►
From YouTube: CentOS Automotive SIG office hours - March 17, 2022
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
A
Okay,
there,
it
is
the
recording,
so
so
yeah
there's
been
a
a
lot
going
on.
Maybe
we
can
just
start
out
if
we
want
to
have
pierre,
do
a
quick
readout
of
the
various
technical
things
that
have
been
going
on
over
the
last
two
weeks
since
our
last
formal
meeting,
I
want
to
put
you
on
the
spot.
B
Sure
sing,
I
guess
the
biggest
one
was
the
order.
We
already
spoke
about
sosd
at
the
last
michigan.
A
I
let
me
we
did
have
talked
about
auto
sd
and
I
can
even
bring
up
the
slides
if
you
like,.
B
Not
yet
not
the
slides
yet,
and
I've
shared
on
the
chat.
The
blog
post,
that
we've
published
on
the
blog.centers.org
about
toto
sd
and
the
idea
of
photo
sd
is
to
have
a
public
development
version
of
the
in.
B
That
he's
working
on
where
we
can
collaborate
and
then
the
automotive
sig
sit
on
the
top
of
the
auto
sdd
of
the
auto
ac
project
and
expand
upon
it.
So
if
we
want
to
do
new
hardware
enablement,
if
we
wanted
to
test
some
features
on
the
kernel,
let's
change
some
things
on.
I
have
no
idea,
you
know
system,
gcc
or
whatever.
B
We
can
do
that
as
part
of
the
sig,
without
impacting
the
the
program,
the
project,
the
product
that
three,
that
is
that
wants
to
work
on,
but
also
that's
also
a
way
for
the
sick
to
actually
contribute
back
to
the
to
the
project.
By
demonstrating
that
hey
look,
we
have
these
patches
on
the
kernel
and
we
have
made
them
and
it's
it
works.
It's
able
we
we
can
test
them,
we
can
work
upon
them
and
then
we
can
knock
on
the
red
hat,
store
and
say:
hey.
B
We
can
tell
you
that
it
is
desired,
so
you
know,
please
have
a
look
at
it
and
consider
it
for
considering
cleaning
it
into
your
product
and
being
as
close
to
the
product
as
possible
by
sitting
just
on
the
top
of
the
sd
would
give
us
that
that
possibility
of
showing
red
hat
what
can
be
done
with
it.
The
idea
is
that
auto
sd
would
be
the
equivalent
to
the
to
the
in
vehicle
os
of
what
santos
stream
is
to
rail.
B
So
it's
it's
very
much
the
public
in
in
development
version
of
the
product,
so
everything
that
is
that
you
know
everything
that
arrives
on
the
touristy
should
be
landing
in
the
product
down
the
line.
So
if
you
want
to
see
you
know
what
the
product
will
look
like
in
in
six
months,.
B
D
B
B
You
can
see
there
is.
B
I
like
doing
that
with
you.
There
is
a
product
built.
This
is
going
to
be
renamed
because
it
doesn't
mean
much,
but
the
the
product
build
is
basically
the
the
current
state
of
the
of
the
authority
project.
So
it's
a
repository
with
a
number
of
rpms.
B
You
have
a
latest
folder
with
a
cs9
and
then
the
the
folder
structure
is
being
reorganized
because
there
are
things
like
we.
We
pull
our
pm
from
different
sources.
You
know
the
different
channels
in
real
different
channels,
equivalent
channels
in
center
stream
and
that
information
does
not
bring
anything
to
total
sd.
So
it's
been
to
be
trimmed
down
a
little
bit.
We
don't
need
to
know
you
know
what
was
coming
from
upstream
code,
ready
builder
or
base
os,
so
it
will
be
a
little
bit
a
little
bit
simplified.
B
The
product
build
folder
is
going
to
be
renamed
as
well,
because
I
see
it's
not
a
product,
so
that
doesn't
make
sense.
But
it's
that's
the
idea
you
can
see
as
well
at
the
top
that
you
have
a
simple
image.
Folder
the
image
is,
this
is
also
going
to
be
reworked.
The
image
is
there.
B
Are
I
don't
know
they
are?
They
have
been
worked
on
by
the
the
pipeline
that
triggers
when
we
have
a
change
done
in
center
stream
and
that
updates
photo
sd
based
on
the
changes
that
are
happening
in
center
stream,
but
I
some
someone
for
example,
one
other
thing
that
is
missing
here.
Is
we
don't
know
how
all
these
images
are?
