►
From YouTube: NixOS Office Hours, 2020-01-24
Description
Today we'll be discussing package quality control and questions from participants.
A
A
A
A
Apologize
I
turns
out,
I
had
broken
out
the
audio
at
some
point,
but
that
should
be
back
always
exciting
yeah.
So
there
is,
there
has
been
for
several
years
now,
an
effort
to
remove
or
fix
packages
that
are
broken.
It's
interesting
because,
on
the
other
hand,
I
know
some
people
have
been
wondering
what
the
harm
is
of
having
broken
packages
as
if,
if
there's
even
a
broken
template,
that's
available,
maybe
that's
better
than
nothing,
but
but.
B
B
A
C
D
B
Can
see
what
you
mean
there
because,
like
an
express
expression,
is
never
a
template,
but
that
is
something
that
could
probably
be
an
RFC.
If
you
wanted
to
do
to
move
out
those
large
model,
expressions
that
are
broken
and
I
think
when
you
mentioned
it
was
a
nur
type
of
thing.
That's
basically
how
I
believe
a
similar
features
that
flakes
could
provide.
D
C
This
sounds
like
hoarding
expressions
which
I'm
not
sure,
is
a
good
idea.
If
they're
broken
they're
broken
there
probably
is
no
interest
to
fix
them
right
now
and
sorry,
oh
yeah,
and
what's
what's
the
actual
good
in
keeping
broken
expressions
that
I
think
a
good
chunk
of
them
going
to
be
trivial
and
could
just
be
rewritten
once
required.
B
A
Of
the
things
but
I
think
about
here,
Oh
what
you
speak
in
just
a
moment
Samuel,
but
is
with
many
distributions
the
channel.
The
release
mechanism
is
not
a
single
coherent
set.
Wake
up
mix
packages
is
not
everything
has
to
build
all
at
the
exact
same
time,
and
so
on
other
distributions,
it's
much
easier
to
keep
something
that
might
be
broken.
Working
for
longer
and
I.
Think
that's
a
natural
tension
of
the
way
NYX
OS
is
architected.
B
Art
the
artist
blattis
mentioned
in
his
office
opinion
that
anacs
packages
attribute
is
a
promise
of
me
being
able
to
build
that
package.
A
broken
package
is
a
broken
promise
and
a
failure.
Next
packages,
maintainer
ship,
so
they're
being
broken
packages
in
NYX
packages.
Maybe
it's
just
because
people
feel
that
they
need
to
contribute
everything
Synnex
packages,
the
half
that
when
it's
definitely
not
the
case,
you
can
have
an
overlay
flakes
would
be
a
great
thing.
We
could
have
that
right
away
to
get
that
stuff
out
there.
B
A
I
agree
with
that
I
think
having
were
a
culture
of
mm
keeping
some
things
to
yourself,
even
if
the
Republic
and
open-source
doesn't
have
to
go
into
NYX
package
itself,
I
one
of
the
things
about,
if
there's
an
expression
that
doesn't
build
being
a
broken
promise,
I
I,
agree,
I
feel
daunted.
By
having.
A
D
Well,
in
go
ahead,
you
can
be
fairly
aggressive
about
marking
packages
as
either
well
as
unsupported
on
some
platforms.
So
basically
you
go
to
hide
Ryu.
You
look
at
so
everything
that's
broken
and
you
you
mark
them
as
as
either
broken
or
unsupported
on
the
platforms
where
they're
broken
and
if
packages
are
you
know,
broken
state
for
a
long
time,
then
they
should
probably
be
removed.
B
Totally
I
think
similar
to
mentioned
that
it's
about
keeping
broken
packages
for
a
defined
list
time
then
dropping
them
once
it's
done.
So
it's
like
a
a
packaging
will
process
I.
Think
we
don't
mean
how
to
process.
Do
that?
Really
it's
not
really
define
the
documentation
about
how
to
remove
something.
Ethically
Florian
I,
see
your
hand
is
raised.
E
E
Don't
really
think,
there's
a
lot
of
reason
on
on
doing
a
lot
of
work
on
moving
broken
stuff
to
another
repository
of
some
sorts
if
it
doesn't
work
in
first
place,
I
agree
on
moving
some
packages,
which
are
not
really
open-source
out
of
mix
packages,
but
if
they
like
stuff
that
is
entirely
broken
and
has
been
for
years,
it
really
doesn't
make
sense
to
migrate
it
over
because
there's
nobody
taking
care
of
it
at
the
moment.
