►
From YouTube: OMR Architecture Meeting 20220303
Description
Agenda:
* Linux core dump code (#6300) [ @davidjmccann ]
* CI pipeline tools (#2254) [ @babsingh ]
A
B
Oh,
thank
you,
hello,
everyone.
So
the
reason
I've
raised
this
issue
is
because
I'm
looking
at
using
the
the
core
dumps
aspect
of
the
code
in
a
containerized
environment-
and
this
provides
different
challenges
than
perhaps
would
be
normally
had
in
as
of
an
on-prem
kind
of
system.
So
there's
always
been
the
issue
of
if
you're,
if
you're,
forking
the
code
and
then
aborting
you're,
subject
to
the
os
restrictions
and
what
happens
with
the
core
dumps.
B
But
it
becomes
a
little
bit
harder
when
you're
in
a
container
environment
and
you're
no
longer
in
control
of
the
host
environment.
So
much,
and
so
you
tend
to
be
subjected
to
whatever
is
set
up
on
the
host,
and
you
can't
necessarily
change
it
in
the
container
environment
as
part
of
your
setup.
And
so
therefore,
if
you
want
to
write
the
core
dump
to
a
particular
location
on
a
so
some
of
the
stories
that
you've
been
allocated.
B
So
the
term
escapes
me
persistent
storage,
if
you,
if
you
wanted
to
write
it
there,
but
the
on
the
host,
there's
some
sort
of
redirection
of
the
core
dumps.
Then
then
it's
gonna
get
sucked
into
the
host
and
not
to
the
persistent
storage
where
you
want
it
to
go
so
so.
This
becomes
a
a
bit
of
an
issue
sort
of
working
in
the
container
sort
of
environments,
and
so
I
was
sort
of
trying
to
investigate
also
to
raise
awareness
of
what
we
could
do
about
this.
B
And
this
was
sort
of
raised
before
by
by
someone
who
raised
against
podman
whether
it
would
be
possible
to
to
override
the
core
pattern.
For
example,
when
you
start
the
container
that
was
sort
of
rejected
as
being
more
of
a
kernel
issue,
but
you'd
have
to
have
a
name
spacing
issue
on
on
the
core
pattern
in
the
kernel.
B
What
I
was
suggesting
was
using
a
mechanism
similar
to
use.
It
was
used
previously
in
things
like
the
google
core
dumper
code
or,
as
as
the
omr
already
has
in
the
on
the
mac
code,
which
is
to
generate
the
core
dump
itself
rather
than
using
the
os
to
do
it
for
it,
in
which
case
it
wouldn't
be
subjected
to
the
same,
wouldn't
be
reset
to
these
restrictions
that
the
os
is
putting
on
you,
and
so,
therefore,
you
could
get
it
to
go
out
to
where
you
wanted
it
to
go.
B
So
it
that's
kind
of
what
I
was
wondering,
what
it
would
be
possible.
The
google
core
dumper
code,
you
know,
was
written
eight
years
ago,
hasn't
been
kept
up
to
date,
no
longer
compiles
on
on
all
the
platforms
that
sort
of
assemble
a
code
that
doesn't
seem
to
work
anymore.
So
I
don't
know
how
this
code
will
be
written
or
what
would
be
required.
A
Okay,
thanks
for
introducing
that
david,
I
don't
know
either
maybe
keith
or
baby.
I
know
you've
got
a
lot
more
experience
in
this
area
of
the
code
than
many
of
us
do.
Do
you
happen
to
have
any?
Do
you
understand
the
problem,
and
do
you
happen
to
have
any
insight
as
to
whether
or
not
omar
could
adapt
to
provide
what
he
needs.
C
C
B
B
C
C
C
D
A
E
E
C
The
deal-
I'm
not
not
sure,
but
my
my
estimate
or
my
expectation
of
the
container
influence
on
this
is
that
the
file
may
be.
