r99402 causes x86_64 regressions

[Michael Hope]

lp:gcc-linaro/4.5+bzr99402 causes va_args related regressions in the x86_64 backend. Investigate and re-merge.

Note that r99402 was reverted for the 2010.11 release.

[Chung-Lin Tang (cltang)]

On mainline trunk, this was fixed at rev.162384,
see discussion here: http://gcc.gnu.org/ml/gcc-patches/2010-07/msg01496.html

After re-applying the r99402 backport, plus this patch, Linaro 4.5 x86-64 now bootstraps and passes those failed varargs cases.

There was also a follow up patch in the same thread here, fixing PR45027:
http://gcc.gnu.org/ml/gcc-patches/2010-07/msg01798.html
However, it does not seem to be needed for Linaro 4.5, as I currently see...

[Michael Hope (michaelh1)]

Looks good. See:
 http://ex.seabright.co.nz/helpers/testcompare/gcc-linaro-4.5+bzr99436~cltang1/logs/x86_64-maverick-cbuild17-crucis/gcc-testsuite.txt?base=gcc-linaro-4.5-2010.11-1

for a comparison against the last release.

I'll also push this into my standard build machine to see if the limits failures are real or not.