►
From YouTube: Diagnostics WG meeting - Nov 20 2019
Description
B
B
Please
put
your
name
to
the
attendees
in
the
mean
time
and
I
see
that
two
items
on
which
is
not
really
in
this
documentary
we
should
discuss.
One
is
that
Girish
made
a
kind
of
a
pool
that
we
should
move
this
meeting
to
another
time,
because
it
doesn't
really
work
for
lots
of
people.
They
didn't
get
many
responses
back,
but
most
of
the
people
who
responded
I
hear.
B
B
B
A
B
A
A
B
It
was
one
topic
which
wasn't
in
the
agenda,
but
I
think
it
was
to
discuss
another
one
is
there
will
be
do
not
interact?
Leave
Canada
and
I
know
that
kibosh
will
have
a
presentation
and
I
kind
of
talked
with
him
about
that.
This
presentation
will
be
little
bit
about
how
the
working
group
is
working.
B
What
kind
of
the
working
group
is
achieving
what
problems
it
solves
kind
of
trying
to
recreate
more
people
for
the
working
group,
and
he
was
asking
if
you
want
to
propose
a
session,
because
I
think
that
would
be
also
opportunity
for
that.
I
won't
be
there,
but
if
there
any
topic
which
she's
so
burning
that
we
should
bring
it
to
the
collaborators
on.
A
B
Okay,
then
we
arrived
to
running
topic,
which
basically,
it's
almost
the
same
so
like
a
months
ago,
kind
of
we
added
sati
to
have
a
deep
dive
on
the
process
crash
and
we
kind
of
identified.
What
can?
What
kind
of
use
cases
it
can
happen,
and
why
and
we
discuss
it
in
a
second
part,
we
will
focus
more
on
the
available
tooling
and
what
you
would
like
to
achieve
in
the
future
on
tooling,
but
then
on.
The
second
one
only
aversion
me
was
so
didn't
really
make
a
progress
on
that.
B
B
C
B
A
B
B
B
B
A
A
A
B
B
B
B
A
B
B
B
A
Think
trace
events,
also
not
that
we
have
everything
we
need,
but
a
you
know
some
in
a
place
they
can
help
you
with.
If
we
have
the
right
trace
point
is
having
something
which
is
always
on
and
and
generates
a
like
a
ring
buffer.
So
you
can,
when
you
crash,
you
could
see
the
last
ten
trace
events
or
a
hundred
race
events,
or
something
like
that.
That
would
be
that'd
be
useful,
so
it's
worth
having
there
it
there,
even
if
we
can't
do
it
today
because
it
might
be,
and
we
should
work
on
something.
B
A
B
B
B
B
B
B
D
C
A
It
basically
runs
your
code
under
some
sort
of
emulation
and
it'll
report
at
the
end,
if
you've
like,
for
example,
it'll
tell
you
whether
you've
allocated
and
not
freed
memory
gives
you
a
summary
at
the
end,
but
the
other
thing
it
does
also
tell
you
about.
Is
it's
like
your
double
freeing
pointers
which
could
be
the
result?
Could
you
know
be
one
of
the
reasons
you
get
a
crash.
B
B
D
B
D
A
D
B
C
B
A
A
A
B
B
C
B
B
C
Second,
one
is
section
that
shows
these
statistics
of
the
JavaScript
heap,
basically
the
allocation
and
usage,
the
free
space,
etc
in
in
the
five
different
areas
within
the
JavaScript
heap,
so
the
old
space,
the
new
space,
the
code
code,
space,
things
like
that,
will
clearly
show
that
ever-growing
pattern
of
memory
so
that
that
should
show
the
first
symptom
of
memory.
Jake.
B
Okay,
which
one
is
what
do
you
think,
which
one
provides
so
just
in
a
user
journey
I,
would
like
to
have
a
kind
of
order
of
operations,
so
you
mentioned
that
there
are
two
ways
to
use.
What
do
you
think,
which
one
gives
the
better
chance
that
the
user
will
find
them?
So
what
we
recommend
that
this
is
the
first
step
and
another
one
is
the
second
step
yeah.
B
B
D
C
C
B
C
C
B
C
B
D
D
C
Yeah
I
mean
the
memory
X
ocean,
which
is
a
culmination
of
a
lot
of
memory
leak
in
the
past.
If
that
is
the
scenario
we're
talking
about
then,
yes,
the
call
stack
would
would
not
truly
represent
the
problem.
Instead,
it
would
represent
only
the
victim
which
piece
of
code
actually
finally
caused
the
exertion
to
happen.
