Eric Blake wrote:
> since vasnprintf has the
> nice API property that, unlike all other *printf functions, it does not
> return an int, it should not need to be limited by INT_MAX; otherwise you
> are introducing an artificial limitation
You're right; I forgot about this argumentation. My patch co
Performing the unit tests on NetBSD 3.0 uncovered this: When the "size of
given buffer" argument passed to vasnprintf() was > INT_MAX, the function
could loop endlessly instead of failing with EOVERFLOW.
Note: NetBSD also has a bug here: snprintf() returns -1 with errno = EINVAL,
where POSIX says
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
According to Bruno Haible on 3/17/2007 6:32 PM:
> * lib/vasnprintf.c (EOVERFLOW): New fallback definition.
> (VASNPRINTF): Fail with EOVERFLOW when the given allocated size is
> larger than INT_MAX, or when it grow to a value larger t
