On Mon, Mar 3, 2014 at 1:47 PM, Kai Tietz wrote:
> Hi,
>
> cygwin64 shares for 64-bit the ABI of Windows native. This is caused
> by different reasons (eg. unwind-table description for prologue, etc).
> So for Windows targets va_list is indeed of 'char *' type. And this
> is ok. The variant of x86_64 abi, which uses indeed a
> structure-variant for variadic-arguments plus some call-abi
> extensions, isn't recommented for on Windows. If your linux code
> relies on structure-variadic x86_64-ABI, it seems to me broken, and
> needs to be ported.
>
> Hope this answers your question.
>
> Regards,
> Kai
I will take that as a negative answer to modifying the internal
definition of va_list...
In case anyone else googles this problem and finds this thread: my
main problem was to make string literals be matched against the
ellipsis overload, and not the va_list overload, when the two are
available. Apparently, gcc is too smart for its own good (perhaps
mandated by the standard) and will always automatically remove the
constness of a string literal and select the va_list overload (and
warn you about it!). I found three solutions, both of which require
modifications of affected code.
1) don't overload, rename one of the functions -- may affect much more
code than what is affected by the original problem.
2) cast your string literal to (void *) -- a bit confusing, and causes
a warning if __attribute__ ((format (printf, ...))) is used.
3) introduce a temporary const char *tmp = "string literal" and pass
that. gcc won't try to cast away the constness of a variable and will
match the ellipsis.
Code example:
#include <cstdarg>
#include <iostream>
using namespace std;
void foo (...) { cout << "varargs\n"; }
void foo (va_list ap) { cout << "va_list\n"; }
int main () {
foo (""); // va_list
foo ((const char *)""); // va_list
foo ((const char *)NULL); // varargs
foo ((void *)""); // varargs
foo ((const char *)(void *)""); // va_list
const char *bar = "";
foo (bar); // varargs
}
-- Irfan
--
Problem reports: http://cygwin.com/problems.html
FAQ: http://cygwin.com/faq/
Documentation: http://cygwin.com/docs.html
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
