►
From YouTube: 2022-09-07 meeting
Description
Open Telemetry Meeting 1's Personal Meeting Room
A
B
A
B
B
Yeah
mark
should
we
discuss
about
the
pr
which
you
have
raised
to
start
with.
Probably
then
we
can
go
with
the
other
stuff.
A
Sure
we
can
start
with
that.
So
I
started
all
the
all
the
investigation
with
one
pr
testing
different
things
and
then.
A
A
A
So
the
next
question
is
for
windows.
The
code
change
to
do
is
a
bit
different
in
structure,
which
is
a
bit
annoying.
A
So
I
tried
to
to
do
the
code
change
for
windows
in
line
with
what
went
actually
investigated
a
while
ago.
The
unit
test
is
called
is
building
for
windows
to
start
with.
So
it's
it's
a
good
thing.
We
have
really
test
that
that
can
be
rebuilt.
A
Unfortunately,
it's
not
working,
and
I
don't
know
why,
because
the
ci
logs,
don't
say
much
about
it,
we
need
someone
with
a
machine
to
actually
build
locally
and
see.
What's
going
on.
B
Okay,
so
so,
if
I
understood
like
the
unit
tests
are
working,
so
we
can
see
the
failure
happening
for
windows.
A
A
Yes
and
the
good
thing
also
is:
I
tried
the
unit
test
alone
without
any
fixed
rider
files,
and
the
unit
test
is
failing
all
over
the
place,
so
the
unit
test
is
really
putting
stress
to
test
what
is
needed.
A
A
So
over
the
unit
test
is
brutal,
but
it
it's
it's.
So
it's
doing
it's
using
several
libraries.
It's
using
two
static
categories.
B
A
A
Yes,
so
can
you
look
at
the
file
video
file
macros,
and
it
contains
a
huge
comment
that
explains
what's
going
on
there.
B
B
B
A
B
Which
is
which
is
kind
of
strange
because,
based
on
whatever
others
have
tested
till
now,
I
think
the
conclusion
was
that
for
windows
without
splitting
header
and
source
we
won't
be
without
splitting
these
headers
into
source
file.
We
won't
be
able
to
do
it,
but
owen
is
saying
that
it's
what
it
possible
to
do.
A
B
A
B
B
B
That's
what
I
think
asan
you
also
mentioned
last
time
right,
yeah,
yeah,
and
because
I
you
know
we
have
to
for
for
windows,
we
have.
We
do
have
to
test
it,
not
just
for
which
for
windows
build,
but
also
using
mingw,
gcc
or
c
lang.
Yes,
okay,
gcc,
on
mingw
how
it
works,
and
there
would
be
some
testing
required
and
just
to
see
if
it
really
works.
So
header
only
should
work
for
ngw
and
gcc,
but
that
also
needs
need
to
be
seen
separately.
B
A
A
B
Yeah,
so
probably
I
think
let's
scope
it.
I
think
this
to
me.
It
looks
like
that
this
should
work
good
for
at
least
gcc
and
c
lang,
and
even
the
unit
test
is
working
good.
So
probably
we
can
scope
it
for
that,
and
I
think
we
can
take
it
that
that
will
at
least
unblock
lots
of
people.
A
B
B
B
A
B
A
A
B
That
good,
that
we
have
at
least
for
gcc
and
c
lang
so
yeah
thanks
for
this
mark
yeah
and
apart
from
that
grpc
upgrade
yeah.
B
A
B
B
B
B
B
A
B
Okay,
let
me
see
because
I
think
from
my
side
I
did
try
to
build
it
with
latest
version
of
bazel.
I
think
I
did
put
the
comment
and
after
that
I
also
tried,
but
unfortunately
and
locally
on
mac.
I
was
still
getting
a
crash,
so
I
could
reproduce
the
issue:
okay,
with
the
older
version
of
bazel,
and
even
after
using
the
latest
bazel
odor
2.2.
B
Still
there
was
a
segmentation
fault,
so
I
thought
that
we
will
raise
a
issue
with
grpc
and
let
them
let
them
handle
it,
because
it's
if
the
grpc
is
failing
with
latest
grpc
build
is
failing
with
the
latest
bazel.
Probably,
I
think
let
them
create
an
issue
with
them
and
let
them
investigate
what
the
segmentation
fault.
B
B
Let
me
see
this
is
on
linux,
okay.
I
think
this
would
be
easy
to
test
at
least
okay.
I
don't
see
any
any
straight
forward
solution
for
this-
let's,
let's
probably
for
this-
let's
create
a
ticket
with
grpc.
