> I always have understood that snprintf does not write a null character when the produced string is longer than the given size.

snprintf always null-terminates when the buffer length is greater than 0 and there's no error. That is, if snprintf returns >= 0 and the buffer length is > 0, the output is null-terminated.

This should be clear from your local snprintf(3) man page (e.g. https://man.openbsd.org/snprintf), but also see the C23 standard (https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3220.pdf) and POSIX (https://pubs.opengroup.org/onlinepubs/9799919799/).

▲

fjfaase 11 hours ago | parent [-]

You are right. I verified this myself. I do not know where I got this wrong. Maybe this was different in earlier implementations.

	▲	wahern 9 hours ago \| parent [-]
		It was first added to the Single UNIX Specification (SUS) in 1997. Perhaps there was some oddball non-conformant implementation at the time, but in those cases the standard would usually use a different name. There's the infamous strncpy, of course, which understandably could cause some confusion. And various non-vendor faux-snprintf implementations, like uprintf, that have different semantics. It's hazardous to reuse a standardized name like snprintf, even as a suffix, in an API but not implement the same fundamental semantics.