https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126858

--- Comment #2 from Vincent Lefèvre <vincent-gcc at vinc17 dot net> ---
(In reply to Vincent Lefèvre from comment #1)
> Here, I don't think that GCC knows that check_vfprintf is like fprintf. So
> the bug may be more general, with function calls having an argument taking
> an address, in which case the analyzer may not be able to determine whether
> the variable is initialized. But this should never generate a false positive.

Indeed, from the code of the MPFR testsuite, I could derive another simple
testcase, but it seems rather different from the above one with printf, for 2
reasons:
1. One needs at least 3 arguments.
2. One needs to use an intermediate function with va_list.

Reported as PR126859.

Reply via email to