►
From YouTube: 2021-05-24 meeting
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
B
I
only
have
one
item
on
agenda.
I
wanted
to
explain
a
little
bit
of
that
upsell
variant,
change
that
I'm
making.
I
actually
made
sure
that
all
tests
are
now
passing.
Yes,.
B
B
Okay,
it
turned
out
to
be
a
bit
bigger
than
I
expected,
especially
because
upsell
already
used
by
otlp
exporter
and
a
different
version
of
it,
so
which
part
of
ultra
big
spotter
is
using
upsell
yeah
it's
in
the
through
the
grpc,
okay
yeah,
and
it
just
happens
that
since
we
require
api-
and
we
also
require
generated
the
grpc
headers,
we
end
up
including
two
versions
of
upsell
in
a
single
compilation
unit
and
then
there's
another
funny
thing
is
that
although
upsell
is
designed
to
be
linked
private
and
you
can
define
a
custom
namespace
for
it,
it
is
not
designed
to
include
two
different
versions
of
it.
B
Unit.
Okay!
Yes,
if
you
included
another
version
from
a
an
api,
it
would
complain
that
some
macros
have
been
already.
B
Yeah
and
I
ended
up
needing
to
change
like
20
or
30
files,
it's
mostly
search
and
replace,
and
the
effort
was
pretty
much
to
isolate
it
into
a
self-sufficient
header.
Only
library
that
is
unique.
That
is
not
conflicting
with
any
other
version
of
upsell.
I
think,
but
it's
a
bit
interesting
exercise.
I
honestly
hit
it
only
after
enabling
the
otlp
exporter,
okay,
because
other
exporters-
usually
don't
don't,
don't-
require
your
pc
anyways.
Let's
wait
for
other
folks
to
join
yeah.
B
And
the
variant
issue
is,
it
seems,
like
there's
a
proposal
in
c
plus
plus
20,
to
resolve
some
errata
about
conversion
of
string
like
object
into
either
string
or
string
view,
but
by
default
in
c
plus
17.
B
If
you
do
not
list
like
cos
charge
star
on
the
variant,
what
happens
is
that
it
just
assumes
it's
bull:
okay
and
it.
It
gets
implicitly
converted
to
bull.
And
if
you
pass
something
that
looks
like
a
string.
B
It
then
end
up
like
false
or
true,
and
it's
very
like
it
took
a
bit
of
time
to
understand.
B
Yeah,
so
what
I
did
is
my
take
here
is:
let's
say
we
still
have
to
deal
with
c,
plus
17
compilers
for
another
period
of
time
and
I'd
say
it's
maybe
another
two
years,
and
even
though
it
seems
like
in
latest
stl,
they
are
fixing
it.
B
I'm
not
sure
it
hasn't
been
released
yet,
and
I'm
not
sure
how
long
it
would
take
for
that
stl
patch
to
be
actually
released
to
most
customers
and
how
long
before,
customers
install
that
so
same
issue
like
no
conversion
to
string
view
applies
to
upsell
as
well,
so
both
upsell
and
standard
implementation.
B
They
have
the
same
ball
and
because
they
have
the
same
bug.
I
had
to
be
very
explicit
about
conversion,
and
I
only
added
it
on
api.
There's
no
need
to
add
it
on
owned,
attribute
value,
because
on
owned
attribute
value
we
already
have
string
type.
So
only
when
we
convert
from
non-owning
api
call
to
recordable.
B
That's
where
I
had
to
add
like
three
lines
of
code
into
every
exporter,
and
I
already
covered
it
for
those
exporters
that
we
officially
support.
Okay,
I
hope
that
it's
not
a
big
maintenance
hassle,
because
I
already
did.
B
I
I
I
would
like
to
add
separately
a
build
loop
for
2015
to
show
that
now
we
cover
2015
with
that
implementation,
because
previously,
in
part,
didn't
even
compile
with
that.
Okay,
now
we're
covering
pretty
much
all
our
bases,
2015
and
the
standard
library
also
works.
