bug#9891: 24.0.90; Duplicated entry at the info directory

[Stefan Monnier]

>> I'm not sure what is being discussed.  Is it that the two above lines
>> are kept, or that one of the two is removed as a duplicate?
> That it is removed.
>> And why is the other behavior better?
> Because these are 2 different glibc functions.
>> Stefan "To my layman's eyes, the two lines are equivalent, just
>> like "* GDB: (gdb)" and "* Gdb: (gdb)" and should be
>> collapsed into one"
> Evidently, this is not always true.  Leaving redundant information is
> a lesser evil than removing non-redundant one, IMO.

I think I'm beginning to understand: the problem is that the "top-level
dir" is used in two very different contexts:
- for interactive use, where it should be shortish and avoid redundancy.
- for non-interactive use, typically to make `info' emulate `man', where
  the toplevel `dir' is abused as an index.  This only works for
  specially built `dir' files and I don't see it used in Debian,
  for instance.
The Info-dir-remove-duplicates is clearly meant for the first use and
I think that's the most important use, so if we want to make it handle
the second case we'll have to make sure it doesn't hurt the first.


        Stefan