Yeah.
D
D
Yeah
I
think
we've
had
this
well
the
discussion
about
the
workflow
for
broken
packages
in
the
past
and
I.
Don't
think
we
ever
really
standardized
anything
but
I
think
the
idea
was
B
was
to
set
a
broken
method
or
broken
on
packages
and
then
in
the
next
release,
remove
all
the
broken
packages,
something
like
that
and
yeah
I
mean
that
could
be
automated,
but
yeah.
B
I
did
mention
on
discourse
that
it
could
definitely
be
some
sort
of
tasks
that
we
do
before
releasing
next
to
us
right
and
one
of
those
was
the
thing
about
metadata,
but
I
think
this
is
because
there's,
like
probably
like
a
least
room
I've,
been
traveling
around
looking
in
X
Packard
history
like
a
30%
amount
of
like
abandoned
offices
old.
That's
that
really
a
difficult
problem.
A
One
of
the
problems
historically
maintain
ership
is
how
do
you
know
if
somebody's
changing
your
package-
and
this
wasn't
a
problem
when
Nick's
packages
was
100
floor
requests
of
a
month
or
a
week,
but
now
that
we're
up
to
over
a
thousand
a
month,
I
think
it
was
too
much,
and
so
there
was
a
effort
to
pay,
maintain
ease
with
work
and
one
of
one
of
the
sort
of
sub
subtle
sub
goals.
There
was
to
try
and
make
maintainer
ship
mean
a
little
bit
more
and
I
wonder
what
might
be
next
for
that.
A
B
D
D
B
D
It
makes
access
control
harder,
the
yeah
not
just
a
mono
repo,
but
also
that
we
have
these
shared
files,
like
all
packages
that
NICs,
which
do
need
to
be
modifiable
by
everything.
So
code
ownership
doesn't
really
prevent
anybody
from
doing
arbitrary
stuff
I
mean
you
could
always
mess
with
all
packages.
Stuff
makes
example
exactly.
A
A
A
A
We're
so
right
now,
if
you
check
out,
if
you
have
a
channel
and
your
say,
you
depend
on
chromium
and
pretend
chromium
wasn't
built
and
required
to
pass
by
the
channel.
You
might
rebuild
your
system
and
then
be
stuck
on
potentially
days
long,
compile
of
chromium
only
to
figure
out
that
it
fails
and
that's
a
pretty
nasty
user
experience.
A
B
I
see
it
you're
saying
so
like
a
package
that
is
known
to
be
broken,
like
actually
maybe
be
implemented
inside
necks
of
us
like
a
package,
that's
known,
to
be
broken
in
like
environment
system
packages.
If
someone
runs
next
to
us
rebuild,
it
fails
saying
that
while
you're
dependence
is
broken
or
your
configuration
yeah.
A
D
So
the
problem
is
that
it
would
require
Hydra
to
be
perfect
in
Ida,
in
distinguishing
between
transients
and
permanent
failures
and,
and
that
historically
has
not
been
the
case.
So,
for
example,
in
the
case
of
chromium
right
often
fails
because
the
disk
is
full
or
it's
or
it
exceeded
some.
Some.
Some
time
limit.
A
D
E
D
A
A
B
For
example,
the
thing
we
had
a
problem
with
like
John
tone,
R
and
heading
for
her
gnome,
is
that
pretty
much
anything
that
comes
under
our
maintain
ership
I
want
to
get
notifications
on
changes
to
that
and
things
like
that.
But
it's
like
okay
after
we
go
through
every
single
package
and
next
packages
and
add
all
of
our
own
stuff
and
like
that,
and
what,
if
one
day
I
want
to
remove
myself
how
to
go
through
and
just
said
all
that
way.
So.
B
A
F
Yeah,
so
so
what
I
was
thinking
listening
to
this
conversation?
Right
now
is
that
I
don't
know
one
one
concern
I'm
kinky,
but
it's
that
it
all
sounds
very
technical
in
the
sense
that
yeah.
So
how
can
we
know?
How
can
we
flag
this
to
a
certain
maintain
our
care?
We
organize
groups
that
contain
hopes
of
maintainers
and
so
on
and
so
on.
This
is
for
me
like
too
technical
like
what's
the
easy
question
is,
and
then
I
mean
you
you
just
like
to
list
before
you
said
this
is
all
very
interconnected
right.
