On 08/16/2010 05:25 PM, Bruno Haible wrote: > Eric, > > Paul Eggert wrote: >> unless this offsetof bug occurs in real code, it might >> be better to not worry about it. > > I agree. It's the first time in my life that I see someone doing > 'sizeof' of an 'offsetof' expression. It's not a realistic use of the > 'offsetof' macro.
Other than to test compliance to other aspects of offsetof. Are there any other C expressions besides the sizeof operator where injecting (cast)(expr) will cause a different parse than injecting ((cast)(expr))? > > And the workaround is trivial: just add parentheses. It also makes the > code easier to read: > sizeof (offsetof (struct d, e)) Then perhaps the simpler patch is to document the bug, but not test for it, by simply committing test-stddef.c with redundant parenthesis; while remembering this thread in case we do find other platforms that do have more severe bugs with using the wrong type. -- Eric Blake ebl...@redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