Let's
let
me
just
because
I
was
able
to
reproduce
the
issue
locally
on
mac.
So
let
me
let
me
probably
test
it
once
again
and
if
it
is
failing
I'll
create
a
ticket
with
for
grpc
and
okay,
then
let's
see
what
what
to
do
for
that.
B
B
Yeah
good
point-
and
I
think
this
is
okay-
you
even
we
have
to
probably
see
what
what
will
happen
if
we
create
a
ci
for
a
big
up
for
a
latest
version
of
grpc
and
building
it
with
only
with
c
plus
14.
B
What's
what's
the
whether
it's
still
failing-
or
at
least
it
works?
Fine?
So
probably
we
can
have.
I
think
that
we
did
discuss
it
last
last
week
and
I
think
we
agreed
that
we
are
going
to
support
for
the
higher
versions
only
with
c
plus
plus
14,
yes
lower,
for
I
mean
so.
If,
if
somebody
is
using
somebody,
if
somebody
is
using
open
telemetry
with
c
plus
plus
11,
they
won't
be
able
to
use
the
higher
version
of
grpc.
They
would
be
only
able
to
use
it
if
they
are
using.
B
A
One
one
question:
in
general:
yeah:
I
don't
get
exactly
who
decide
which
version
of
grpc
to
use
this
development
telemetry
source
code
or
is
it
whoever
is
building
it
when
specifying
a
path
to
an
open
to
a
grpc
package
somewhere.
B
A
B
Should
test
documentary
yeah,
we
should
test
and
document
it
for
all
the
scenarios
which
we
have
okay
right
now
we
have
is,
we
have.
We
do
have
constraints
using
with
gcc
4.8,
and
we
have.
We
have
documented
that
so
even
we
have
to
document
for
grpc
version,
higher
grpc
version
with
c
plus
plus
one
compilers,
which
won't
be
supported
so
for
our
ci
in
our
ci.
We
would
basically
our
ci
would
be
testing
all
those
scenarios.
B
We
should
have
a
ci
test
to
test
all
those
scenarios,
but
at
least
but
from
application
side,
the
user
side.
They
should
be
able
to
use
whichever
version
of
grpc.
They
want
to
use.
A
B
A
B
A
B
Okay,
and
next
is,
I
think,
documentation
yeah.
A
So
I
guess
the
documentation
team
has
a
big
question
as
to
where
to
put
the
dark
and
we're
very
true.
So
there
are
two
different
repositories:
there
is
open
telemetry.
I
o
that
contain
the
dock
of
a
website.
If
I
understand
correctly,
and
then
we
have
some
documentation
in
the
open,
telemetry
cpp
repository.
A
But
after
thinking
more
about
it,
it's
it
might
be
more
convenient
for
us,
but
it
won't
be
convenient
for
everyone
in
the
long
term
and
it
will
cause
also
more
problems
because
of
mergers
because
of
people
who
are
not
aware
of
that
and
do
change
in
the
world.
But
anyway,
and
things
like
that.
So
I
guess
I'm
okay,
with
checking
in
code
in
open,
telemetry,
io
directly.
A
It
will
be
less
convenient
for
us,
but
at
least
it
will
be
only
one
place,
and
the
advantage
I
see
to
that
is
that
there
are
also
some
editorial
changes
like
I
don't
know,
fixing
typos
fixing
the
table
of
contents,
changing
styles
and
whatnot,
which
the
web
team
can
do
directly
without
needing
a
pr
on
the
open,
telemetry
cpp
code.
So
it
has
some
advantage
also.
B
A
B
A
B
We
have
these
rst
files
which
get
converted
to
html
and
it
get
published
in
read
the
docs,
the
same
same
for
sdk
and
apis.
So
we
have
kind
of,
I
think,
a
general
a
generic
process
for
both
non-api
and
api,
and
we
publish
everything
on
read
the
we
docs.
A
Probably
need
clarification
based
on
the
type
of
documentation
like
if,
if
a
documentation
is
for
the
external
website,
it
needs
to
be
written
in
that
way
and
within
that
place.
If
it's
internal
process
guidelines
like
how
to
do
a
tagging
git
to
publish
your
release,
which
is
only
for
developers,
it
would
be
in
another
way
done
in
another
way
in
another
place.
Things
like
that
so,
but
there
is
also
the
old
oxygen
documentation
which
is
tied
to
the
code.
B
A
A
That,
if,
if
we
do
that,
then
people
raising
a
defect
against
io
cannot
fix
that
in
io.
They
will
need
to
do
a
push
request.
