusion
when
initially
trying
to
contribute
that
could
potentially
be
made
easier.
So
I
guess
we
wanted
to
just
sort
of
go
over.
Some
of
these
points
discuss
if
there's
anything
that
we
can
do
to
ease
the
frustration
and
pain
around
them.
D
C
A
C
A
C
A
C
C
A
Thinking
that
like,
if
we
sort
of
enumerate
some
things
and
then
have
a
sense
of
what
are
the
bits
that
like
demetrius
pointing
out
some
of
this,
is
mitigated.
But
what
are
possibly
like
some
high
value
things.
We
can
come
out
with
like
a
couple
of
things
to
make
improvements,
but
they're
the
high
value
ones.
That
would
be
most
useful.
C
A
A
C
A
C
And
I
think
this
is
definitely
a
case-by-case
basis,
but
like
the
wording
of
expected
something
to
match.
Something,
may
give
the
impression
that
what
follows
expected
is
what
is
expected,
especially
when
the
two
match
is
sort
of
thrown
in
in,
like
this
big
block
of
yaml
and
stuff.
That
may
be
a
little
easy
to
miss.
A
A
A
A
C
I
think
here
I
I
don't
think
that
it
was
related
to
too
much
time.
I
think
that
was
a
separate
issue,
but
I
think
that
was
addressed
by
suggesting
using
like
running
your
specific
test
while
developing
and
then
running
the
full
suite
yeah.
I
agree
with
their
approach,
but
some
places
just
missing
more
unit
or
specific
ede
tests,
for
example,
check
that
image
set
tags
have
been
really
created
on
the
registry.
C
A
A
D
You
know,
I
guess
everybody
has
various
kind
of
preferences
and
so
like
there's
a
kind
of
a
general
question
of
like
what
should
be
a
preferred
way
to
run
your
test
in
terms
of
like.
Should
you
be
using
a
a
local
registry
or
a
registry
that
you
like
docker
hub
or
whatnot
right,
and
so
I
guess,
depending
on
those
preferences
right,
some
people
will
benefit
from
the
default
setup.
Some
people
will
have
to
like
try
to
find
out
more
about.
D
D
A
A
C
C
A
C
A
Docker
hub,
typically
and
and
like
in
the
moment,
it's
you
know
we're
playing
a
story,
we're
just
trying
to
get
it
going.
So
we
didn't
necessarily
like
capture
all
the
instructors
like.
Oh,
this
is
bro
cause.
I
need
oic
dug
into
the
code.
We
need
right,
a
proper
url
for
an
image,
oh
okay
or
a
repo,
and
just
put
that
in
but
but
worked
backwards
from
that.
A
That
was
one
of
those
things
where
I
can
understand
if
somebody
doesn't
have
a
pair
with
them,
etc
that
it's
not
the
greatest
experience
to
start
with
an
error
message
and
work.
My
way
back
in
order
to
get
the
tests
going,
that
if
I
had
context-free
instructions
like
okay,
assuming
you're
using
docker
hub
you've
got
an
account,
you
can
and
you've
got
a
repo
to
spare
for
those
who
are
using
a
free
version.
A
D
A
C
D
D
D
You
know,
kind
of
a
person's
familiarity
person's
environment,
like
some
people
may
not
have
docker
hub,
but
some
people
you
know
like
I
barely
use
docker
hub,
but
it's
very
easy
to
configure
it
to
run
it
for
me
in
this
scenario,
so
it's
kind
of
like
it
in
my
head.
It
depends
on
the
person's
preference
of
how
to
set
that
up.
Like
I
I
would
not
like
I.
D
I
went
through
the
mini
cube
local
setup
once
to
figure
out
for
ci,
and
I
don't
even
remember
actually
going
through
it
entirely
because
it
was
too
painful
to
do
it.
So
I
just
you
know,
made
sure
that
the
ci
itself,
you
know,
is
able
to
do
it,
and
then
I
never
looked
at
it
again,
since
you
know
it's
just
too
painful
for
me
to
run
it,
but
it
sounds
like
sounds
like
it
was
helpful
for
this
case.
