I have a target where, due to a gcc defficiency, one cannot set a breakpoint at 
the very end of a function.

This causes problems in the gdb testsuite because several tests presume one can 
set a breakpoint or step to the final } of a function.  The tests which do this 
are not actually testing for that functionality, but merely using it.  For my 
target, this leads to a lot of spurious test failures.

Although it would be good to fix gcc, that's not practical right now, and I 
thought it would be better to defensively code the gdb testsuite anyway.  To 
that end I

*) added an explicit test in gdb.base/break.exp to test for breaking at the end 
of a function.  IMHO, that is where such a test belongs
*) recoded the spurious fails to add real code just before the end of the 
function in order to guarantee we'd stop there.
*) fileio.exp was one such test, it has a lot of explicit line numbers in the 
.exp file, making adjustment of the .c file practically impossible.  I've 
recoded it so that we have a single breakpoint in a 'stop' routine.

ok?

nathan

-- 
Nathan Sidwell    ::   http://www.codesourcery.com   ::         CodeSourcery
nathan@codesourcery.com    ::     http://www.planetfall.pwp.blueyonder.co.uk