Well,
because
I
I'm
adding
that
type
on
api
surface
and
even
if
we
compile
with
the
studio
all
tests
pass
and
all
values
string
values
get
properly
populated.
B
So
I
think
it
it.
It
honestly
took
me
a
bit
more
time
than
I
originally
planned
because
of
all
these
you
know
quirks
with
upsell
conflict
and
the
the
lack
of
proper
conversion
of
for
the
variant
string
type.
B
Yeah,
but
I'm
fixing,
like
both
of
these
the
older
one
that
I
loved
about
the
implicit,
the
wrong
conversion
to
boolean,
as
well
as
getting
rid
of
impart
variant.
I
just
deleted
it.
Maybe
I
should
keep
it
like.
You
know
no,
but
I
I
cleaned
it
as
if
it
never
existed
and
now
as
if
this
upsell
is
the
default.
B
Now
I
was
thinking,
maybe
I
should
also
provide
a
script
like
search
and
replace
script.
When
I
do
a
snapshot
of
upsell,
perhaps
I
should
explain
the
the
reasoning
why
it
it's
a
private
snapshot
so
that
it
doesn't
clash
with
the
otlp
upsell,
and
I.
B
The
other
option
is
for
somebody
who
builds
the
entire
thing
a
from
source,
including
jrpc,
and
if
they
are
statically
linking
and
if
they
do
not
use
a
dll
at
all,
then
in
theory
they
have
that
option
with
underscore
upsell.
B
Yeah,
that's
right,
yes,
yeah
and
then
we
don't
need
our
local
snapshot,
which
seems
like
I
mean
I
can
measure
how
smaller
it
is.
It
would
probably
not
be
a
lot
of
so.
B
But
it
also
depends
on
a
bunch
of
tricky
code.
So
it's
not
just
it's
variant,
header
and
all
the
utility
functions
associated
with
that.
A
I
think
so
another
question
is:
do
we
need
to
update
this
snapshot
for
for
newer
version
of
upsell
great
question?
I
noticed.
B
That
grpc
uses
more
fresh
version
of
it,
so
I
was
thinking-
maybe
I
should
see
like
on
my
machine
what
version
I
get
right
now
and
I
think
it's
a
release
from
2021,
because
I
I
originally
took
this
snapshot
in
2020.
It
was
like
march
21
in
snapshot.
Now
there's
a
more
fresh
version.
I
I
can
refresh
that
and
perform
as
I
refresh
that
I
can
also
supplement
the
readme
file
with
the
script
that
I
run
to
just
search
and
replace
respecting
the
case.
So
it's
like
absl
capital
underscore
right
now.
A
B
Yeah-
and
it's
like
honestly,
I
didn't
expect
that
it's
gonna
be
kind
of
that
hard
and
it
it
is
ugly,
but
I
don't
see
any
other
better
way,
because
without
that
patch,
the
two
up
sales
are
going
to
clash
while
compiling
the
otlp
export.
B
For
jrvc
right,
let
me
try
that
that's
a
good
question,
so
let
me
try
that
locally.
I
can
do
that
and
I
hope
that
it's
still
going
to
pass
like
I
pretty
much
locally
tested
that
already
on
on
my
like.
I
know
that
ci
passes
for
all
these
platforms,
but
I
had
to
spend
a
bit
of
time
on
apple,
gcc
and
windows.
Visual
studio,
like
I
had
tried
a
few
versions
and
that's
all
working
okay,
yeah.
B
Yes,
yes,
and
I
think
it's
it's
a
kind
of
big
bit
risky
change,
so
I
want
to
give
you
guys
enough
time
to
take
a
look
at
that
and,
from
my
end,
I'll
try
to
see.
Maybe
I
should
just
refresh
that
to
2021
snapshot
of
upsell.
Okay,
that
sounds.
A
All
the
items
in
your
topics-
yes,
that's.
B
The
only
topic
that
I
wanted
to
cover,
I
added
that
to
the
document
like
the
word:
docs
docs,
google
docs
document.
