Use R0/R1 instead of R4/R5 for eh_return on ARM

Use R0/R1 instead of R4/R5 for eh_return on ARM

This should unblock removal of  _Unwind_PNaClSetResult{0,1} entirely.

BUG= https://code.google.com/p/nativeclient/issues/detail?id=4171
R=jvoung@chromium.org

Committed: https://gerrit.chromium.org/gerrit/gitweb?p=native_client/pnacl-gcc.git;a=commit;h=a2fe95dbfd10df1b8d4e375e4668dc1e98e237e7

[Derek Schuff, 2015-06-03 00:28:28]

dschuff@chromium.org changed reviewers:
+ jvoung@chromium.org

[jvoung (off chromium), 2015-06-03 17:21:05]

lgtm

https://codereview.chromium.org/1151383009/diff/1/gcc/unwind-pnacl.h
File gcc/unwind-pnacl.h (left):

https://codereview.chromium.org/1151383009/diff/1/gcc/unwind-pnacl.h#oldcode100
gcc/unwind-pnacl.h:100: #else
I wonder if mips is going to have the same problem...

[Derek Schuff, 2015-06-03 20:51:11]

https://codereview.chromium.org/1151383009/diff/1/gcc/unwind-pnacl.h
File gcc/unwind-pnacl.h (left):

https://codereview.chromium.org/1151383009/diff/1/gcc/unwind-pnacl.h#oldcode100
gcc/unwind-pnacl.h:100: #else
On 2015/06/03 17:21:04, jvoung wrote:
> I wonder if mips is going to have the same problem...

Yeah, it seems likely. I still have the suspicion that there was something that
was kind of mitigating this problem before, that maybe isn't working now.
Because those registers were always getting clobbered before, and it seems kind
of inconceivable that there were never any situations where they were being used
until now.

[Derek Schuff, 2015-06-03 20:51:55]

Committed patchset #1 (id:1) manually as
a2fe95dbfd10df1b8d4e375e4668dc1e98e237e7 (tree was closed).