[tree-ssa] mainline merge as of 2004-04-10

[David Edelsohn]

>>>>> Diego Novillo writes:

>> If that is the case, then the coverage from the bootstrap tests are
>> meaningless.

Diego> With respect to hot/cold?  Of course it is.  Just as it is meaningless
Diego> for any of the dozens of -f options that are not enabled by default.  I
Diego> fail to see what you are trying to point out here.

	The patch itself caused the failures, not the optimization.  One
did not need to test with the commandline option enabled.

>> All of the breakage from the hot/cold partitioning patch is now in
>> the tree-ssa branch.

Diego> Yes.  Are you experiencing new problems because of it?

	As Andrew pointed out the original hot/cold partitioning patch,
without the follow-on patches developed over the weekend, was broken on
all targets except *-gnu-linux and *-darwin.

	The merge was bootstrapped on five targets, none of which exposed
a known bootstrap failure.  Diversity in all the target axes would help
achieve better coverage for testing.

David