A
A
B
I
hope
that
since
we
already
kind
of
tried
a
lot
of
things
with
c
plus
plus
20,
at
least
when
I
build
with
standard
lib
on
the
plus
plus
20
compiler,
it
all
builds.
I
see
warnings.
B
There
are
certain
things
which
have
been
duplicated
already
since
c,
plus
plus
17,
and
I
also
saw
a
few
warnings
with
clang,
which
are
not
usually
seen
on
gcc
and
the
old
visual
studio.
I
think
we
might
expect
some
more
warnings
with
the
new
compiler.
I
can
give
it
a
try.
A
B
A
B
Yes,
first
thing:
we
can
give
them
all
just
to
press
it
this
way,
and
this
is
the
option
you
can
use
to,
suppress
it,
but
we
gotta
be
prepared
to
maybe
clean
it
properly
so
that
we
don't
have
to
suppress
it's
going
to
be
much
nicer.
That
way.
B
I
think
we
should,
I
think
we
should.
I
think
we
should
and
in
case
when
we
can
detect
the
compile
the
the
standard,
the
language
standard.
Maybe
we
should
have
that.
B
You
know
check
if
language
standard
does
this
there
was
that
one
thing
with
the
std
result
or
for
something
like
feature
that
got
removed:
okay
plus
plus
17,
but
I
I
am
hopeful
that
it's
gonna
be
easy,
because
my
understanding
is
c
plus
plus
20
is
the
the
latest
standard
we
kind
of
already
tried
building
with
that
on
other
platforms,
and
I
think
latest
visual
studio
is
anchoring
to
just
have
better
more
complete
support
of
that
language
standard.
B
So
there
should
not
be
a
lot
of
surprises,
but
yes,
we
need
to
to
focus
on
that,
and
unfortunately
the
bracket
is
wide.
B
I
mean
main
reason
is
why
I
want
to
cover
15
is
some
of
our
customers
in
azure
were
asking
for
e-world
compiler,
and
I
look
at
that
parallel
to,
for
example,
like
how
we
still
support
gcc
4.8,
which
is
old,
but
we
still
agree
to
support
it
and
I
think
on
windows.
It
should
be
at
least
for
now
about
the
same
attitude
regarding
visual
studio
2015.
B
Perhaps
we
can
say
we
require
update
2,
because
upsell
itself
requires
update
2,
which
means
that
your
pc
itself
would
also
require
update
2,
because
it
depends
on
upsell,
and
we
should
kind
of
spell
out
that
this
is
the
compiler
that
we
still
support.
B
Years,
it's
a
great
question
like
I
would
cautiously
say
that
I
had
the
experience
when
I
spoke
to
there's
a
bit
of
a
disconnect
between
I.
I
realized
that
there's
an
expectation,
for
example,
that
we
always
use
recent
latest
compiler.
It
has
all
the
features
and.
A
B
And
there's
a
push
like
guys
use
latest,
but
at
the
same
time
there
are
customers
who
are
kind
of
stuck
and
they
say.
Oh,
we
know
old
compiler
works
for
us
and
we
don't
want
to
upgrade
to
latest
because
it
just
worked
for
us.
Even
if
there
are
some
glitches,
we
haven't
hit
those,
but
then
there's
a
bit
of
a
disconnect
like
how
far
back
we
should
look
and
when
all
of
our
customers
would
be
at
least
on
2017.
B
Yes,
yeah,
and
that's
going
back
also
to
that
issue
that
we
discussed
with
lolita
about
the
variant
patch
fine,
even
if
it's
fixed
in
2020
or
2022,
the
customers
are
still
gonna,
get
only
to
2017
like
this
or
next
year,
which
means
that
they
still
haven't
gotten
that
variant
proposal
that
resolved.
B
So
it's
like,
we
still
have
to
look
back
and
keep
it
working
with
the
legacy
stuff,
that's
kind
of
broken,
but
it
was
operational
and
there
are
ways
to
get
around
it.
So
I
mean
I
do
not
see
us
getting
rid
of
15
for
a
year
and
I
think
for
another
two
to
three
years.
We
would
like,
I
think,
15
for
a
year
and
17
is
probably
for
another
two
years
or
three
years
from
now.