We
don't
know
when
you
know
what
they
correspond
to
so
this
is
also
going
to.
D
B
B
B
There
is
a
fun
fn
set
of
containers
that
interact
to
each
other
that
talk
to
each
other
via
some
ip
and
that
you
can
basically
build
an
image
with
the
container
pre-built
and
systemd
starting
the
container
on
boot
and
have
the
container
to
each
other
using
some
ip.
As
a
as
the
underlying
message
bus,
I
guess
we've
updated
the
documentation
quite
a
bit
as
well.
B
If
you
haven't
read
it
have
a
look
at
it
and
that's
for
the
most
visible
part
of
what
we've
been
doing,
we're
still
working
on
the
santos
infrastructure
side
about
enabling
you
know
a
nicer
workflow.
B
So
we're
working
the
leukocyte
cache
structure
which
would
allow
to
develo
a
more
flexible
git
structure
because
currently
the
when
you
upload
the
source
to
this
git
on
on
center's
side,
it
incorporates
the
git
branch
name
because
of
the
way
it's
currently
linked.
It's
currently
set
up,
but
by
moving
with
github,
we
can
drop
the
git
branch
structure
that
say
that
was
currently
imposed,
but
the
lucasi
cache
relies
on
this
structure.
B
So
we
are
working
on
changing
the
lookup
side,
cache
structure
so
that
we
can
drop
the
git
branch
requirements
on
on
the
good
side
so
that
it
should
just
make
our
life
easier
than
the
line.
That's
that's
the
case
of
it
dropping
dropping
hardcoded,
structured
and
decisions
to
for
a
simpler
for
simpler
workflow.
B
I
don't
know
if
there
are
any
questions.
Yeah,
the
doc
has
been
updated.
If
we've
liked
the
you
know,
we
start
by
presenting
the
different
artifacts
that
the
sig
is
is
working
on
auto.
D
C
I
did
I'd,
be
cure
or
encourage
people.
I
know
most
people
in
red
hat
are
probably
familiar
with
this,
but
if
you
don't
know
what
pierre
means
by
it,
this
has
a
similar
workflow
to
centos
stream
only
for
the
automotive
product
to
explore
centos
stream
a
little
bit
more.
It
really
is
a,
in
my
opinion,
a
great
improvement
that
red
hat
has
created,
starting
with
real
nine
and
centos
stream,
nine
and
we're
trying
to
piggyback.
On
top
of
that,
so
it
really
does
show.
B
I'm
also
happy
to
take
any
questions
I
may
have
you
know
thrown
on
a
number
of
terminology
terms.
Words
that
may
not
be.
You
know
that
you
might
not
be
familiar
with
I'm
happy
to
explain
what
the
leukocyte
cache
is.
What
the
disk
structure
is,
how
one
relates
to
the
other,
it's
a
self-interest.
A
E
A
Well,
there's
been
a
number
of
different
things
that
we've
discussed
over
the
last
couple
of
months.
I
haven't
gotten
any
responses
to
that
query
in
particular,
but
things
that
have
come
up
include
things
like
the
nvidia
jetson
there's
a
hardware
is
always
of
a
lively
discussion
whenever
it
comes
up,
so
we
could
certainly
talk
about
some
of
the
hardware
options
out
there
right
now,
since
this
is
a
an
open
source
project,
you
know
anybody.
A
Admitting
somebody
we
welcome
any
hardware
enablement
that
anybody
would
like
to
do.
A
Nothing
discussed
already,
nothing
that
had
you
know
nothing,
nothing
new.
You
know,
there's
a
certain
boards
that
have
have
repeatedly
come
up
like
the
jets.
A
Just
so
everybody
knows
the
initial
reason
that
we
chose
the
raspberry
pi,
as
the
initial
target
is
simply
because
of
its
ubiquitous
availability,
that
everybody
has
access
to
one,
or
at
least
they
did
before
the
chip
shortage
started.
So
a
lot
of
people
end
up
having
a
raspberry
pi
on
their
desk
if
they
want
to
play
with
real
hardware.
A
A
The
question
going
on
on
the
chat
about
the
the
raspberry
pi
4.,
I'm
in
canada,
and
I
have
been
unable
to
get
a
raspberry
pi
for
for
the
last
year.
I
think
eight
or
ten
months
that
I've
been
looking.
F
Yeah
so
availability
is
your
mileage
may
vary
a
relative
capacity
at
best.
C
A
Yeah
I
was
looking
on
my
shelf
for
all
of
the
really
ancient
hardware.
