On 07/06/2017 16:29, Jakub Jelinek wrote: > On Wed, Jun 07, 2017 at 05:16:44PM +0300, K wrote: >>> ui16 *ptr = (ui16*)buf; >>> >>> There's no need for any of this messing about with pointer casts, as has >>> been explained. >> >> Sorry, but I still can't get the idea. Cast from udp_pseudo to uint8_t >> doesn't have an aliasing problem (std 8.8) and a cast from uint8_t to ui16 >> still doesn't have an aliasing problem (std 8.6), or may be I missed >> something? > > This is not the right list to learn C or C++, so this should be moved > to gcc-help. The casts themselves are not the points of UB (unless the > pointer is misaligned for the pointee type you cast to), the problem is that > you have an object (tmp and its fields) that doesn't have unsigned short as > its dynamic type, nor cv-qualified version thereof, nor similar type, etc. > (see [basic.lval]/8 for details) and you are accessing that object > using a glvalue with unsigned short type. And that invokes the UB: > "If a program attempts to access the stored value of an object through a glvalue > of other than one of the following types the behavior is undefined:" > ... long list of what is allowed. > > Accessing it e.g. using glvalue of char, unsigned char or std::byte would be > fine, or say using ui16 __attribute__((may_alias)) type (by having > typedef ui16 ui16a __attribute__((may_alias)); > ui16a *ptr = (ui16a*)buf; > ... *ptr ... > or similar, or union that includes the type of tmp and ui16 as type of its > members), etc. Anyone messing around with pointer casts (other than pointer to first field of struct foo cast to/from pointer to struct foo) should be using -fno-strict-aliasing (as is done in Linux) https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fstrict-aliasing If that makes any difference, then likely UB has been tickled. Regards.