A
B
I
haven't
hit
much
issues
with
that.
It
has
variant,
it
doesn't
have
spam
so
spam.
We
still
use
custom
spam
and
let
me
check
about
string
view.
I
don't
remember,
I
think
string
view
was
stream
view.
A
string
of
view
was
also
there
so
string.
A
B
A
B
No
std
span
and
for
the
no
std
span
we
do
have
two
options
right
now.
Our
custom
no
std
span
and
I
have
that
check
for
the
gsl
spawn
like
microsoft,
gsl
spam
gsl
spam
claims
that
it
provides
hundred
percent
drupal
and
replacement
for
the
standard
spot.
So
it's
like,
I
think.
When
I
looked
at
upsell
spam,
they
say
that
they
may
not
are
guarantee
hundred
percent
drupal,
whereas
microsoft
guidelines
support
library.
It
actually
claims
that
it
should
be
like
precisely
matching
okay,
the
standard
span,
so
we.
B
A
B
Okay,
great
and
it's
a
header,
only
template
library,
so
it
just
links
into
customer
code.
There's
no
need
for
the
lib
file,
so
we're
pretty
good
with
17.
I
would
say
I
think
tom
one
thing
about
the
dll:
let's
see
if
we
want
to
build
the
dll-
and
I
think
this
is
a
conversation
more
for
version
1.1.
B
B
Upsell
variant
and
verify
that
the
dll
can
still
be
loaded
and
operational
if
the
main
executable
is
built
with
either
17
or
19,
and
it
should
should
work
that
way.
It's
like
because
we
wanted
to
have
that
run
time.
Library,
loading,
abi
stability,.
B
Yeah
in
in
after
g
after
the
first
release,
okay,
we
were
kind
of
pushing
it
back
to
1.1
or
1.2
whatever,
like
later
this
summer,.
B
The
I
think
there
was
some
a
person
asking
for
header,
only
api
headers.
B
Yes,
they
were
working
on
sdk,
which
itself
does
not
implement
any
exporters
and
I
think
their
expectation
is
they
would
be
able
to
kind
of
dynamically
load,
the
actual
implementation
of
the
sdk,
and
that's
where
I
think
we
would
need
to
consider
this
api
stabili
bi
stability
guarantee
so
that,
irrespective
of
what
the
other
products
are
compiled
with
15,
17
or
19,
we
can
provide
a
recipe
for
shipping.
A
pre-built,
open,
telemetry
sdk,
maybe
with
the
htlp
exporter,
also
included
in
it
yeah
with
the
standard
export.
I
think.
B
Linux
for
sure,
because
for
linux
I
think
originally
when
we
started
all
this
abi
stability
mumbo
jumbo,
which
is
a
complex
topic,
we
we
wasted
so
much
time
on
that
stl
of
our
own.
The
original
ask
was
from
one
of
the
customers.
B
I
think
it
was
envoy
proxy
or
something
that
may
that
should
dynamically
load,
an
instrumentation,
elaborate
runtime,
and
the
issue
was
that
products
for
linux
may
cover
older
enterprise,
linux
versions,
which
were
also
built
with
gcc
4.8,
and
I
hit
an
issue
with
std
string
personally
when
I
was
trying
to
compile
with
different
steel
libraries
like
gcc,
four
and
gcc,
I
think
seven.
I
had
very
odd
things.
It
just
didn't
match,
because
it
doesn't
provide
the
same
memory
layout,
that's
okay,
linux,
first
and
for
windows.
B
I'd
say
with
the
customers
that
I
currently
work.
They
did
not
ask
for
deal
awarding,
but
my
view
is
very
limited.
There
might
be
others
who
need
it.
So
since
we
delivered
it
for
linux,
I
think
we
should
also
deliver
that
for
windows
as
well.
B
A
B
And
that's
why
we
have
these
types
now
added
also
like
bite
buffers.
