►
From YouTube: Implementations Sync: 2020-08-20
Description
* Status Updates
* Release Planning
* Meta: Meeting Scheduling
* Profile script bug: https://github.com/buildpacks/lifecycle/issues/359
B
B
C
B
D
B
C
B
Life
cycle
maintainer,
I
feel
like
on
some
level.
It
should
be
my
responsibility
to
facilitate
this,
but
because
there
have
been
several
sub-team
sinks
for
both
sub-teams
that
I've
not
been
to
yet
because
I
was
on
vacation.
I
was
wondering
if
someone
else
wanted
to
kick
us
off
and
walk
through
what
the
format
is.
C
B
Sounds
good
to
me,
I
can
kick
off
with
an
update
this
week,
got
off
to
a
bit
of
a
slow
start
for
me,
because
I
was
catching
up
from
a
deep
backlog
of
vacation
things.
One
of
the
things
I've
been
doing
is
trying
to
create
spec
and
lifecycle
issues
to
mirror
all
the
rfcs
that
have
passed
so
that
we
can
plan
the
work
for
the
next
couple
releases
also
been
spending
some
time
trying
to
work
through
the
plan
for
stackpacks,
and
that
will
continue
later
today.
Working
group.
B
B
B
All
right,
so
I've
created
two
milestones,
one,
I
think
a
we
should
do
one
more
patch
of
life
cycle
092,
because
we
have
a
couple
bugs
in
our
issues
right
now
and
then
also
a
plan
for
our
next
feature:
release
lifecycle,
10..
So,
let's
start
with
nine
two
I
threw
most
of
the
bugs
in
here
that
don't
require,
like
rethinking
in
the
spec
to
complete.
B
B
The
other
things
I
was
thinking
are
this
issue
filed
by
ryan
right
now,
if
there's
one
bad
cash
layer
and
it's
bad
because
it's
empty,
therefore
there's
nothing
to
cash.
We
fail
the
whole
build
he's
suggesting
that
we
should
warn
instead
and
comes
up
with
some
realistic
examples
of
when
this
might
happen
when
it
isn't
a
shouldn't,
be
a
fatal
situation
and
I
think
that's
reasonable
and
something
we
could
get
out
in
a
pretty
short
amount
of
time.
B
B
B
B
I
would
like
to
try
to
get
this
in
0
9
2,
but
because
it's
more
complicated,
I
think,
there's
a
chance
depending
on
what
we
discover
here.
This
may
result
in
spec
changes
or
a
larger
change,
so
we
need
to
go
through
a
different
process
and
it
could
get
booted
into
the
next
release,
but
I
think
it's
worth
trying
to
explore
and
see
if
we
could
fix
this
in
the
patch.
D
B
A
B
B
B
Back
to
our
patch
release,
planning,
there's
another
windows
issue
here,
where,
if
the
command,
I
haven't,
actually
filled
this
in
with
details,
yet
I
just
found
it
right
about
so
I
know
I
went
on
vacation,
so
I
apologize
for
this
blank
issue
here
I
will
fill
it
in
the
issue.
Is
that
the
way
profile
scripts
are
executed?
If
your
command
references,
an
environment,
variable
that's
set
in
the
profile
script,
it's
the
cmd
is
trying
to
interpolate
it
before
the
profile
scripts
run.
So
you
don't
actually
get
the
value.
B
So
if
your
command
calls
a
like
a
batch
script
that
references
the
profile
variable,
it
will
work
because
then
it
gets
evaluated
when
the
batch
script
runs.
But
if
your
command,
the
whole
thing
exists
on
the
command
line,
then
it
doesn't
get
evaluated,
and
this
is
actually
something
I
wanted
to
talk
about,
because
there's
some
ambiguity
about
how
we
could
go
forward
with
this.
The
way
this
works
in
linux
is,
we
basically
have
two
nested
invocations
of
bash.
B
We
can't
do
this
on
the
command
line
in
windows,
because
windows
command
doesn't
allow
for
a
batch
script
in
a
string
like
bash
does
so.
The
only
path
forward
I
could
come
up
with
here
is
to
actually
write
a
temporary
batch
script
containing
the
command
and
then
invoke
it,
which
is
because
you
have
to
write
a
file
like
not
the
most
performant,
but
it
is
correct
wondering
if
anyone
had
any
better
ideas
on
this.
B
C
B
B
I
think
that's
the
strategy
we
can
go
forward
with,
so
I
can
put
on
my
list
to
like
flush
this
out
and
then
we
can
solve
the
profile
script
bug
because
profile
scripts
aren't
particularly
useful
if
they're
they
hold,
they
exist
to
set
these
values.
So
if
you
can't
use
the
value
in
the
command,
then
it's
not.
It's
not
great.
D
D
D
B
So
if
we
don't
pass
cnb
registry
off,
our
auth
falls
back
on
the
docker
config
json,
but
there
are
cases
where
neither
of
these
things
are
set
and
we
could
still
proceed
successfully
because
anonymous
auth
would
work
to
read
the
types
of
images
we
want
to
read
and
instead
of
proceeding
with
anonymous
auth,
this
blows
up.
So
I
think
we
should
change
this
to
fix
this
bug.
We
should
fall
back
on
anonymous
if
neither
of
these
strategies
are
available.
D
D
Like
where
the
fallback
should
occur
so
right
now,
I
guess
we're
like
not
being
particularly
smart
about
the
home
directory.
We
just
try
to
stat
it,
and
if
we
don't
have
permission
we
we
fail.
I
wonder
if,
like
should
we
just
after
that
failure,
try
to
do
anonymous
auth
or
is
there
some
sort
of
like
logic
that
we
might
need?
That's
you
know.
Oh.
C
I
don't
think
so.
It
kind
of
feels,
like
you
know,
a
good
error
message
or
a
warning
would
go
a
long
way
like
just
being
like
we
couldn't.
We
couldn't
do
this
so
that
you
could
backtrack
a
failure
to
something
that's
unexpected,
but
I
wouldn't
want
to
fail
the
whole
bill.
We've
had
similar
kind
of
not
just
with
analyzer
even
the
exporter.
B
This
seems
reasonable
to
me.
I
think
we
should
warn
when
neither
of
the
inputs
are
available
that
doesn't
actually
capture
every
anonymous
situation,
like
sometimes
you
get
a
config.json,
but
it
doesn't
have
an
entry
for
your
registry
and
you're
still
anonymous.
So
I
don't
think
we
need
to
warn
in
all
anonymous
cases,
but
I
think
we
should
warn
if
we
don't
find
any
configuration
for
authentication
at
all.
B
B
B
So
I
do
want
to
indicate
that
these
are
ready
but
keep
an
eye
on
them.
You
know
what
I
mean
and
I
had
an
ugly
status
label
that
was
status
pending
final
spec,
but
it
was
wordy
and
I
stopped
attaching
it
to
things,
because
I
thought
it
just
looked
confusing,
and
I
was
wondering
if
maybe
at
the
next
sub
team
sync,
we
all
wanted
to
come
in
with
different
labeling
strategies.
So
we
could
think
about
how
to
indicate
types
of
status
that
are
particular
to
the
life
cycle
and
its
relationship
to
the
api.