[PATCH] Fix PR target/6522 (dwarf-2 regression from 2.95.x)

[Mark Mitchell]

> The following patch fixes this by doing the same adjustements
> on DECL_ALIGN as stor-layout.c does.

Ugh, ugh, ugh.  If DECL_ALIGN is not telling the truth, that is just
plain disgusting.  The correct fix is to have place_field and 
place_union_field update DECL_ALIGN for the field after they decide
what alignment they're going to give the field.

For the mainline, at least, that should be the fix, so I reject your
patch for the mainline. :-)  For the branch, it's OK; I'm a little
nervous about what actually setting DECL_ALIGN correctly might break.

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com