We
have
it
and
we
have
unsigned
64
bit
integer
added,
because
if
and
it's
a
good
thing
because
now,
spec
of
for
the
logging
at
least
have
that
provision
for
the
byte
buffers-
and
we
already
had
it
on
api
surface
because
we
added
it
ahead
of
spec,
because
we
cannot
change
the
variant
layout,
we
have
to
freeze
it.
B
We
have
to
define
everything
we
can
is
adding
an
extra
field
on
that
variant
will
prob,
it
seems
like
it
will
break
ebi
compatibility.
We
have
to
just
you
know,
deliver
it
good
as
part
of
version
1.0,
g8
and
yeah.
B
It
seems
like
it's
just
two
of
us.
I
don't
see
other
folks
joining.
Yes
did
you
want
to
go
over
some
other
issues?
We
got
in
general.
B
Yes,
so
these
are
all
done,
I
mean
I
think
we
spent
like
good
15
minutes.
I
explained
you
the
merit
of
why
I'm
doing
it.
Yes,
I
need
to
get
feedback,
I
think
from
other
companies
as
well,
maybe
from
josh
and
evgeny,
because
it'd
be
good
to
get
feedback
from
other
companies
outside
of
microsoft.
B
B
B
I'll
add
the
comments
I'll
edit.
My
last
comment
that
all
the
tests
passed
and
I'll
explain
the
logical
reason
why
the
course
chart
in.
A
So
you
you
add
this,
this
discussion
items
to
the
pr
right,
so
yeah,
yeah
yeah.
B
I
I
don't
know
I
wish
we
had
the
better
tool.
Maybe
we
should
because
right
now,
when
it
fails
it,
it
creates
that
red
cross,
yeah
and
and
it's
confusing,
because
you,
you
must
keep
a
tr
thinking
that
it's
not
ready
because
it's
waiting.
B
But
then
the
only
thing
that
is
failing
is
that
good
coverage.
A
B
B
Maybe
it
would
be
sufficient
if
it
just
passes
and
leaves
the
link
to
report
like
I
mean
how
to
tell
like,
so
that
it's
fyi
kind
of
thing.
Yes,
it
shows
the
report,
but
it
doesn't
add
that
failure
mark
because
the
failure
market-
yes.
A
A
B
For
the
license,
I
spoke
to
elite,
there's
that
format
spdx
okay
and
it's
actually
recommended
by
linux
foundation.
Originally.
So
it's
like
cncf
is
a
sub
sub
project
in
linux
foundation
and
linux
foundation
itself
is
promoting
this
spdx
format,
which
is
just
pretty
much
two
or
three
liners.
B
B
Yeah
yeah
and
full
screen
I'd
say
we
should
adopt
this
spdx
as
well,
because
it's
just
a
modern,
better
way
of
documenting
what
your
license
type.
A
B
A
B
A
Look
yes,
and
besides
that,
I
think,
do
we
need
to
add
a
ci
task
to
check
whether
some,
whether
new
files
comes
in
and
make
sure
it
contains
slice,
information.
B
My
understanding
is
we
shouldn't.
Let
me
elaborate
on
this.
My
understanding
is
that
a
license
file
in
the
root
folder
already
covers,
even
if
these
files
don't
have
copyright
by
default,
the
license
file
applies
to
all
files,
it's
more
like,
for
example,
if
somebody
steals
the
source
code
and
copies
it
elsewhere.
We
avoid
unintentional
alteration
of
the
license,
so
it's
like.
If
we
included
it
in
there,
it's
there.
People
are
not
removing
it
and
if
they
just
copy
that
file
elsewhere,
the
license
is
retained.
B
Okay,
it's
more
like
for
consistency,
but
from
the
legal
standpoint.
My
understanding
is
that
license
file
in
the
root
folder
already
comes
so
it
covers.
B
You
can
do
code,
clone
search,
find
that
same
code
in
open
source
and
say
hey
by
the
way
the
root
folder
license
is
apache
license
v2.