F
The
real
question
is
like
what
are
the
what
are
the
actual
bottlenecks,
and
you
know
what
what
has
to
be
what
has
to
be
done
in
order
for
me?
First
of
all,
actually
even
you
know
it
even
what
what
is
the
problem
right,
okay
and
then
what
has
to
happen
for
for
the
problem
to
get
better
like
is
the
problem
like
the
typical
and
of
the
typical
of
of
PRS
and
unreached
PRS,
and
so
on.
That
issues
is
the
problem.
F
Package
is
not
being
not
being
fixed
for
a
long
time
and
so
on
and
so
on,
and
then
even
before
we
said
you
know,
and
even
even
if
we
had
a
technical
solution,
a
perfect
technical
solution
for
flagging
maintenance
and
so
on
and
so
on.
What
is
the?
What?
What
does
it
actually
mean?
You
know
to
be
maintained
or
and
so
on
and
so
on.
So
so
these
very
technical
terms
of
how
to
organize
maintainers
on
packages
and
so
on
is
I.
A
A
F
So
so,
in
a
way,
I
feel,
like
the
technical
part,
has
to
follow
the
cultural
part.
So
we
can,
we
can't
really
necessarily
say
so.
These
are
the
technicalities
of
how
being
a
maintainer
works,
but
more,
like
maintenance
have
to
evolve
into
something
specific
and
and
then
we
can
adapt.
You
know
we
can
adapt
a
technical
part
to
that
new
way
of
how
maintain
SPF
I
don't
know
so,
maybe
also
it's.
Maybe
it's
maybe
a
bit
tougher
to
specify
in
order
for
someone
to
be
a
maintainer.
F
B
I'm
gonna
do
that
right
now,
I'm
gonna
get
the
screen
sharing
going.
Okay,
everyone
can
see
everything
yeah
everything
looks
right.
I
can't
reach
my
doc
yep,
okay,
I
guess
we'll
just
have
to
ignore
that
I
can't
use
my
doc
all
of
a
sudden.
Okay,
so
I've
been
working
with
much
sheets
to
package
cinnamon
for
a
couple
weeks
now
growing
merrily
lungs,
and
this
will
go
right
into
it.
B
B
B
So
first
thing
I
see
right
away,
is
to
do
here
for
a
network
manager,
integration
and
I
believe
that
has
to
do
with
something
with
like
the
go
object,
introspection
stuff
not
being
available
in
network
manager
something's
up,
because
that's
out
of
scope
for
the
particular
issue
and
there's
some
things
that
are
commented
here
at.
What's
really
shown.
C
So
far,
this
is
old
stuff
that
I
found
from
somebody
else
or
the
old
cinnamon
package
set
and
I,
don't
know
if
it's
still
needed.
I've
just
basically
left
it
in
so
we
know
those
packages
have
existed
and
we
can
reevaluate
whether
they
are
necessary
and
that
thing
you're
looking
at
right
now.
This
hacky
thing
is
because
the
test
was
failing
and
I
just
somehow
made
the
test
work
by
using
some
dark
magic,
and
now
it
works.
C
B
It
it's
probably
okay,
I
think
the
only
suggestion
I
have
here
is
the
use
of
like
pre-install
phases
and
stuff
like
that
and
like
little
custom
attributes,
I'm,
not
sure
I
feel
about
that.
Usually
it's
just
like
something
I
would
do
in
like
pre
build,
and
this
would
be
the
this
would
be
pre-installed
I,
don't
know
it's
not
really.
Some
sort
of
like
I
mean,
like
I,
can
say
a
style
that
I
see
a
lot.
B
So
in
the
install
file,
actually
the
overhangs
not
exactly
the
uncracked
rose,
but
in
the
install
phase
there's
a
hook,
pre
build
not
in
this
office
and
there's
a
pre
installed
hook
and
a
post
install
hook.
So
this
runs
before
the
install
phase
runs
and
after
but
I'm
overriding
would
be
like
putting
install
phase,
which
is
completely
different.
You
shouldn't
do
that
because
then
it
won't
run
those
hooks
anymore
because
he's
completely
overridden
what
was
in
and
setups
on
sage.
B
C
B
Exactly
so
I
guess,
looking
at
the
build
dependencies
is
kind
of
difficult,
so
we
can
probably
look
through
that
and
sort
of
match
that
this
is
a
lot
of
picky
little
stuff,
having
a
check
each
and
every
single
one,
and
sometimes
because
it's
next
to
us,
you
have
other
ones
that
are
not
exactly
on
direct
dependencies
that
need
to
be
in
building
puts
because
of
how
the
gnomes
setup
hook.
