Re: Remove useless if-before-free and if-before-xfree tests

["Stephen J. Turnbull" <stephen@xemacs.org>]

Jim Meyering writes:
 > Kim F. Storm wrote:
 > > Jim Meyering <jim@meyering.net> writes:
 > >> I removed a bunch of these last year.
 > >> Since then, a few have snuck back in, so here's another round.
 > >> I've just committed the following two change sets:
 > >
 > > I haven't looked specifically at the changes, but a test before xfree
 > > may not be useless performance-wise if the normal case is "ptr is NULL".

These "normally NULL" cases are likely only in "last chance" cleanup,
which should be outside the loop.

 > Hi Kim,
 > 
 > If adding that extra "if" test-before-free leads to a measurable
 > performance improvement, please let me know.  I've removed hundreds
 > of such tests in many projects, and so far no one has taken the time
 > (or been able?) to demonstrate such a thing.
 > 
 > Besides, even if there is a measurable performance difference, in my book,
 > the cleaner code trumps what can only be a negligible performance hit.

"inline" is your friend.  If you don't trust GCC (and on this kind of
thing, who does? :-), you can force xfree to inline the test by
defining it as a macro.