B
You
are
not
supposed
to
change
the
license
of
it,
so
even
if
we
don't
it's
still
covered,
but
it
would
be
a
better
nicer
gesture
if
we
explicitly
add
licenses
to
the
header
so
that
if
somebody
accidentally
copies,
they
would
still
remember
what
that
original
license
of
the
quote
was,
I
think,
okay
yeah,
but
I
think
that
makes
sense
then
yeah,
it's
more
like
nice
to
have,
but
it's
not
required
from
the
legal
standpoint.
B
We
can
yeah,
I
don't
think
we
should
enforce.
We
can
give
a
hint
to
people
who
add
during
code
review
manually-
hey,
please
add,
but
we
don't
have
to
enforce
because
legally
it's
already
clean.
A
Okay,
I
think
could
reveal
to
copper,
should
cover
that
okay,
yep,
okay,
we
can
go
through
their
recent
issues
quickly.
Let
me
show
you
on
the.
B
A
B
B
Let
me
take
a
look
because
I
scoop
yeah.
A
B
A
B
No
there's
no
need
for
that.
It
was
more
like.
I
think
I
wanted
to
show
different
time.
Timestamps,
that's
the
only
true
reason,
but
we
can
remove
it
from
the
test
suite.
It's
not
important.
A
B
Remove
those
sleeves
those
slips,
I
only
relied
on
these
to
actually
visually
inspect
the
delta
between
timestamp
generate,
but
let's
clean
it
yeah.
I
agree:
okay,
okay,
sleep.
A
A
B
Because
I.
A
I
have
a,
I
have
one
question
when
look
at
this,
like
for
their
script
like
a
setup,
build
build
gg
test,
but
it
installs
the
build
artifact
artifacts
into
temp
folder.
Not
I
I
can.
B
Was
needed,
the
only
reason
why
it
was
needed
is,
if
you
run
on
ubuntu
14,
like
old,
you
wouldn't
from
2014.
B
It
may
have
google
test
installed
already
like
with
debian
package
manager
and
that
that
version
was
all
too
old
and
it's
like
so
we
we
have
an
issue
here
that
older
distros
may
already
have
g
test,
but
it's
not
good
enough
for
us,
because
it's
too
old,
so
what
I
was
doing,
I
was
building
a
local
g-test
just
in
the
build
3
kind
of
thing
and
linking
against
that
one
by
explicitly
specifying
the
path
to
that
local
g-test.
B
Because
that's
like
alternatives,
how
do
we?
How
do
we
handle
multiple
versions?
If
the
distro
comes
with
old
version?
How
do
we
install
a
temporary,
more
recent
version
of
it,
and
I
didn't
want
to
install
it
under
user
local,
so
I
use
the
temp.
Maybe
we
should
use
the
cmake
binary
output
directory
or
something
like
so
that
when
we
build
the
project,
we
put
a
dependency
into
a
sub
directory
of
the
build
3
and
link
against
that.
B
I
can
take
a
look
at
my
machine.
That's
the
only
case
is
when
distro
comes
with
the
version.
That
is.
A
B
A
B
B
Build
root
variable,
or
at
line
eight,
nine.
Yes,
that's
right.
B
I
had
this
invoked
from
the
docker
image,
so
maybe
let
me
think
maybe
I
should
write
the
document
that
explains
how
to
use
that
script
and
where
it
is
invoked,
because
I
give
that
freedom,
whether
I
build
an
under
tmp
build
or
if
I
respect
the
build
root
variable
which
has
to
be
set
somewhere
outside
and
then
I
can
actually
install
this
custom
g
test
into
the
same
directory.
B
Where
b
we
built
the
rest
of
it
and
the
again.
The
only
scenario
where
this
is
needed
is
if
we
are
running
on
very
old
system,
because
if
we
are
running
on
fresh
ubuntu,
like
ubuntu
20,
for
example,
most
likely,
we
should
allow
to
install
a
g
test
that
comes
with
the
distro,
like
whatever
apt,
get
and
style,
google
tester.
Somehow
like
latest
and
okay.
We
usually
work
fine
with
latest
again.
A
I
see
so
on
the
and
the
old
build
ubuntu
build.
If
the
user
has
also
an
old
g
test,
we
need
it
is
building
using
that
wall.
