Re: [RFA/libiberty] Fix documentation issues in filename_cmp.c

[Eli Zaretskii <eliz at gnu dot org>]

> Date: Thu, 5 Apr 2007 23:12:18 -0700
> From: Joel Brobecker <brobecker@adacore.com>
> Cc: gcc-patches@gcc.gnu.org, gdb-patches@sourceware.org
> 
> Eli,
> 
> > > 2007-04-05  Joel Brobecker  <brobecker@adacore.com>
> > > 
> > >         * filename_cmp.c (filename_cmp): Improve documentation.
> > 
> > Thanks, I am happy, as far as the documentation goes.
> 
> Am I to understand that you are not happy with the code?

Well, I did have comments about that, and they were left unresolved ;-)

>   1. Fold multiple consecutive slash/backslash characters into
>      single slash/backslash;
> 
>      To do that without modifying the source filenames, we need to
>      allocate some memory locally to manipulate a copy of that filename.

This is a misunderstanding: I didn't mean that we should modify the
source file names.  What I meant is that the comparison should ignore
multiple consecutive slashes, so that, say, "/foo/bar/baz" compares
equal to "/foo//bar////baz" (and to "\foo//bar\/\baz" on Windows).
(However, note that, as Chris pointed out, double slash at the
beginning of a file name, as in "//foo/bar", are significant on
Windows.)

>      My take on this is that the file names we have seen, at least in
>      the debugging information, have been consistent; and thus this
>      enhancement would end up having no actual effect. I would wait
>      until we come across a case where this is a problem before
>      going that way.

I've seen quite a few of such situations, actually.  And in any case,
good engineering doesn't need examples to know what's Right ;-)

>   2. The possible introduction of a bug with certain locales because
>      the function used in place of strcasecmp uses tolower.
> 
>      For this one, as I said, I don't know, and I'm not sure where
>      to start looking for the answer.

How about using your function with LANG set to various values?  That
should at least tell us whether strcasecmp and tolower do the same
thing; if they do, there's no problem here.