This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [patch 2/9] Code cleanup: Drop IS_ABSOLUTE_PATH checks
On Fri, 18 Jan 2013 20:20:24 +0100, Eli Zaretskii wrote:
> > Date: Fri, 18 Jan 2013 19:39:38 +0100
> > From: Jan Kratochvil <jan.kratochvil@redhat.com>
> > Cc: gdb-patches@sourceware.org
> >
> > On Fri, 18 Jan 2013 08:32:25 +0100, Eli Zaretskii wrote:
> > > > int
> > > > compare_filenames_for_search (const char *filename, const char *search_name)
> > > > @@ -171,7 +171,8 @@ compare_filenames_for_search (const char *filename, const char *search_name)
> > > > to put the "c:file.c" name into debug info. Such compatibility
> > > > works only on GDB built for DOS host. */
> > > > return (len == search_len
> > > > - || IS_DIR_SEPARATOR (filename[len - search_len - 1])
> > > > + || (!IS_ABSOLUTE_PATH (search_name)
> > > > + && IS_DIR_SEPARATOR (filename[len - search_len - 1]))
> > > > || (HAS_DRIVE_SPEC (filename)
> > > > && STRIP_DRIVE_SPEC (filename) == &filename[len - search_len]));
> > > > }
> > >
> > > I don't understand why the "match up to a slash" rule is now limited
> > > to non-absolute file names.
> >
> > FILENAME may contain for example: /path/to//file.c
> > Then we may request to put a breakpoint to: /file.c:main
> > And without
> > the '!IS_ABSOLUTE_PATH (search_name) &&' part it would falsely match.
>
> Thanks for explaining. However, IMO i's wrong to test for double
> slash with IS_ABSOLUTE_PATH, because the same could happen with
> foo//bar.c, right? So why not explicitly test for consecutive
> slashes instead?
I do not understand much what do you suggest.
Remaining problem is that /path/to//file.c does not match /path/to/file.c or
that /path/file.c does not match /path/to/../file.c and so on.
I find these bugs outside of the scope of this patchset, this patchset only
tried to handle correctly /path/to/file.c matching, which was big enough IMO
to post it and upstream.
This specific patch is "Code cleanup" so it should have no visible changes, it
is only simplification of the caller-callee implementation so it mimics in the
callee (compare_filenames_for_search) what every of the callers was doing.
Thanks,
Jan