The
build
process
give
the
error
message
explicitly
or
like
your
gta.
B
Suggestion
I
think
I
have
seen
this
not
only
with
g
test,
but
also,
I
think,
with
protobuf
protobuf
was
also
of
an
old
version
in
ubuntu
in
ubuntu
1404.
B
I'm
just
thinking
here
I
can
see
how
how
we
can
avoid
using
that
build
g
test.
Maybe
by
documenting
it
I
see.
Yes,
that
will
be
very
helpful.
Yeah
see,
that's
the
end.
The
other
thing
that
I
mentioned
proto
pro
protobuf.
A
B
Yeah,
so
for
this
one:
yes,
I'm
just
thinking
here
what
other
better
options
we
have.
B
A
B
B
It's
the
same
person
we
can.
We
can
help
out.
A
B
At
it
yeah
we
don't
support
conan.
We
we
have
tried
vc
package.
We
support
vc
package.
We
may
be
able,
in
future
to
add
support
for
for
concrete
distro
packaging
like
deb
or
rpm.
B
Here
I
would
say
if
the
person
knows
how
to
fix
this,
I
would
rather
expect
a
pr
from
them,
because
formally
we
should
say
we
do
not
currently
have
plans
to
support
conan
and
if
you
have
ideas
how
to
fix
it.
Please
provide
the
patch
and
we.
B
Be
relatively
low
priority
to
us
right
now
because,
again
with
support
bazel,
we
support
cma,
and
we
have
a
portfolio
for
vc
package
wrapper
on
top
of
cma.
So
we
do
have
at
least
three
build
systems
and
should
cover
a
lot
yeah
and
supporting
yet
another.
Our
build
system
to
us
is
probably
a
bit
of
an
overkill,
at
least
not
for
the
for
the
first
release.
B
A
B
Are
from
lolita
about
the
compliance.
B
The
other
two
jagger
compliance.
A
B
The
support
matrix
here
my
feedback
was
as
follows:
a
person
was
trying
to
build
our
sdk
with
a
compiler
that
didn't
even
have
support
possible
and
his
proposal
was
to
fix
the
language
standard
at
11..
B
I
think
okay,
and
it's
like
I
don't
know
if
there's
an
option
in
c
make
that
allows
to
say
I
want
mil
11,
but
anything
above
I
would
work
either
with
17
or
20,
and
so
basically,
I
recommended
the
person
to
maybe
create
a
document
which
describes
what
operating
systems
and
what
compilers
we
are
known
to
work
with
and
having.
That
document
is
a
good
reference
point
we
can
share
with
others.
B
B
Yeah
and
I
I
I
re-run
locally
and
right
now-
it's
so
good
with
my
latest
pr
that
moves
to
upsell
and
cleans
up
the
variant.
So
we
are
one
step
closer
to
actually
add
the
stdlib
to
ci.
So
it's
going
to
be,
like
probably
add
the
latest
windows
and
I'll
add
latest
linux,
to
verify
that
c,
plus,
plus
20
works
well
and
all
tests
pass
and
I
got
there.
A
B
That's
why
I
request
the
changes.
I
don't
want
that
person's
behind
merged.
A
Okay
on
us,
this
one.
Let's
take
a
look.
A
B
Yeah
yeah,
I
have
I
understand
that
we
would
record
the
message
and
probably
not
like
you
know
in
in
java,
it's
easier
to
get
a
back
trace
and
in
most
other
languages
it's
easier
to
get
a
backspace
in
c
plus,
plus
it's
not
as
easy.
I
guess
the
best
we
can
get
is
the
well.
I
guess
we
may
have
an
address,
and
that
is.
B
Yeah
yeah
I
mean
there
are
usually
other
libraries
being
used
for
unwinding
like
google
breakpod
and
all
these,
like
exception
is
captured,
but
I
I
don't
think
we
can
afford
integrating
with
that.
So
from
our
perspective,
it's
probably
going
to
be
just
a
message
and
just
the
address.