D
A
A
B
A
A
A
A
Well,
that
makes
sense,
would
hopefully
do
that
ongoing.
I
guess
the
question
is
whether
or
not
we
want
to
proactively
go
seek
out,
because
we've
gotten
feedback
that
there
is
some
tests
that
have
confusing
messages
and
the
assertions
if
we
want
to
proactively
go
look
for
those
and
make
them
make
those
improve
them.
F
D
D
D
A
D
So
in
any
way,
in
ytt,
most
of
the
tests
are
actually
you
know
component
level
tests
they're,
not
the
cli
level
tests
right.
So
you!
Actually,
if
you
look
at
ytt,
there
is
a
e
to
e
test.
That's
actually
a
bash
script
that
compares
you
know
it
actually
calls
ytt
blah
blah
and
compares
with
the
expected.
D
A
I
think
that's
what's
on
the
table,
as
I
understand
it
that
well,
if,
if
the
tests
are
invoking,
I
hear
the
I
hear
the
part
that
you're
talking
about
it's
like
well,
what
is
the
true
api?
What
is
the
thing?
That's
actually
in
practice
being
used
in
for
k,
build
it's
it's
the
binary,
and
so
I
don't
want
to.
I
don't
want
to
fast
forward
past
that
point.
A
But
I'm
hearing
this
like
pragmatic
desire
to
be
able
to
downshift
into
debugging
doesn't
necessarily
mean
that
we
have
to
have
that.
We
can't
shell
out,
but
we
would
need
to
find
a
way
to
be
able
to
do
that
like
bring
back
that
capacity
in
an
easy
way
or
write
the
tests
in
a
way
that
they,
they
do
run
in
process
like
cable,
runs
and
process.
With
the
test.
D
The
you
know,
I
don't
know
what
the
debugging
technology
is
where
it's
at
nowadays
I
I
rarely
use
the
kind
of
id
enhanced
debugging
myself,
so
I'm
not
familiar
with
what's
possible,
what's
not
possible,
I
don't
know
you
know.
Maybe
somebody
else
on
the
call
can
talk
more
about
that.
If
they're,
a
heavy
user.
A
A
A
But
I
do
think
in
general,
when
I,
when
I
see
products
that
have
a
test
suite
that
stand
back,
that
try
and
test
at
the
highest
level
that
there's
often
points
where,
because
implementation
details
are
allowed
to
float,
they
can
be
really
handy
to
just
okay,
in
this
particular
case,
run
this
high
level
test
and
set
my
break
point
at
the
at
the
part
that
I
really
care
about,
and
I
can
pretty
quickly
get
that
feedback
about.
Okay.
This
is
the
piece
that
broke
that
ultimately
bubbles
up
to
this
more
general
assertion.
Failing.
D
I
mean
you
know
it's,
it's
probably
possible
to
the
way
that
the
e2e
tests
are
designed
is
there
is
a
k-build
object
that
represents
the
k,
building
location,
and
so
you
could
probably
shim
it
in,
but
I
would
argue
that
I
would
still
want
to
have
ability
and
probably
what
the
ci
should
do
is
run.
The.
A
F
It's
something
that
I've
run
into
and
I
ended
up
using
delve,
which
wasn't
immediately
clear
how
to
use
and
wasn't
straightforward,
with
the
way
that
the
tests
were
set
up.
So
I
did
feel
that
pain
a
bit,
but
I
feel,
like
I
have
a
workaround
now
that
works
for
me,
but
it
took
me
a
while
to
get
there,
so
I
think
that
there
could
be
some
room
for
improvement
there.
F
C
One
thing
to
call
out,
though,
that's
very
dependent
on
the
fact
that
our
setup
is
long
lived
for
all
of
these
tests
like
yeah,
okay,
build
an
image
package,
create
sort
of
setup
images
or
something
like
that
and
then
run
the
command
under
test,
and
we
never
actually
delete
those
images.
But
if
there's
a
test
suite
that
does
have
clean
up
of
some
of
the
setup
steps,
then
that
you'd
have
to
somehow
kill
it
at
the
right
moment
or
put
a
break
point
in
the
test.