But
if
you
are
able
to
detect
the
pattern
early
like
a
memory,
leak
has
been
observed
through
diagnostic
report
or
whatever
monitoring
means
then
probably
capturing
the
report
when
the
allocation
is
happening
may
show
up.
D
D
C
Yeah,
because
if
you
look
at
the
memory
exertion,
we
don't
generally
see
these
stacked
with
v8
doesn't
provide
you
a
stack
trace.
That's
because
it
doesn't
make
sense
to
show
the
called
stack
at
that
point
in
time,
because
the
the
current
call
stack
doesn't
make
it
make
sense
at
all.
The
the
exhaustion
is
basically
a
result
of
a
lot
of
bad
things
in
the
past.
B
Okay,
so
imagine
that
we
look
the
odd
space,
the
kind
of
have
a
high
confidence
that
we
have
a
memory
leak.
That's
caused
the
crash,
we
look
to
the
Java
sip
stack
and
you
go
unlucky.
What
would
be
the
next
step
to
identify?
What's
causing
the
memory
leak
or
we
would
not
point
to
the
memory
leak,
use
cases
and
that's
it.
C
We
could
have
other
scenarios
like
native
memory
leak
I,
exactly
have
no
customer
experience
with
that,
but
on
and
off
I
see
from
the
issues
in
the
core
reports:
elf,
about
the
buffers
and
shared
buffers
XLR,
which
am
native
backends,
potentially
causing
memory
leak.
They
not
necessarily
short
as
a
pattern
on
the
JavaScript
heap,
but
I,
don't
know
which
tool
or
which
section
the
report
is
going
to
show
that.
B
D
B
D
Production,
it's
a
bit
more
tricky
because
most
of
the
tools
you
can
use,
we
will
introduce
some
overhead
also.
They
are
more
system
specific.
So,
for
example,
a
two
that
we
work
on
Linux
will
probably
not
work
on
Windows
or
on
FreeBSD,
for
example,
for
Linux
you
can
use
BPF
if
you
have
trace,
for
example,
and
trace
Malick's
and
freeze,
but
you,
since
those
calls
will
be
frequent.
B
B
D
C
B
D
B
B
B
D
That's
not
a
tool
that
I
expect
just
rip
developers
to
need
to
use
to
be
honest,
okay,
this
will
be,
for
example,
if
you
are
developing
a
native
module,
you
have
a
a
crash
on
that
native
module.
Then
you
might
want
to
use
the
four
grind
or
if
you
are
working
working
on
that
car,
for
example,
but
in
other
cases
it's
it's
probably
not
going
to
be
to
be
necessary.
D
B
B
C
Yeah,
maybe
sorry
God,
probably
you
guys,
would
have
discussed
in
the
beginning
of
the
meeting,
but
just
wanted
to
quickly
touch
base
on
two
things.
One
is
I
set
up
a
duty
for
a
new
time
proposal
expecting
that
that
would
work
for
majority
of
the
folks,
but
looking
at
the
minutes
I
see
there
is.
There
is
no
acceptable
time
slot
that
was
identified
is
that
is
that
the
right
assumption.
B
We're
looking
to
do,
oh,
it
looks
lighter
to
me.
There
is
no
single
one
that
it's
overlapping
and
I
think
it's
it's.
We
discuss
the
of
really
hard
and
other
people
in
Europe
Asia
and
the
u.s.,
so
micro
propose
that
other
working
groups
as
alternative
times
and
like
one
working
group
using
this
time.
Another
one
is
having
like
every
second
week,
is
changing
kind
of
between
the
two.
That
would
be
an
eternity
where
we
are
trying
to
ask
people
who
are
really
off
like
in
this
case.
B
B
C
A
B
Okay
and
another
topic
which
we
discussed
at
the
collaborative
summit-
yes,
one
day,
trivia,
give
this
talk
about
the
dynastic
working
group
work
and
trying
to
make
people
excited
and
that
you
also
mentioned
that.
Maybe
there
is
not
what
in
the
first
session
and
you
kind
of
didn't,
have
anything
in
our
mind
to
to
bring
it
to
the
session
yeah.
That's
it.
D
C
I
believe
I
tagged
the
diagnostic
working
group,
but
then
later
I
remembered
that
the
the
submit
proposal
is
not
only
no
J's
organization,
so
the
tagging
does
not
really
work.
So
what
I'm
going
to
do
is
I'm
going
to
open
a
gh
issue
in
this
report,
link
to
that
proposal
and
basically
asking
for
feedback
or
if
you
want
to
propose
another
session
with
wider
content,
and
things
like
that.
