[Bug middle-end/101836] __builtin_object_size(P->M, 1) where M is an array and the last member of a struct fails

["siddhesh at gotplt dot org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101836

Siddhesh Poyarekar <siddhesh at gotplt dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |siddhesh at gotplt dot org

--- Comment #5 from Siddhesh Poyarekar <siddhesh at gotplt dot org> ---
Vim explicitly disables _FORTIFY_SOURCE to keep its use of 1-sized trailing
arrays:

https://github.com/vim/vim/issues/5581

so either they haven't tested with more recent gcc or they're hitting a corner
case where __builtin_object_size does return the subscript value for the
trailing member.

I inherited the __builtin_object_size behaviour in
__builtin_dynamic_object_size to remain consistent with current behaviour:

ok:  sizeof(*working) == 24
ok:  sizeof(working->c) == 16
ok:  __builtin_object_size(working, 1) == -1
ok:  __builtin_object_size(working->c, 1) == 16
ok:  __builtin_dynamic_object_size(working, 1) == -1
ok:  __builtin_dynamic_object_size(working->c, 1) == 16
ok:  sizeof(*broken) == 24
ok:  sizeof(broken->c) == 16
ok:  __builtin_object_size(broken, 1) == -1
WAT: __builtin_object_size(broken->c, 1) == -1 (expected 16)
ok:  __builtin_dynamic_object_size(broken, 1) == -1
WAT: __builtin_dynamic_object_size(broken->c, 1) == -1 (expected 16)


However in theory if the pass can see the allocation, it should be able to
build the right expression for object size.

I'm updating the patchset to meld the two passes into one and I could add
-fstrict-flex-arrays as one of the patches to make this stricter.