https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87533
Bug ID: 87533
Summary: bogus assume_aligned attribute silently accepted
Product: gcc
Version: 9.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: msebor at gcc dot gnu.org
Target Milestone: ---
The manual describes the assume_aligned attribute as follows:
The assume_aligned attribute is used to tell the compiler that the function
return value points to memory, where the returned pointer minimum alignment is
given by the first argument. If the attribute has two arguments, the second
argument is misalignment offset.
Clearly, the first argument must be a power of two, and second argument should
presumably be greater than (or perhaps equal to) zero and less than the value
of the first argument. Finally, the attribute only makes sense on functions
that return a pointer. Yet GCC silently accepts the following non-sensical
declaration:
$ cat c.c && gcc -S -Wall c.c
__attribute ((assume_aligned (-1, -2))) void f (void) { }
Clang, on the other hand, issues:
c.c:1:15: warning: 'assume_aligned' attribute only applies to return values
that are pointers or references [-Wignored-attributes]
__attribute ((assume_aligned (-1, -2))) void f (void) { }
^~~~~~~~~~~~~~~~~~~~~~~ ~~~~
After the return type is changed to void*, Clang then issues the following:
c.c:1:15: error: requested alignment is not a power of 2
