GNATS cleanup: high-priority 3.2/arm bugs

[Joe Buck]

Hi,

I've closed all the PRs marked as high-priority 3.2 regressions (but
not 3.3/3.4), except for 8582 (which is still an ICE on the trunk),
and three ARM errors:

9790, 10194, 10195.

I checked the PRs that did not have a strong statement from Wolfgang
or someone else saying that the bug is fixed for 3.3 and the trunk.
Otherwise, I took Wolfgang's word.

Could someone with access to ARM hardware (or simulator, or who builds
cross-compilers) check these to see if they are present on 3.3 or the
trunk?  If not, they can be closed.