A
A
I
heard
something
along
the
lines
of
make
certain
pattern-based
improvements
to
test
assertions,
so
carrie
gave
us
an
example
of
hey.
If
you
look
for
a
particular
kind
of
assertion,
probably
get
real,
you
can
get
a
good
bang
for
your
buck
and
then,
like
dimitri,
was
saying
as
we
go
along
if
you
didn't
feel
empowered
to
improve
the
test
or
look
for
assertions
that
weren't
necessarily
intuitive
you
now
do
like
that,
should
be
part
of
what
we're
improving.
A
A
B
A
E
E
So
I'm
looking
at
the
as
a
devops
right,
I'm
looking
at
the
developed
first
experience,
and
I
I
have
to
I'm
looking
into
this
project-
provide
an
environment
so
that
the
developers
can
be
have
a
very
high
productivity,
and
so
what
I
would
want
to
do
is
package
their
functions
into
you
know.
Ultimately,
they
will
work
on
the
functions,
but
I
want
to
do
all
the
pipeline
of
automations,
essentially
building
the
image
with
cabled
or
a
similar
tool,
and
is
this:
is
this
the
type
of.
E
E
E
D
D
B
D
And
the
way
that
buildblocks
operator
is
there
is
no
support,
if
you
will
for
docker
files.
Kpack
is
a
system
that
effectively
provides
you
a
few
crds
in
your
kubernetes
cluster
to
convert
your
source
code
with
build
packs
into
those
docker
images
right.
So
it's
actually,
you
know
you
could
imagine.
Buildbacks
is
a
thing
that
can
convert,
but
you
have
to
kind
of
wrap
it
into
something
that
orchestrates
it
right.
When
you
say,
oh
some
new
git
code
came
into
git
repository
right.
D
You
want
to
be
able
to
you
know
you
want
to
be
able
to
build
it
and
then
publish
it
as
an
image.
Right,
yes,
now
separately
from
that
k,
build
is
a
tiny
tool,
part
of
the
carvel
tool
suite
to
make
it
easy
for
users
who
are
dealing
with
kubernetes
yaml
to
build
and
resolve
the
image
references,
and
you
can
kind
of
see
k
build
as
more
of
a
thing
that
maybe
calls
out
to
other
tools
to
build
things.
D
So
you
could
imagine
like,
for
example,
k,
build
calling
out
to
k
pack
to
build
something.
Currently,
it
only
calls
out
to
two
things:
it
calls
out
to
docker
daemon
where
it
can
just
does
docker
build
so
that
docker
files
are
supported,
and
then
it
also
supports
pax
cli,
which
is
effectively
like
the
cli
version
of
kpac,
which
also
runs
locally
and
so
that
that
uses
you
know
that
uses
the
same,
build
packs
to
convert
your.
D
If
you
are
trying
to
kind
of
build
kind
of
a
diy
function
as
a
service
system
right,
you
could
imagine
using
build
packs
to
convert
your
source
code,
the
function
right
into
something
that
runs
right.
In
fact,
I
think
there's
probably
a
few
people
in
the
community
who
has
been
and
we're
looking
into
that
and
probably
shipping
products
around
that.
D
D
Will
you
you'll
probably
would
want
to
make
a
system
that,
maybe
I
mean
it
depends
I
can
imagine.
I
could
imagine
various
ways
how
one
can
use
and
not
use
cape
building,
this
kind
of
thing
right,
okay,
so
it
kind
of
depends
where
you're
drawing
the
obstructions,
what
kind
of
apis
you're
you're
trying
to
offer
what's
running
behind
those
apis
as
an
implementation,
detail.
E
D
E
E
D
I,
to
be
honest,
you
know,
there's
there's
various
teams
in
in
the
organization
there's
definitely
folks
that
are
looking
at
stuff
like
that,
I'm
not
sure.
If
there's
any
kind
of
you
know
product
store
on
that,
but
we
do
have
a
few
folks
that
are
the
founders
and
maintainers
of
k,
native
working
for
vmware.