That's
also
perfectly
fine
from
my.
C
C
B
Mean
this
session
made
sense
to
me
to
be
honest:
okay,
like
one
thing
that
I
really
would
love
to
to
discuss
with
people
that
I
think
nobody
I
mean
that
wasn't
really
conversation
around
having
the
dynastic
summit
next
year
and
I.
Think
there
should
be
one
and
we
are
head.
We
are
happy
to
host
at
Netflix.
Just
is
always
a
problem
that
people
doesn't
want
to
talk
about
too
much
yeah
and
I.
B
Don't
think
there
will
be
any
conference
not
related
in
California
that
I
know
about
it,
but
maybe
I
oughta
make
it
could
be
that
we
discussed
in
the
company
that
we
have
other
offices
and
I.
Don't
know
how
big
those
offices
are
so
I
definitely
need
to
ask
them,
but
maybe
that
could
be
one
in
it's
closer
for
most
of
the
people,
but
it's
like
I
think
it
would
be
great
if
you
could
poke
on
the
collaborate
on
the
collaborative
Sami.
B
C
Of
the
other
thing
that
probably
happened
between
this
session
proposal
and
raise
that
the
note
convey
you
lab
on
the
diagnostic
report,
a
majority
of
the
folks
who
have
been
using
the
report
wanted
to
know.
Can
we
use
it
in
production?
They
really
ordered
to
use
it
in
production.
So
definitely
there
is
no
performance
overhead,
because
once
you
enable
this
in
the
production,
there
is
a.
C
There
is
absolutely
nothing
happening
in
the
background
only
when
the
report
is
being
generated,
a
trigger
is
being
put
and
some
some
additional
activities
happening
only
at
that
site,
but
unlike
profiling
and
all
there
is
no
recurring
overhead,
so
they
really
want
to
use
it.
But
given
that
it's
experimental
I'm
thinking
should
we
propose
or
seek
the
feedback
from
the
wider
collaborators
about
making
this
table
or
if
you
are
proposing
or
if
you
are
collecting
the
feedback
on
the
diagnostic
summit
in
2020
early
2020.
B
C
B
D
C
D
B
Think
the
TLDR
is
that
none
of
us
has
a
really
urgent
topic
to
discuss.
So
idea
could
be
just
a
general
like
grouping
together
and
dynastic
and
like
the
working
group
and
session
and
the
summit
and
the
future
of
this
whole
things
or
if
you
have
a
topic
or
under
the
NSC
report,
I've
seen
it
was
a
good
one.
I,
don't
think
anyone
else
has
a
urgent
topic
to
to
bring
to
the
collaborate
assume
it's
up
to
you.
C
B
One
thing
that
I
forgot
at
the
meeting
at
the
beginning
of
the
meeting
and
they're,
really
sorry
that
Adam
I
see
you
are
in
the
call
and
Kirsten
II
I.
Don't
really
know
you
so
do
you
know
a
little
bit
talk
about
yourself
what
you
are
doing
and
why
you're
interested
in
the
in
the
lancet
working
group?
Yes,.
E
A
E
E
We
have
just
kind
of
thought
a
proof-of-concept
and
had
made
it
make
a
few
API
changes,
but
now
we're
kind
of
we're
in
a
good
place,
I
would
say,
and
really
what
this
is
about
is
just
we
I
saw
on
Twitter
that
the
ASA
cooks
spec
could
meet
could
use
some
help.
The
next
stage
and
I'm
just
here
to
say
I,
would
like
to
help
move
that
one,
if
possible,
and
any
other
help
that's
needed
as
well.
E
B
Awesome
that
sounds
great.
To
be
honest,
the
whole
I
see
the
whole,
not
yes,
community
and
working
group.
How
it
works
is
really
like
picking
up
things.
It's
like
there
is
no
like
test
assigning
and
I
did
all
project
management
behind
it.
It's
really
kind
of
you
pick
up
things
that
you
are
interested
in,
so
your
interests
in
a
scene
hooks.
B
E
B
So
I
stopped
yeah,
so
also
start
I
would
recommend
to
we
have
to
issue
I
think
in
both
in
the
North
core
and
the
in
the
domestic
repo
around
the
Asian
cooks.
So
as
a
beginning,
I
would
I
would
just
say,
like
kind
of
hider
that
that
you
have
that
you
want
to
work
on
this.
You
are
willing
to
work
and
like
that,
we
were
looking
for
context,
and
hopefully
people
who
were
involved
will
shatter
historical
context.