I
think
I've
got
an
original
beagle
board
and
probably,
I
think,
a
a
dragon
board-
a
qualcomm
dragonboard
14c
about
eight
years
ago
six
years
ago,
so
we'll
just
have
to
start
digging
things
off
the
shelf
and
putting
them
back
into
production
right
now.
Unfortunately,.
A
So
is
anybody
doing
any
works
on
on
these
images
or
on
the
within
the
sig
that
they
would
like
to
have?
They
have
challenges
with
like
to
bring
up?
We've
got
a
lot
of
the
a
lot
of
the
maintainers
are
here
on
the
call
so.
E
G
So
this
is
philip
talking,
you
know,
at
least
on
our
side.
We,
we
don't
have
any
difficulty
with
hardware.
We
have
a
tensorboard.
G
One
thing
issue
is
that
we
still
have
a
very
recent
project
where
people
are
using
very
low
end
hardware,
just
to
give
you
an
idea.
We've
been
talking
with
people
working
on
a
on
a
camera
to
detect
if
the
driver
is
really
looking
the
road-
and
you
know
when
you
are
under
autopilot
and
the
goal
for
the
full
system
is
30
euro
okay,
which
includes
the
hardware,
the
cameras,
the
softwares,
everything,
okay,.
D
G
It's
not
going
to
have
eight
cores,
it's
not
going
to
have
eight
gig
of
ram.
Okay,
it's
going
to
be
more
something
like
one
core
with
multiple
microcontroller,
maybe
a
dsp
fpga
whatever
to
do
the
work
and
for
those
type
of
board
linux
is
still
necessary
because
you,
you
need
to
have
some
high
hand
functionality.
G
You
need
to
talk
with
the
rest
of
the
car
with
the
communication
gateway.
Typically,
you
need
to
have
cyber
security
built
in
and
typically
is
that
what
micro
controller
cannot
do
so
they
need
linux
for
that.
But
fitting
linux
in
such
an
architecture
is
a
challenge
by
itself,
and
this
is
sending
me
to
so
so
I
would
say-
hopefully
you
know
stefan
he's
not
in
the
call
he
was
supposed
to
be,
but
probably
he's
locking.
G
Something
else
is
due
to
come
with
some
boards
that
we
could
support
and
maintain
directly
at
iot
busy
days
to
have
an
image
for
zoosport.
We
still
have.
I
still
have
an
issue
and
I'm
not
sure
how
we're
going
to
deal
with
that,
which
is
how
we're
going
to
maintain
the
specific
part
of
a
board
board
support
package
and
a
generic
kernel.
Okay
in
yocto.
Typically,
we
have
layers,
and
so
you
may
have,
like
you
know,
a
standard
core
for
the
core
functionality
of
your
kernel.
G
And
then
you
have
you
know
some
external
file
where
you
have
the
specificity
of
your
hardware.
We
don't
have
that
in
a
in
current
kernel,
config,
where
more
or
less
we
have
one
monolithic
configs
that
include
everything.
G
So
one
thing
we
are
asking
here
at
iot
business
is
whether
we
could
have
something
somehow
similar
where
we
could
have
on
one
under
configurations
that
define
the
generic
kernel
and
to
have
some
complements
that
we
we
built
or
we
had
or
we
overload.
Whatever
is
the
term.
We
want
to
use
to
complement
this
kernel
to
have
some
specific
for
a
given
board
without
this
mechanism,
it's
very
hard
to
keep
a
generic
kernel
in
place
because
everyone
is
willing
to
shift
to
a
new
one
so
that
that's
my
first
concern.
G
The
second
question:
I
think
that
this
sig
has
not
answered
yet
is
what
is
going
to
be
different
in
between
the
linux
ship
by
this
sig
and
a
standard
linux
ship
by
by
santos,
for
a
specific
board.
Okay,
if
we
ignore,
you
know
the
notion
of
board
enablement,
what
are
the
specific
functionality
and
typically
in
the
in
the
automotive
sector,
we
are
asked.
G
We
are
asked
for
skim
down
of
the
kernel,
because
people
want
to
boot
in
a
few
seconds
and
on
a
very
small
hardware,
we
are
asked
for
built-in
security
with
things
like
dm
verity.
You
know
activated,
but
also
read-only
file
system.
We
are
asked
for
some
cryptic
crypto
file
system
for
all
the
user
data
people
want
to
everything
and-
and
that's
something
that
is
not
that
easy
to
make
okay
is.