Infrastructure
works,
like
you'll,
need
certain
things
in
there
that
have
setting
schemas,
so
I
can
get
put
up
into
the
wrapper.
B
So
that's
a
lot
of
things,
so
you
also
you
even
just
look
at
like
other
distros
packaging
like
here's.
A
Debian
control
file
that
has
all
the
dependencies
but
like
the
most
true
way,
would
just
feel
look
at
the
actual,
build
expression
and
look
for
like
package
check
modules
and
each
one's
checking
make
sure
that's
the
dependency
in
there,
but
those
won't
be
on
negative
dependencies.
Those
will
be
on
building
puts
so
from
its
going
to
like
side
by
side.
Compare
these.
A
B
Concede
yo
eunuchs,
which
is
surprising.
It
need
the
see
flags
hoc
to
fix
that
I.
Guess
they
have
it
right
here
it
depends
on
slips.
Cinnamon,
Control
Center
and
the
colon
modules
which
are
up
here.
Gt
k3g,
live
g3
ideals
are
all
in
G,
live
and
cinnamon
desktop
and
I
see
those
you
have
notified
there
and
also
sets
what
required
versions
that
has
which
are
strict.
B
B
B
So
yeah,
if
you're
reviewing
something
that
uses
this
tilt
system,
you
sort
of
look
at
these
particular
keywords
to
see
what
the
dependencies
are.
They,
someone
put
checks
in
for
that,
otherwise
just
have
to
look
into
the
source
code
and
stuff
like
that,
which
is
the
pain.
I,
don't
see,
GDK
fix
both.
B
B
C
C
Yeah
configure
face
doesn't
find
the
library
and
its
network
major
applet
I,
believe
I
have
added
that,
but
it's
still
not
finding
that,
and
that
is
really
bad
I,
don't
know
if
it's
actually
breaking
anything
but
I
suppose
not
having
network
manager
in
control
center
means
all
network
manager.
Related
settings
won't
work
and.
B
They
probably
won't
be
there,
so
I
think
when
it
fails
like
that,
it's
because
there's
a
missing
required
library
in
the
package
config
file,
it's
not
present
in
the
current
environment
and
as
we'll
see
here
as
I
mentioned,
it
need
to
propagate
these
on
libraries,
because
it's
in
package
configs
requires.
So
if
you
could
PM
me,
the
exact
error
I
could
figure
out
why
you
can't
find
that
right
manager.
I
can't
remember,
was
an
issue
you
open
about
that
anywhere.
B
B
B
C
So
it's
my
laptop
right
now,
which
I'm
using
and
I've
got
cinnamon
installed
here.
I've
got
a
few
tweaks
and
settings
and
the
reason
that
works
is
because
I've
made
those
changes
on
Ubuntu.
So
if
I
open
the
control
center,
it
doesn't
open,
which
is
part
of
the
thing
we're
trying
to
fix
the
last
review.
C
The
actual
cinnamon
desktop
right
now
doesn't
look
very
polished
because
we
still
have
some
rough
edges.
So
if
I,
let
that
login,
we
get
a
very
untamed
and
just
pretty
basic
cinnamon
and
its
really
what's
also
missing
our
desktop
plugins.
So
my
desktop
isn't
actually
so
clean
because
if
my
desktop
clean
but
because
the
plugin
isn't
running
for
the
desktop
I
can
still
show
up
and
once
this
locks
in,
we
should
get
a
sim
desktop.
C
Yeah
I
think
it's
taking
its
sweet
side
but
notice
I'm,
trying
to
make
a
demo
because
it
worked
and
now
it
works.
So
what
we
also
have
trying
to
fix
is
8,000,
sometimes
to
take
it
out
for
acceleration.
On
my
laptop,
it
doesn't
show
the
message
that
my
video
drivers
are
broken,
but
I
do
these
are
messages
from
OpenGL
about
things
like
this
meta
sync
ring
not
working
exactly.
B
C
We
have
a
pretty
decent
desktop
and
you
know,
and
do
things
what's
also
very
interesting
to
note,
for
people
trying
to
pour
desktop
environments.
Is
you
can
always
just
launch
an
external
session
and
launch
your
desktop
environment
to
get
all
the
error
messages,
because
logging
in
on
the
first
try
after
parting,
never
works.