On Sun, Jan 15, 2023 at 11:48 AM Masahiro Yamada <masahiroy@xxxxxxxxxx> wrote:
>
> On Sat, Jan 14, 2023 at 8:15 AM Miguel Ojeda
> <miguel.ojeda.sandonis@xxxxxxxxx> wrote:
> >
> > On Thu, Jan 12, 2023 at 7:23 AM Masahiro Yamada <masahiroy@xxxxxxxxxx> wrote:
> > >
> > > Maybe, your purpose is to use sed consistently, but
> > > perhaps you can avoid forking sed if you know the
> > > format of the first line.
> >
> > The most unknown format would be the one of the libclang check, where
> > there may be other lines before the one we are interested in. However,
> > the pattern expansion would still match newlines, right?
> >
> > > BTW, what is missing here is, you do not check if
> > > ${RUSTC} is really rustc.
> > >
> > > I can fool this script to print
> > > "arithmetic expression: expecting primary: "100000 * + 100 * + "
> >
> > We can test if nothing was printed by `sed` for that (or do it with
> > shell builtins).
> >
> > Having said that, I would say fooling the script on purpose is an more
> > of an oddity compared to the case `MAKEFLAGS` attempts to cover
> > (please see my reply on the other patch). So if we cover this, then I
> > would say we should really cover the other one.
>
>
>
> get_canonical_version() in scripts/as-version.sh has
> a little more trick to avoid
> "arithmetic expression: expecting primary: "100000 * + 100 * + "
> but it is up to you.
My code accepts anything that is separated by dots
(and non-numerical strings are silently turned into zero).
Your code takes exactly the "([0-9]+\.[0-9]+\.[0-9]+)" pattern,
so it works very safely.
I think using sed is fine.
--
Best Regards
Masahiro Yamada