You
know
it's
either
in
a
location
that
is
visible
only
within
the
container
or
it's
only
in
it's
in
a
location,
that's
visible
only
from
outside
the
container,
and
so
both
of
those
present.
You
know
they
present
different
kinds
of
problems,
and
so,
if
it's
outside
the
container,
the
application
won't
be
able
to
see
it.
A
Okay,
I
I
think
that
we
should,
I
mean,
have
a
little
bit
of
a
discussion
still
in
the
issue
itself
to
sort
of
document.
What
we've,
what
we've
discussed
here
and
and
it's
kind
of
the
the
the
solution
that
we
that
we
think
is
going
to
work
going
forward,
so
it
it.
However,
the
unfortunate
thing
is:
it
doesn't
sound
like
you'll
have
a
solution
david
any
time
in
the
short
term.
This
is
going
to
be
a
a
bit
of
a
longer
term
item
that
that
may
get
done.
B
Effectively
configuration
for
a
single
kubernetes
node,
for
example,
and
which
can
be
done
it's
just
how
how
much
you
want
people
to
go
through
hoops,
and
you
know
to
get
that
capability,
and
so
it's
what's.
There
is
something
to
start
with
and
then
moving
forward
if
it
can
be
made
more
fit
for
purpose
in
the
container
environment.
That
would
be
great.
A
Okay,
all
right,
so
I
mean
certainly
appreciate
you
raising
the
issue
and
bringing
this
to
our
attention.
So
this
is
maybe
something
that
we
can
make
some
progress
on
in
the
in
the
near
future.
So
we'll
summarize,
the
the
discussion
today
in
the
issue
any
other
discussion
around
this
topic
that
anybody
wants
to
have
bring
up.
A
A
E
E
E
So
currently
we
do
these
checks
manually
and
recently.
I've
noticed
that
a
lot
of
copyrights
are
missing
and
if
there
is
oversight,
even
the
dates
can
become
inaccurate,
so
the
manual
checks
are
difficult
because
it
takes
a
lot
of
reviewer
time
and
also
contributors
time
so
and
it
becomes
infeasible
when
there
are
a
large
number
of
files
in
the
pr.
E
A
F
E
A
Well,
I
mean,
I
think
we
have-
I
mean,
there's
there's
working,
I
mean
sorry
from
from
the
history
of
this.
I
mean
there's
working
code
in
the
openj9
project
right
they
have
a
copyright
checker
that
works
reasonably
well
and
probably
does
pretty
much
what
we
want
it
to
do.
So
I
think
it
would
just
make
sense
to
somehow
surface
what
they
have
in
open
j9
in
the
omar
project.
I
think
that
some
of
the
discussion
some
of
what
derailed
or
slowed
the
discussion
last
time
was
around.
A
A
A
E
C
A
C
H
G
F
G
F
A
G
C
F
G
F
G
G
G
F
C
H
I
F
C
Right
yeah,
I
agree
so
if,
if
you
want
to
kind
of
draw
out
the
sort
of
the
longer
term
plan,
that
would
be
my
suggestion,
make
a
copy
of
what
we
have
in
openg9
now
put
that
new
copy
into
omr
and
that
that
can
mutate
into
the
github
actions
someday
and
then
once
that
happens,
open
g9
can
reuse
it,
but
we
get
the
the
you
know
the
benefit
of
the
this
automated
checking
in
omr
earlier
by
by
just
clone.
You
know
just
copying
it.
C
The
question,
though,
is
there
a
way
to
set
this
up
to
be
automatic,
which
is
one
of
the
goals?
As
I
understand
it,
you
know
a
committer
on
omr
will
say
you
know
jenkins
build
all
or
whatever
do
they
have
to
remember,
to
do
the
copyright
check
as
well,
or
is
it
going
to
be
automatic
somehow
for
jenkins
to
forgive
for
the
omar
project
when
somebody
opens
a
pull
request
right?
Sorry,.