B
A
B
Yes,
which
is
fine
right
in
I
mean
anyway,
they
have
to
any
documentation,
has
to
go
through
us.
You
know
I.
I
don't
think
that
if
it's
a
c
plus
documentation
just
open
telemetry,
I
o
should
die
the
approvers
at
open,
telemetry
io
should
be
able
to
approve
it,
and
so
either
open
telemetry
I
have.
I
o
should
have
some
documentation
on
how
to
really
update.
B
A
Oh,
by
the
way,
I'm
finally
part
of
open
telemetry.
B
Okay,
but
you
can
see
your
yourself
right
now
in
open,
telemetry
org
right
get
up
open,
telemetry,
okay,
which
is
good
here
then.
Finally,
it's
dogs
and
sorry
reference.
B
B
A
B
A
B
Which
is
strange
even
the
also
has
everything
coming
here.
Okay,
I
think
which
is
fine.
I
think
that
if
they
really
want
to
go
on,
you
know
if
they
want
to
pull
all
those
documentation.
I
think
this
is
fine.
Let
them
do
it
and
I
think
we
can
ask
them
if
they
want
to
do
all
the
heavy
lifting
create
a
pr
on
our
repo
and.
B
B
A
B
Okay,
thanks
mark,
would
you
want
to
comment
on
this.
B
B
B
A
B
B
Yeah,
okay,
I
don't
see
anything
else.
Probably
we
can
quickly
go
to
the
specs
changes.
Events
and
logs
api
is
already
merged,
so
I
think
we
have
to
start
looking
one
sometime
on
the
changes
which
we
have
to
do
on
log
api
and
also,
I
see
a
pr
which
is
already
being
raised
to
align
the
log
sdk
with
the
api.
A
B
And
explicit
bucket
boundaries.
I
think
we
discussed
this
last
week
this
pr
is
we
had
it
last
week
it
was
the
issue:
was
there
to
change
the
bucket
boundaries,
mostly
for
the
performance
reasons,
and
the
issue
was
there
now,
the
pr
is
there.
So
I
think
this
once
this
is
merged,
we
may
have
to
do
our
code
changes
in
our
matrix
sdk
to
take
to
use
this
to
the
new
chased,
buckets
and
non-periodic
metric
reader
here
this.
This
is
a
good
issue,
I'm
just
saying
as
of
now.
B
This
is
somehow
missing
in
the
specification
we
only
have
periodic
metric
leader
which
will
read
the
matrix,
which
will
collect
the
matrix
data
periodically,
but
there
is
no
non-periodic
metric
reader,
which
can
be
called
from
outside
dsdk
so
as
and
when
needed.
So
this
is
something
a
good
addition.
I
think
once
we
did.
This
get
public
once
just
get
added
in
the
specification.
B
B
A
So
we
would
expose
an
api
to
do
that,
but
if
somebody
still
has
to
export
a
exposure
port
or
something
new
with
the
glue
to
actually
call
that
api,
I'm
assuming
we,
we
are
not
exposing
like
a
tcp
port
to
do
scraping
directly
or
things
like
that.
A
A
B
And
yeah,
then
the
scope
attribute:
this
is
a
this
is,
I
think,
not
a
recent
pr,
but
this
change
was
introduced
a
couple
of
months
back
and
we
never
aligned
to
these
changes.
A
Or
a
logger,
I
I've
seen
a
lot
of
changes
to
the
portal
buff
or
idea,
or
for
that.
So
is
it
the
case
that
it's
part
of
a
lot
of
change,
but
we
are
not
using
it
yet
so.
B
A
B
B
It
would
be
part
of
instrumentation
scope
and
I
think
we
need
to
add
this
sometime.
It
would
be
a
breaking
change
as
and
when
we
add,
even
though
it's
optional,
but
it's
it's,
it's
abi,
breaking
change
I'll
say
not,
may
not
be
api,
but
definitely
abi.
B
So
yeah,
I
think
probably
sometimes
we
have
to
definitely
just
convert
things
yeah
and
anything
else.
Probably
we
can
quickly
go
through
the
issues
and
the
prs.
Apart
from
what
we
have
already
discussed,
I
think
matrix,
I'm
definitely
not
state
is
not
discussing
this
week.
I
don't
see
any
substantial
changes
from
last
couple
of
weeks.
Please
correct
me.
I
think
I
sound
refuse,
but
I
haven't
done.
We
haven't
done
much
progress
in
matrix.
A
Yeah
yeah,
but
is
that
one
thing
I
mean
it's
not
really
related,
I'm
currently
working
on
the
one
issue
on
kubernetes,
I
think
ingress
engines
controller.
