Re: [RFA] Should openp open directories?

[Joel Brobecker <brobecker@ACT-Europe.FR>]

> I think this change is a good idea, but I have one comment: if stat 
> fails, isn't it better to return non-zero?  Yes, I know: it shouldn't 
> happen, but if we return non-zero in that case, we keep 
> back-compatibility in case there are some obscure filesystems when that 
> could happen.

It seems a sensible idea to return True when stat fails, except when the
file does not exist. I suggest the new following implementation (rough
code, not compiled):

   status = stat (filename, &st);

   /* A comment here explaining why we return true when errno != ENOENT.
      Or maybe in the function description ??? */
   if (status != 0)
     return (errno != ENOENT);

What do you think?

> Also, did you try your patch when there's a directory by that name, but 
> no executable program?  If so, what does the patched GDB say and/or do?

Yes. I tried "./gdb/gdb bfd" at the root of the gdb source directory and
got the following error message:

          bfd: No such file or directory.

The "or directory" part may be confusing...

-- 
Joel