http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50581
--- Comment #6 from Wolfgang at Solfrank dot net 2011-10-02 16:08:51 UTC --- (In reply to comment #5) > On Sat, 1 Oct 2011, Wolfgang at Solfrank dot net wrote: > > > > Passing va_list as a function argument is generally hard, whether or not > > > variadic, since you don't know whether it will be passed by reference or > > > by value or what the type of the address of a va_list parameter will be. > > > Portable code needs to pass a pointer to va_list or a structure > > > containing > > > va_list or use some other such means of avoiding dependence on whether > > > va_list is an array. > > > > Huh? What about vprintf and friends? They are defined to take a va_list as > > their last parameter. > > There are some things you can do - for example, calling those functions in > accordance with the rules given in the C standard. There are various > things that cause problems - for example, taking the address of a > parameter declared as a va_list (because the parameter type may have been > changed from va_list to pointer-to-element-of-va_list as part of the > parameter type adjustment of parameters declared as arrays, so the type of > ¶meter may not be va_list *). But I don't want to take the address of a va_list parameter. I just want to handle it with stuff defined in stdarg.h. Actually, I'm not sure why va_list is defined as an array in some of the architectures. The problem wouldn't arise if va_list were just defined as the structure that it's currently defined as an array of. Anyway, I still consider it a bug that gcc when called with -std=c99 compiles va_arg with array type without any hint into code that cannot possibly be useful, as it expects the array to be passed by value to the variadic function.
