Re: rcsdiff + revision symbolic names BUG

[Romain Francoise]

"Aaron S. Hawley" <address@hidden> writes:

> A potential patch that fixes the behavior is attached.  It deletes all
> instances of a flag variable named `no_diff_means_no_output'.  It's
> not clear what its purpose is.

If diff(1) is called with -y or -D, it produces output even if the files
are identical.  RCS recognizes this behavior and doesn't optimize out
the case where it knows that the files/revisions will be identical if -y
or -D are used (this is the purpose of `no_diff_means_no_output').

Your patch removes this 'intelligence' and as a result, rcsdiff will
always skip diff generation when the revisions are identical, which is
wrong if the user specified -y or -D.

The "bug" is that RCS considers a long option name as a case where we
don't know if diff(1) will produce output or not; indeed if the long
options are --side-by-side or --ifdef (the long counterparts of -y and
-D) then there will be output even if the revisions are identical.

-- 
  ,''`.
 : :' :        Romain Francoise <address@hidden>
 `. `'         http://people.debian.org/~rfrancoise/
   `-