The design of the attribute exclusion framework includes
support for different exclusions applying to different
kinds of declarations (functions, types, and variables
or fields), but the support is incomplete -- the logic
to consider these differences is missing. This is
because the differences are apparently rare. However,
as the bug below points out, they do exist.
PR middle-end/84108 - incorrect -Wattributes warning for
packed/aligned conflict on struct members, shows that while
declaring a non-member variable aligned is enough to reduce
the its alignment and declaring it both aligned and packed
triggers a -Wattributes warning:
int a __attribute__((packed, aligned (2))); // -Wattributes
a struct member must be declared both aligned and packed in
order to have its alignment reduced. (Declaring a member
just aligned has no effect and doesn't cause a warning).
struct S {
int b __attribute__((packed, aligned (2)));
int c __attribute__((aligned (2))); // no effect
};
As a result of the incomplete logic GCC 8 issues a -Wattributes
for the declaration of b in the struct.
By adding the missing logic the attached patch lets GCC avoid
the spurious warning.
I considered adding support for detecting the ineffective
attribute aligned on the declaration of the member c at
the same time but since that's not a regression I decided
to defer that until GCC 9. I opened bug 84185 to track it.
Tested on x86_64-linux with no regressions.
Martin