Re: warn about conflicting skeleton-generated files

[Joel E. Denny]

On Thu, 18 Jan 2007, Paolo Bonzini wrote:

> No Unix utility behaves like this with respect to data files (even
> though your comparison with how shells look in the PATH is right).

info --file=FILE

> IMHO, the right way to do it would be: If the file name does not contain
> a `/', and is not found in the current directory, it's one of Bison's
> installed files.  This is the same as include paths in C, and it fits
> better than the comparison with shells looking up the commands.

Hans Aberg, Paul Eggert, and I already discussed this possibility.  To 
summarize, I prefer the current logic because it's very simple and 
backward-compatible.  The C include path is a different animal with more 
complex needs.

> This is
> partially backwards-incompatible, in case someone has a lalr1.cc file in
> their current directory, but people should switch to %language anyway.

That seems dangerous to me.

> Also, you're not modifying -S if I understand correctly.

-S and --skeleton should work right.  See the test cases.