G
There
is
a
lot
of
know-how
before
you
arrive
to
to
build
a
system
where
your
your
root
file
system
is
read-only,
so
this
clearly
would
be
a
difference
with
the
standard
linux,
as
santos
can
ship
it,
and
the
other
element
that
is
also
we
see
specific
in
the
automotive
sector
is
all
the
capability
to
drive
the
specific
devices,
especially
the
can.
G
Okay,
we've
done
a
lot
of
work
for
automotive
grade
linux
to
to
implement
modularity
and
some
you
know
some
way
to
access
scan
from
from
automotive
description
of
the
message,
and
so
my
question
is
where
this
group
is
willing
to
go.
Is
it?
G
Is
it
willing
to
focus
on
the
very
low
level,
providing
the
kernel
and
board
enablement
only
or
you
also
look
at
providing
some
higher
functionality,
either
on
the
security
side
or
on
the
functional
side,
like
the
cans
introspection,
some
ipc
mechanism,
whatever
we
we
may
need
to
have
in
in
automotive
sector,.
C
I
can
try
and
provide
an
initial
response.
I
think
our
from
the
red
hat
perspective,
our
effort
so
far
has
been
towards
moving
to
enough
standardization
of
the
hardware
and
enough
upstreaming
of
the
enablement
that
we
can
continue
to
have
a
kernel
model
similar
to
what
we
do
with
rel,
where
we
are
able
to
have
a
shared
kernel
and
the
enablement
is
done
in
the
upstream.
C
Much
of
it
is
modularized.
You
use
the
modules
that
you
need.
I
do
think
it
would
be
interesting
to
talk
about.
Are
there
are
there
models
short
of
that?
Are
there
integration
paths
like
or
sort
of
interim
steps
along
the
way
to
that
where,
for
example,
you
could
use
the
images
and
some
of
the
output
that's
coming
from
this
sig,
but
then
swap
in
something
that
looks
like
a
bsp
in
the
way
that
you
can
with
yakto.
I
don't
think
that's
really
well
fleshed
out.
C
I
haven't
seen
it
well
fleshed
out
in
the
in
the
rel
or
fedora
or
centos
space.
We
say
what,
if
I
want
what,
if
I
do,
need
to
bring
in
entirely
distinct
kernel
rpm
in
order
to
produce
an
image
that
can
boot.
I
think
that
could
be
interesting
and
we
could
maybe
there
is
that
I
suppose
so
that's
one
one
potential
discussion.
What
would
that
be
useful
and
if
so,
which
platform?
Is
there
a
sort
of
reference
platform?
C
C
G
G
G
But
I
I
I
think
it's
a
it's
a
it's
a
fair
point.
We
try
to
convince
industry
that
they're
not
going
to
be
in
a
position
to
maintain
one
kernel
for
board
for
20
years:
okay,
because
two
days,
that's
what
they
do.
C
Yeah,
and
also
I
mean
these-
there's
a
pretty
important
delineation
between
because
there
are
models
for
you
know.
I
think
rpm
fusion
has
examples
of
these
there's
a
mod.
There
are
models
for
delivering
out
of
tree
drivers
if
they
can
be
delivered
as
modules
in
a
way.
That's
purely
additive,
I'm
sure
you're
almost
certainly
familiar
with
this,
but
then
I
gather,
I
suspect,
in
among
the
600
patches
that
you're
talking
about
in
this
case
are
things
that
are
not
can't
just
be
built
as
modules.
G
H
H
No,
I
just
wanted
to
add
that
yeah,
for
example,
on
top
of
the
of
the
settings
kernel,
you
have
something
like
800
patches
and
if
you
look
in
the
in
those
patches
you
you
see
that
there
is
a
big
part
which
are
drivers
which
did
enter
in
the
in
the
mainline
kernel.
For
many
reasons
I
would
say,
the
quality
of
the
code
is
one
reason
and
or
maybe
for
proprietary
reasons
too,
because
they
don't
want
to
upstream
some
some
kind
of
firmware
which
is
distributed
to
special
customers.
H
So
there
are
many
reasons
so,
but
quality
is
one
reason
and
you
see
other
patches,
which
are,
I
would
say,
acts
on
top
of
the
kernel
that
wouldn't
be
accepted,
mainline
just
to
make
those
crappy
drivers
work.
H
So
so,
at
the
end,
most
the
the
biggest
part
of
the
patch
is
relate
to
to
the
drivers
somehow,
but
not
on
the
on
the
generic
features
of
the
kernel.