They
are
trying
to
use
this
engine
mix
module
from
country
and
they
had
problem
with
compiling.
If
you
remember
yes
issue,
I
fixed
that
one.
B
B
A
B
B
B
Yeah
multiple
cumulative
metric
collection
without
measurement
recording
yeah.
This
is
fixed
and
I
think
thomas
review
tom
had
tom
had
some
good
comments
on
this,
and
I
think
I
updated
that
and
the
the
original
the
person
who
created
the
issue
he
did
test
the
changes
and
it
works
fine,
as
he
mentioned.
B
B
Okay
yeah!
I
think
he
has
to
join
some
other
meeting.
He
has
to
join
the
log
sig
meeting
that
reminds
me.
We
do
have
to
discuss
about
the
timing
of
this
meeting
as
it's
conflicting.
The
30
last
30
minutes
are
conflicting
with
the
logsic
and
tom,
and
probably
even
I
may
have
to
join
going
forward
in
the
log
segmenting,
so
yeah,
let's
discuss
it
afterwards,
so
yeah.
This
is
something
I
think
this
is
good,
should
be
good
too
much
I'll
check
with
tom.
B
If
you
have
any
other
comment-
and
I
said,
if
you
have
time
to
review
it-
just
just
have
a
look
into
that
quickly.
I
think
it's
it's
a
very
simple
small
change,
not
big.
Okay,
yeah
thanks
and
remove
reference
of
tracer,
provided
this
is
etw
exporter.
This
is
something
which
we
use
it
internally
in
microsoft,
so
nothing
I'll
I'll
work
with
tom
on
getting
it
through
and
that's
all.
I
think
we
don't
have
anything
new
for
issues.
B
B
A
B
It
is
what
yeah,
and
this
is
something
which
probably
we
can
avoid.
A
A
B
A
B
A
B
A
B
B
A
Question
in
general,
so
let's
say
there
is
an
exporter
that
talks
to
http
somewhere
and
the
part
is
not
up
so
it's
talking
to
nothing.
Then
there
is
a
an
error
which
is
raised
inside
the
open,
telemetry
sdk
eventually
gets.
It
gets
logged
in
the
debug
log
if
or
error
log,
if
it's
enabled,
but
from
my
application
is
there
a
way.
B
B
How
about
I'm
just
just
thinking
so
yeah,
it's
very
cool
to
that.
B
A
A
B
A
I'm
getting
a
message
that
I
can
put
into
my
own
work
file,
okay,
but
things
like
if
we
cannot
connect
to
the
remote
service,
for
example.
I
would
like
this
error
to
be
to
be
counted
so
that
I
know
how
many
times
I
failed
to
connect
to
the
to
a
collector,
for
example,.
A
A
A
Itself
is,
is
working
or
failing.
B
B
B
B
Okay,
we
are
getting
started
documentation
here.
This
is
something
which
we
have
to
do
it
and
probably
just
created
these
matrix
issues
just
to
track
the
ga.
What
are
things
which
we
need
for
ga
so
add.
B
Yes,
yes,
yes,
we
do
have
to
write
those
md
files
and
probably
they
have
to
pull
it.
Yes
in
that
website
doc.
So
the
way
they
want
it
here,
the
decode
values
for
add
support
for
poor
matrix
exporter
yeah.
This,
I
think
I
said
you
already
have
it,
and
this
is
just
targeting
for
the
ga
I
mean
whenever
you
have
time
nothing
pressing
as
of
now
decode
values
for
hotel
resource
attributes,
and
we
already
discussed
that
this
also
we
discussed
these
are
sdk
matrix,
sdk.
B
I
haven't
seen
that
I
haven't
really
done
any
investigation.
A
On
this
yeah,
I
have
seen
so
while
debugging
for
the
single
turn
debugging
unit
tests.
Okay,
I've
seen
a
bug
where
I
compile
one
file
and
it's
compiled
with
stl.
So
it
has
fcpp
stlib
and.
A
A
file
and
it's
not
compiled
with
the
same
flag
and
when
linking
the
two
it
creates
a
crash
in
in
the
no
steady
shot
pointer
exactly,
and
so.
The
issue
is
that,
in
my
case
it
was
that
the
build
is
not
consistent,
because
the
build
was
done,
presumably
with
stl,
but
some
code
was
compared
with
stl.
Some
code
was
compiled
with
no
std
and
linking
the
two
across
a
crash,
so
it
could
be
related
to
that.
Also,
maybe.