A
B
Yes,
it's
just
I
I
remember
that
for
the
crashpad
great
pad
there
was
that
need
to
capture
the
symbols
first,
and
then
there
was
a
separate
unwinder
library
and
from
my
olden
times
I
remember
that's
not
always
trivial
task
across
all
possible
platforms,
including
arm
platforms
to
reliable
and
consistently
unwind.
B
A
B
Yeah
anyways:
let's
see
what
exactly
is
expected,
I
mean
in
other
languages
just
easier
because
in
other
languages,
you'd
probably
get
a
structured
list
of
entries,
addresses
and
function
names.
I
don't
know
like
in
python
or
in
java.
You
get
that.
C
But
you.
A
B
Like
an
array
of
addresses
or
functions
and
file
names
and
line
numbers
and
all
that
beautiful
structure,
but
in
supply
spots,
not
as
easy.
A
A
So,
what's
the
problem
of
std
copy,
is
it
security
yeah?
They
run
with
the.
A
B
Also,
a
list
of
banned
methods
that
they
don't
want
to
see
so
there's
no
problem
per
se,
because
here
we
know
that
object,
size
and
we
know
the
destination
object
side.
So
it's
something
that
otherwise
you
can
go
through
security
and
argument
that
it's
safe
because
it's
not
doing
anything
bad,
but
it
would
be
easier
to
replace
it
with
the
for
loop,
for
example,
okay,
nice
and
then
it's
no
longer
gonna
be
tripping,
triggering
a
warning
here.
B
It's
it's
silly
because
it's
not
a
real
issue.
It's
an
issue
with
the
code
analysis,
tooling
and
I'd
clean
it
up.
B
I
know
it's
still.
I
know
it's
silly
because,
but
then
it's
like,
I
know
that
the
array
is
of
that
size,
I'm
going
to
loop
over
the
array
size,
and
I
know
that
the
original
object
is
eight
bytes.
A
And
this
is
only
for
eight
bytes,
so
maybe
just
a
for
loop
is
fine,
you
know,
but
for
bigger
for
bigger
range.
I
think,
but
yeah
that
could.
A
B
What
what
I
don't
like
is,
I
see,
watched
other
libraries
like
I
looked
at
json
http
and
I
looked
at
google
upsell
and
they
used
this
construct
elsewhere.
B
So
if
this
code,
checker,
which
is
too
straight,
is
run
on
the
other
parts
that
we
always
use
is
still
gonna
complain,
I
mean
for
now
for
etw
exporter
being
header,
only
library
that
does
not
require
json
hvp
and
that
does
not
require
or
gop
and
does
not
require
upsell
per
se.
I
might
be
able
to
just
replace
this
with
the
forward,
but
in
general
I
don't
think
that
the
the
code
analysis
checker
that
they
are
using
is
a
good
checkup.
I
think
it's
overly
paranoid
sitting
that
they
have
in
there.
A
B
Just
the
st
yes
and
right
now,
immediate
issue
is
only
this
part
was
checked
and
that's
why
only
this
part
was
triggering
the
the
warning
okay,
but
if
they
check
the
entire
code
base,
I
think
they
would
still
see
an
issue
like
in
jason
hpp,
but
in
in
my
case
that
specific
customer
uses
just
that
header.
They
don't
need
the
other
parts.
B
That's
why
the
other
parts
don't
trigger
warnings,
but
if
someday
they
used
your
pc
exporter,
I
think
they
would
still
hit
some
of
these
similar
issues
and
anyways
I'd
like
to
solve
this
immediately.
Just
just
here
and
then
later.
We
when
we
hit
another
issue
like
this,
we
can.
We
can
talk
the
other
one.
B
A
Okay,
since
we
covered
the
issues
in
the
last
week,
yep
so
in
a
while
running
out
of
time-
and
I
will
comment
a
few
issues-
you
mentioned
yeah
with
follow-ups.
B
Right,
yes,
okay,
if
you
can
take
a
look
at
my
big
pr
with
the
variant.
B
Take
a
look-
and
I
still
have
that
fool
up
to
verify
the
latest
snapshot
of
upsell
I'll.
Do
that.