bug#8254: race condition in dired.c's scmp function

[Paul Eggert <eggert@cs.ucla.edu>]

On 03/16/2011 06:19 AM, Richard Stallman wrote:
> I see that uppercasep uses inline.  That works fine in GCC, but does
> it work in all compilers anyone wants to use?

I expect so, yes.  That is, for all compilers that anybody
wants to use,  I expect that typically the performance difference
will be so small that it will be hard to measure and will not be
worth worrying about.  I attempted to simulate this by compiling
without optimization, both without and with the change,
and measured roughly a 3% performance degradation on a simple
large test case (running emacs interactively to upcase an 80 MB
text file that was mostly lower case).

This is sort of the worst case, since it assumes no inlining.
In the normal case, when optimization is enabled, I observed
a 3% performance improvement (on the same benchmark) due to
the change.

These measurements are noisy and approximate, but they indicate
that there's not much performance impact either way in practice.