gtk-jamesa-971010 considered bogus

[Tor Lillqvist <tml iki fi>]

I looked into a problem with the file selector on Win32, and noticed
that the behaviour of correct_dir_fullname() and correct_parent() in
the case of going to ".." didn't seem quite logical. Also, the
comments in correct_parent() didn't match the semantics (the "hard
way" was done every time). So, I checked with cvs and noticed that the
suspect stuff had been introduced in gtkfilesel.c's version
1.2. What's more, the original version of this code corresponded
exactly to what I had been thinking how the code should look...

(Note: This is all on Unix, it isn't Win32-related.)

So, I tried to reverse the patch, and now the file selector IMHO works
better. Specifically, if you with the current fileselector go
somewhere via double-clicking on a symlink, and the go up again by
double-clicking on "..", you end up in the parent of the symlinked-to
directory. Without gtk-jamesa-971010, you end up where you came from,
which I assume is the originally intended behaviour.

I also couldn't notice any strange behaviour with the original code as
alleged by gtk-jamesa-971010.patch.README. (Can be found in
ftp.gtk.org/pub/patches/old/1997, IIRC.)

If nobody disagrees, I will commit a reversal of said patch. (Except
for the line that sets cmpl_state->directory_storage = NULL in
prune_memory_usage(), as that's not related to this.)

--tml