Re: [PATCH] testsuite: prune DOS drive letter in test outputs

["Clément Chigot" <chigot@adacore.com>]

Hi Jan,

On Wed, Feb 22, 2023 at 9:26 AM Jan Beulich <jbeulich@suse.com> wrote:
>
> On 21.02.2023 17:14, Clément Chigot wrote:
> > --- a/binutils/testsuite/lib/binutils-common.exp
> > +++ b/binutils/testsuite/lib/binutils-common.exp
> > @@ -687,6 +687,22 @@ if ![string length [info proc prune_warnings]] {
> >      }
> >  }
> >
> > +# prune_dump_output OUTPUT
> > +#
> > +# Clean up the output from system specific or unwanted characters.
> > +# This allows to simplify the regexp inside dump tests.
> > +proc prune_dump_output { output } {
> > +    if [ishost "*-*-mingw*"] {
>
> Cygwin as well then?

I don't think so. AFAIK, Cygwin is using Unix path not DOS like ones.
Moreover, include/filenames.h seems to say the same. HAS_DRIVE_SPEC is
not defined for Cygwin.

> > +     # Prune DOS drive letter from an absolute path if it appears
> > +     # at the beginning of a line.
> > +     regsub -all "(^|\n)\[\[:upper:\]\]:" $output "\\1" output
>
> Drive letters aren't guaranteed upper case, are they?

I thought so, but it seems not. I'll change that.

> I also think this
> may end up too lax, as there may be messages which don't start with a
> path. I would expect that you want to also match "warning:", "error:",
> and alike infixes. Plus wouldn't you better match (but not remove) the
> first \ or / following the colon?

Yeah, that was my opinion too. But I had trouble with TCL regexps not
wanting to detect \... So, I still submit a first patch to have your
feedbacks on that. I'm not sure how common "LETTER:" can appear in
binutils tests, thus it might have been false concerns on my side. But
it seems not. I'll try to improve that.

> > --- a/ld/testsuite/ld-elf/noinit-sections-2.l
> > +++ b/ld/testsuite/ld-elf/noinit-sections-2.l
> > @@ -1,5 +1,5 @@
> >  #...
> > -(|.:)[^:]*: warning: orphan section `.noinit.var_noinit' from \S+ being placed in section `.noinit.var_noinit'
> > +[^:]*: warning: orphan section `.noinit.var_noinit' from \S+ being placed in section `.noinit.var_noinit'
> >  #...
> > -(|.:)[^:]*: warning: orphan section `.gnu.linkonce.n.var_noinit2' from \S+ being placed in section `.gnu.linkonce.n.var_noinit2'
> > +[^:]*: warning: orphan section `.gnu.linkonce.n.var_noinit2' from \S+ being placed in section `.gnu.linkonce.n.var_noinit2'
> >  #pass
>
> Did you commit the earlier patch, which now needs reverting?

No, this is not the same. I submitted two commits dealing with that issue.
 - ld/testsuite: handle Windows drive letter in a noinit test. which
was merged on Monday.
 - ld/testsuite: handle Windows drive letter in persistent section.
which ends up with this patch.

Thus I'm reverting the one that I've been already merged.

Clément