[Bug tree-optimization/78428] [5/6/7 Regression] wrong code at -O2 and -O3 on x86_64-linux-gnu in 64-bit mode

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78428

--- Comment #6 from Martin Liška <marxin at gcc dot gnu.org> ---
Created attachment 40107
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=40107&action=edit
Tentative patch

Attaching tentative patch where I added bitregion_{start,end} to
store_constructor_field.
Actually there are multiple calls of the function, one for
{RECORD,UNION,QUAL_UNION}_TYPE (that's what we are interested in), another for
ARRAY_TYPE and one for VECTOR_TYPE. As the problematic mode reset (done in
expr.c:6230):

            mode = DECL_MODE (field);
            if (DECL_BIT_FIELD (field))
              mode = VOIDmode;

does not happen last 2 types, I hope we don't have to care. I'm attaching a
test-case and I'm wondering how to set a hard timeout of the test? Setting
dg-timeout just prints WARNING and the test is still running.

Thoughts?