B
Okay
good
point
here
so,
but
that
will
only
happen
if
somebody's
purpose,
if
somebody
is
by
mistake,
doing
that
right.
Otherwise
it
should
not
our
power
build
files.
Cmake
should
not
be
doing
that.
B
Okay,
yeah,
probably
I
think
you
can
just
comment
if
that
that
that's
the
potential
that's
one
of
the
potential
issues
for
this
is.
A
Yes,
so
basically
welper,
although
I
don't
know
what
equality
is
because
velcro
has
to
be
header
only
as
usual,
so.
B
Yeah,
that's.
That
was
one
of
the
things
we
wanted
to
avoid
at
the
api
level,
creating
any
dedicated
types
and
objects
for
any
of
the
expand
data
span,
attributes
which
we
use.
So
this
this
just
correct
me.
This
solution
is
something
which
will
create
a
span
link
which
will
have
reference
to
context,
and
I
have
probably
maintained
the
reference
to
context
in
reference
to
the
attributes
or
something
like
that.
A
B
A
B
A
Is
that
when
I
start
to
add
a
link,
the
api
was,
I
mean
it
took
me
a
day,
literally
yeah
to
add
one
span
context
as
unique,
because
the
gap
between
this
pattern
context
type
and
what
the
api
expects
is
too
too
big.
I
had
no
idea
what
what
iteratable
value
or
what
or
whatever,
to
give
and
how
to
get
one
so.
B
B
Yeah,
I
agree
with
you
this.
This
is
bit
complicated.
The
way
we
pass
links
in
start
span-
and
I
think
the
using
this
iterable
triple
is
a
bit
complicated,
probably
good,
to
have
some
some
examples:
how
to
do
it,
but
I'll
be
more
interested
to
see
if
this
is
really
causing
any
performance
issues
or
in
terms
of
sending
a
single
link.
B
A
B
B
So
so
I
mean
what
what's
the
suggestion
right
now
I
mean
it's.
I
really
don't
want.
B
Yeah,
okay,
fine,
okay,
yeah
probably
will
we
can
we
can
you
can
track
this
issue
and
add
an
example,
and
I
think
that
you
should
be
good
yeah,
at
least
in
the
examples.
It
should
be.
Okay
in
the
examples
right.
Yeah,
probably
you
can
add
that
yeah.
I
think
that
should
be
good.
A
So
this
one,
I
don't
think
we
use
json
for
meta
files
anymore.
B
B
B
B
B
B
A
Changes,
I
will
do
a
crap
again
and
comment,
but
I
think
it
was.
It
would
be
the
easiest
solution.
B
A
It's
not
used
in
sdk
header
files.
It's
definitely
not.
B
A
B
A
Just
to
say
that
I've
seen
a
lot
of
progress
in
a
matrix
while
I
was
in
vacation.
So
that's
good.
B
B
Probably
that
took
lots
of
my
time.
I
don't
know
if
I
raised
a
pr
for
that
in
the
con
trip.
This
is
something
very
specific
to
internal
to
microsoft,
but
yeah.
This
is
some
some
microsoft
internet
exporters,
but
it's
something
which
we
already
publishing
it
on
on
github,
so
has
been
spending
more
time
on
this
to
make
it
work.
But
I
think
this
should
be
done
by
by
this
weekend
and
I'll
be
spending
more
time.
On
back
to
my
textbook.
A
Okay,
yeah,
and
so
the
other
thing
is,
I'm
I'm
hopeful
to
finally
put
a
nail
to
these
singleton
things,
so
that
would
be.
A
B
Yeah
and
okay,
thanks
thanks
for
joining,
have
a
good
rest
of.
A
B
B
Sooner
yeah,
so
right
now,
if
we
see
the
calendar,
the
problem
here
is
that
okay,
let's
open
yeah
this,
this
is
right.
Now
this
is
phd
time,
so
it
starts
our
time,
9
30
and
then
at
10
o'clock.
There
is
a
log
sig
meeting.
B
A
Yeah
well
I'm
playing
for
longest,
I'm
in
europe,
so
it
works
for
me.
I
don't
know
if
people
in
san
francisco,
if
you
can
make
it
for
nine
or
or
not.
A
A
B
B
No,
there
won't
be
any
community
pr,
because
community
just
has
links,
so
this
just
have
links
to
the
google
doc
yeah.
It
has
to
change
here.
Sorry
yeah,
yes,
yeah
yeah,
so
we
have
to
change
the
calendar.
We
have
to
change
here
and
probably
we
have
to
change
our
reports.
If
we
maintain
women,
if
we
mention
the
time
there,
so
three
things
has
to
be
done
here.