bug#17476: 24.3.91; [PATCH] unimplemented info entry

[Eli Zaretskii]

> Date: Mon, 12 May 2014 19:54:05 +0300
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: 17476@debbugs.gnu.org
> 
> > Cygwin info command treat them as info files.
> 
> No, it doesn't.  A reference such as 'foo::' is parsed as a _node_
> named "foo" in the 'dir' file.  Info then searches for such a node in
> the file 'dir', and doesn't find it.  It then falls back on a kludgy
> feature, whereby typing "info foo" from the shell prompt looks for a
> _file_ 'foo' or 'foo.info', on the assumption that the
> bozo^H^H^H^Huser forgot to type -f, as in "info -f foo".
> 
> IOW, the above dir file abuses a kludge in the Info reader introduced
> to shut up all those who complain because they cannot be bothered to
> invoke the reader as prescribed in the documentation.
> 
> In any case, this kludge is for invoking the Info reader from the
> shell prompt, not for navigating inside Info files.  IOW, this is
> actually a buglet in Info, which therefore can disappear without
> notice.
> 
> > With the attached patch, Emacs regards the above format in a dir node
> > as info file names.
> 
> This patch is incorrect.  It treats the dir file specially, whereas it
> is simply an Info file with a menu, and nothing prevents the Cygwin
> maintainers from putting such menu items in any other Info files.
> More importantly, it tramps on a useful feature, whereby a reference
> such as "foo::" is actually a shortcut for "(dir)foo::".
> 
> So I hope we will not accept this change.

Maybe I should explain why I'm so agitated about this.  You see, I
wrote the code which is responsible for interpreting "foo::" as a
reference to a file "foo" if there's no menu item by that name.  I
wrote that code, under protest, more than 12 years ago, only because
the head maintainer asked for that, in order to stop complaints that
"info SOMETHING" didn't show the manual for that SOMETHING.

So imagine how I feel 12 years later when I see that a kludge
concocted in order to cater to some user sloppiness is now treated as
a first-class feature, and Emacs is requested to follow suit.

Sorry if my response sounded harsh.