On 03/16/2014 11:42 AM, Stefan Weil wrote: > We have a similar situation with other attributes, too. Smatch also > complains about missing QEMU_NORETURN in our implementation code (we add > it only in header files for global functions). Do you think it would be > good to make header and implementation match there, too?
Not without changing our CODING_STYLE. I'm not really fond of void attributes some_function(arg, arg, arg) since its quite easy to wind up with too long lines. It's one of the things I really like about the gnu style void attributes some_function(arg, arg, arg) layout. (In addition to being an old codger who really liked to be able to do "grep ^some_function *.c" to find the implementation.) r~