H
H
What
what
happens
with,
for
example,
free
scale,
is
a
good
example
of
freaks
can
now
nxp
they
there
are
times
in
the
past
to
to
push
their
patches
to
the
kernel.
So
at
some
point,
when
you
were
building
some
yucca
stuff
with
nxp
or
with
freescale,
it
was
horrible.
It
was
a
very
old
kernel
with
a
lot
of
patches
and
they
made.
H
I
think
that,
when
they
they
were
bought
by
nxp,
they
made
some
effort
to
to
to
get
a
good
level
of
quality
in
the
code
and
and
finally
they
they
pushed
a
lot
of
drivers
in
the
men
like
kern
kernel.
So
so
now,
if
you,
if
we
talk
about
imx6
or
imx8,
we
are
quite
at
the
same
level
as
as
renaissance.
H
Typically
or
I
wouldn't
say,
intel
intel
is
the
best
the
best
citizen
in
this
in
this
area,
because
quite
everything
is
upstream
then
renaissance
and
nxp
are
also
good
citizens
because
they
push
most
of
the
codes
they
achieve
that
challenge,
and
there
are
other
ones
like
xilinx
or
specifically,
for
for
very
new
platforms.
C
Yeah,
as
I
say,
all
of
us
who
have
who
have
sort
of
landed
in
this
and
are
seeing
it
in
this
level
of
detail
for
the
first
time
understand
that
the
any
new
any
new
piece
of
hardware
right
now
that
arrives
here,
if
it
runs
linux
at
all,
almost
certainly
has
some
element
of
what
you've
described,
and
so
I
think
path.
One
for
that
is
bsps
are
then
generally
associated
with
yakto
yakto's
built
to
allow
this
to
happen,
fedora
and
rel
and
centos
generally.
C
It's
not
that
they
aren't
it's
just
they
don't
facilitate
it.
The
way
that
yakdo
does-
and
I
feel
like
ubuntu
has
landed
somewhere
in
the
middle,
where
they
make
it
easier
to
generate
an
alternative
kernel
to
install
with
things
so,
and
I
don't
think,
there's
no
reason
we
couldn't
do
that.
We,
it
has
been
done
in
the
fedora
space
repeatedly.
C
It's
just
people
kind
of
hold
their
nose
when
they're
doing
it,
whereas
if
we
could
present
it
as
a
step
along
the
way
towards
up
streaming,
for
example,
that
would
be
better,
which
is
certainly
the
path
that
we
can
intend
to
continue.
Taking
on
the
red
hat
side,
it's
what
we've
done
with
the
enterprise
product,
there's
with
the
one
notable
exception
of
the
very
famous
out
of
tree
preempt
rt
patches,
there's
nothing!
That's
ever
shipped
in
a
rel
kernel
that
isn't
upstream.
G
I
just
take
one
example,
which
is
the
advanced
ip
processing
driver
from
qualcomm,
okay,
that
is
never
going
to
enter
in
the
kernel
that
is
quite
widely
used
in
in
the
automotive
industry,
because
people
want
to
have
the
firewall
running
in
the
dsp
and
not
in
the
kernel.
G
This
also
allows
you
to
continue
having
a
tcp
stack
when
you
stop
linux,
okay,
which
is
a
sometime
very,
very
useful,
but
I
agree
with
your
point:
nothing
prevent
us
from
doing
something
that
could
you
know,
mitigate
both
requirement
and
and
find
an
intermediary
step
before
we
arrive
to
more
standardized
hardware,
which
is
not
going
to
happen
in
the
next
five
years.
Okay,
it's
going
to
be
longer
than
that.
G
G
That
is
completely
independent
of
the
hardware,
and
so,
if
we
could
have
some
nice
model
to
do
that
separation,
it
would
be
very
useful,
yeah
and-
and
we
don't
have
to
go
as
far
as
yocto,
because
yokto
is
a
nightmare.
People
can
change
any
option
at
any
time,
which
is
make
your
life
impossible,
but
we
need
some
some
way
to
do.
You
know
a
minimum
of
customization.
I
think
that
would
be
coming
back
to
what
would
be
unique
to
this
automotive
centos
distribution.
G
This
could
be
one
of
the
point.
Okay.
Today
there
is
no
red
hat
santos
fedora
distribution,
where
people
can
tweak
the
kernel
without
breaking
everything.
Okay,.
