This patch actually fixes the bogus RTL seen in PR101697.


Basically we continue to use the insn condition to catch most of the 
problem cases related to autoinc addressing modes.  This patch adds 
constraints which can guide reload (and hopefully LRA) away from doing 
blind replacements during register elimination that would ultimately 
result in bogus RTL.  The idea is from Paul K who has done something 
very similar on the pdp11.  I guess it shouldn't be a big surprise that 
the H8 and pdp11 need the same kind of handling given some of the 
similarities in their architectures.


Anyway, this has been tested in my tester without regressions. In fact, 
it fixes several bugs where the testsuite was tripping over the same 
problem.  Given this issue is covered by the testsuite, I haven't added 
a new test.


Pushed to the trunk.

Jeff