B
Yeah
yeah
sorry,
I
was,
I
just
wanted
to
say
that
nothing
prevents
the
sig
from
shipping
out
of
three
patches
in
the
kernel.
It's
gonna
be
a
different
canon
than
the
one
that
lands
on
auto
sd,
but
we
also
have
rpm
repository
on
that
is
available
to
the
six.
So,
if
we
as
a
collective,
we
want
to
create
a
kernel
that
you
know
raspberry
by
six
or
some
of
the
changes
that
philip,
which
is
probably
more
pragmatic
than
the
the
basic
that
doesn't
exist.
Yet
that's
something
that
we
can
do.
B
B
G
G
What
we
don't
have
today
and
and
what
we
provide
with
to
them
pre-built
is
all
the
binary
for
the
packages,
which
is
already
a
lot
of
times
that
they
can
save
so
difficulty
when
you
you
do,
that
is
to
make
sure
that
you
test
automatically
what
has
been
built,
because,
obviously
you
want
to
know
if
it
works
or
not
and
the
more
you
build
version
and
flavor
of
your
kernel,
the
more
you
have
to
run
test,
which
means
you
need
to
have
a
lot
of
hardware
and
a
lot
of
cpu
to
do
all
that
work.
G
We
would
love
to
have
an
intermediary
solution
with
some
kind
of
reference
configuration
for
the
kernel
and
not
tell
people
you
can
do
whatever
you
want,
but
you
can
do
modification
on
something
we
have
tested.
I
think
that
would
be
a
big
plus,
because
people
do
a
lot
of
mistake
by
making
their
own
stuff
and
very
often
they
don't
need
that
many
modification.
Okay,
they
pretend
they
need
a
lot
of
modifications.
They
need
something
that
is
100
specific,
but
it's
not
true.
G
Most
of
the
time
they
could
use
something
that
is
90
standard
and
have
only
10
customization,
and
today
we
don't
have
that
mechanism,
it's
either.
They
use
a
standard
kernel
we
ship
or
they
use
some
something
they
build
themselves
and
we
don't
have
any
intermediary
and
I'm
not
aware
of
any
distributions
that
have
this
model
to
have
intermediary
patches
on
on
ongoing
kernel,
outside
of,
as
we
said,
of
the
module
and
building
the
module
out
of
tree,
which
is
not
something
that
the
embedded
community
is
ready
to
accept.
G
G
G
Yeah,
it
is
true,
that's
the
way
it's
just
shipped
by
yokto,
that's
all
for
the
good
news,
but
that's
true.
G
It's
I
could.
C
The
I'm
just
trying
to
think
the
analog
for
us
would
be
to
demonstrate
a
way
that
you
could
maintain
an
alternate
kernel
package
build,
and
we
have
ways
of
doing
this
so
like
if
you
look
at
in,
in
fact,
so
what
you're
talking
about
is
something
a
set
of
patches
that
are
invasive
enough,
that
they
need
the
primary
kernel.
Binary
is
different.
It
can't
be
the
same
monolithic
binary
and
the
analog
we
have
for
that
in
centos
stream.
Right
now,
which
is
related
to
rel,
is
the
is
what
we
do
with
rt.
C
If
we,
if,
instead
of
rt,
it
is
renesas
ranasis's
reference
platform,
the
starting
point
would
probably
be
something
like
that,
where
you
have
a
either
the
upstream
as
the
thing
that
you're
basing
against
or
the
auto
sd
kernel
as
something
that
you're
rebasing
against,
and
then
you
have
patches
in
another
tree
a
fork
that
you're
constantly
making
sure
reapply.
C
And
then,
if
you
do
that
and
if
that
works,
then
you
could
probably
then
mostly
copy
the
rpm
spec
file.
That
turns
that
into
an
rpm
that
can
then
be
used
with
the
image
building,
tooling
and
everything
else.
So
I
think
what
you're
describing
boils
down
to
explain
how
the
concepts
of
a
bsp
map
from
yakdo
to
auto
sd-
and
I
think
that's
the
rough
sketch
of
how
it
would
work-
and
I
think
the
way
to
do
it
to
flesh
it
out
further
would
be
to
decide
on
something
that
we
would
build.
C
This
way
find
something
that
has
a
bsp.
That
is
a
desirable
hardware
platform
to
support
and
actually
do
this,
create
the
sourcetree
extract.
The
patches
apply
them
on
top
of
whatever
and
turn
them
into
an
alternative
kernel.
That
can
run
the
hardware,
so
we
just
need
to
choose
a
piece
of
hardware
and
do
it.
C
G
G
Most
of
the
development
board,
you
need
to
sign
an
nda
before
receiving
the
board
okay.
So
it's
it's
not
going
to
fly
but
yeah.
I
agree
finding
what
is
going
to
be
the
good
reference
board
for
end
of
this
year
and
next
year.
C
Yeah,
I
think
this
is
something
we
could.
We
could
chew
on
a
little
bit,
I'm
looking
at
trying
to
read
francisco's
reaction
to
this
and
see,
if
there's
anyone
else
from
colonel
in
harvard,
but
if
it's
something
we
can
buy
without
an
nda
that
has
huge
appeal
and
you're
saying
both
of
these
do
not
they're
not
supported
fully
by
upstream.
Both
platforms
would
deliver
a
bsp.
G
One
thing
that
is
typically
never
supported
by
upstream
is
the
interaction
with
the
microcontroller
you
have
on
the
board.
Okay,
most
of
the
harm
board
have
a
microcontroller
either
m5,
m4
or
r7
internally.
Here
we
run
zephyr
on
those
microcontrollers,
some
other
people
around
some
other
stack,
but
the
interaction
with
that
microcontroller
is
typically
never
a
part
of
the
of
the
standard
linux
distribution,
because
it
rely
on
specific
hardware
mailbox
internally.
G
G
We
could
also
take
a
vision
board.
There
are
some
development
board
for
vision,
especially
in
the
adas
world,
but
they
might
be
too
expensive.
I
think
we
also
probably
have
to
limit
the
cost
of
the
board.
I
think
less
than
one
thousand
euro
is
probably
one
thousand
euro
is
probably
a
maximum
for
for
development
board.
C
I
It
does
I
brought
it
up
earlier
and
it
was
considered
to
be
too
slow
of
a
cpu
for
some
of
the
work
that
we
were
looking
at
it
back
then,
but.
G
G
C
G
Think
it's
also
important
to
prove
the
concept
with
both
people
using
development
and
demos,
because
if
we
don't
take
the
board
people
using
demos-
I
mean
you
know.
One
thing
I
think
is
very
easy
to
understand
is
no
eom
or
vendors
is
going
to
do
a
demo
at
cs
with
a
raspberry
pi.
Okay,
that's
not
going
to
happen.
Okay
university
may
do
some
demo
and
some
tests,
but
are
not
going
to
do
that
so
they're
going
to
come
with
some
more
professional
board,
renesas
and
xp
qualcomm
zillings,
and
so
on.
G
G
So
that's
not
an
issue.
Okay,
the
issue
is
really
to
run
and
answer
the
strongest
market
today,
for
linux
in
a
car
is
on
the
controller
gateway
on
the
you
know,
on
the
li
on
the
leader,
you
say
lidar,
or
how
do
you
pronounce
that
in
english,
lidar,
okay,
so
lidar?
Typically,
they
have
a
real-time
operating
system
to
detect
all
the
point,
but
they
have
a
linux
to
put
some
name
on
the
object.
G
G
G
H
We
can
push
something
I
was
replying
to
stephen.
In
fact,
if
you
look
at
at
the
renaissance
models
you
you,
you
will
see
that
there
is.
There
are
plenty
of,
let's
say
inaudible,
because
they
have
a
let's
say
a
generic
one.
M3
was
one,
h3
is
another
one,
and
then
they
have
some
kind
of
specialized
chips
for
different
applications
in
the
car,
typically
ivi
or
ibi
will
use,
I
would
say,
a
middle
range
soc,
with
graphics
and
multimedia
capabilities.
H
You
will
have
less
cores,
maybe
less
multimedia
capabilities,
but
still
you
can
have
a
cam
controller,
or
so
you
have
common
ips
in
each
soc,
and
you
have
specific
ones,
and
the
one
pointed
by
by
stephen
is,
for
example,
for
adas,
typically,
because
they
are
they
probably
added
an
npu
for
analyzing
or
to
make
some
kind
of
deep
learning
about.
H
I
don't
know
what
the
cameras
will
return
or
whatever
stream
they
want
to
analyze
so,
and
then
renaissance
is
is
doing
like
this,
because
I
want
to
optimize
the
prices
probably
and
they
pay
for
external
ips
to
other
other
companies.
H
So
it
gets
also
even
more
difficult
to
support
these
in
the
kernel,
because
you
see
that
there
is
a
huge
range
of
features
to
support
and
you
can
you
in
fact
you
you
can
use
a
generic
renaissance
kernel
to
say
yeah,
I'm
able
to
boot
the
board,
but
at
some
point
you
will
have
to
do
something
specific
to
really
use
that
npu
or
that
gpu
or
whatever
ip
specific
you
have
in
the
in
the
soc.
So
that's
a
challenge
for
production,
but
maybe
not
for
for
this
sig.
In
fact,.
I
Thank
you
that
helped
me
I
I
had
been
looking
looking
for
I'd,
seen
a
couple
of
demos
and
I
was
trying
to
find
the
board
that
were
being
used,
but
one
of
the
issues
was.
The
demos
were,
of
course,
pre-covered
and
the
boards
were
now.
H
I
don't
want
to
disclose
anything,
but
what
I
know
is
that
the
the,
for
example,
the
the
next
generation
success,
are
out
now
or
this
quarter,
but
the
board
yeah.
But
the
boats
are
not
there
yet
yeah.
H
Raspberry
pi
because
the
socs
are
there,
but
you
know
that
you're
missing
some
we're.
G
G
Sure,
but
to
come
back
to
the
discussion
we
had
before,
if
this
sick
could
help
finding
a
solution
on
maintaining
you
know
customization
on
the
kernel
for
20
years,
which
is
the
requirement
from
the
industry,
it
would
be
a
big
plus.
Okay,
I
don't
have
the
answer,
but
clearly
that's
a
big
issue,
and
you
know
every
people
we're
talking
to.
Even
when
we
say
you
know
we
can
propose
to
maintain
the
kernel
for
10
years.
They're
like
this
is
not
enough.
Okay,
we
won
15
or
20
years.
G
So
clearly
we
have
a
lack
on
the
technology.
Today
you
know
to
tell
people
you
know
we
can
take
your
specific
part,
try
to
minimize
that
to
isolate
this
specific
part,
and
we
can
glue
that
to
something
that
is
more
standard.
That
has
a
an
official
maintenance
for
a
long
run
and
we
don't
have
the
right
tools
today
for
that
and
yes,
this
would
be
clearly
something
we
would
love
to
have.
I
I
don't
have
the
magic
answer.
Okay,
let's
be
clear,
I
don't
have
a
magic
wand,
but
it's
a
big
issue.
I
G
That's
the
reason
why
they're
looking
for
20
years
and
and
everyone
understands
they're
not
going
to
do
some
functional
update,
we're
talking
about
you,
know
important
or
critical
security
vulnerabilities
you
want
to
fix,
but
this
still
means
that
you
should
be
able
to
do
some
modification
on
existing
system
for
a
very
long
time-
and
I
agree
with
you:
okay,
20
years,
it's
probably
a
dream,
but
even
10
years
is
a
big
issue
and,
and
today
what
they
do
is
they
have
one
specific
kernel
for
every
system
they
ship,
which
is
not
going
to
fly.
A
G
Yeah
we,
we
are
part
of
that
project,
but,
honestly
I
I
don't
think
it's
going
to
fly
in
the
in
the
automotive
sector.
A
G
I
If
they're
going
to
five,
what
is
it
a
cent?
Is
it
sent
in
euro
yeah,
because
when
I
dealt
with
this
back
in
the
90s,
if
you,
if
you
could
save
four
cents
on
a
chip,
you
went
with
that
chip
on
a
motherboard,
whether
yeah.
G
H
I
A
Well,
we're
at
the
top
of
the
hour,
love
to
you
know
preserve
as
much
of
this
conversation
as
possible.
I
will
take.
I
have
my
own
notes
that
I
will
keep
and
anybody
if
you
want
to,
please
feel
free
to
add
some
hardware
ideas.
I
just
posted
the
the
page.
I've
set
up
a
wiki
page
for
hardware
ideas
if
it's
okay
with
you
folks
at
if
with
the
philip
and
stuff
on
I'd,
be
glad
to
add
the
hardware
that
you
mentioned
and
just
say
you
know
this
is
just
as
a
starting
point.
A
D
A
I
think
that
we
should
it'd
be
great
to
mention
the
things
that
are
going
on
and
to
call
them
out
as
but
make
sure
that
the
status
on
that
page
stays
up
to
date.
Otherwise,
if
he
mentions
it,
you
know
something's
being
worked
on
and
and
that
work
completes,
but
the
page
never
so
yeah.
I
think
this
is
a
great
place
to
do
that.