size_t mbstowcs(wchar_t *restrict pwcs, const char *restrict s, size_t n);

According to my interpretation of the specification of mbstowcs (and
wcstombs) in the C standard, these functions are not allowed to write
beyond a terminating 0. The question was discussed in comp.lang.c today.
So far, it seems everyone agrees.

However, Ubuntu does not. On Ubuntu 18.04 with all GCC versions I tried
(5.5.0, 6.5.0, 7.5.0, 8.4.0), these functions will write up to n bytes,
i.e. beyond a 0 in the string.
On the other hand GCC and LLVM on Debian, and LLVM on Ubuntu will not
write after a terminating 0.

The GNU/Linux manual page states
"size_t mbstowcs(wchar_t *dest, const char *src, size_t n); […]
The programmer must ensure that there is room for at least n wide
characters at dest."

So clearly, people interpret the specification of these C functions differently.

I guess we need a proposal to make WG14 decide and clarify?

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Am 11.05.20 um 20:27 schrieb Florian Weimer:

I don't think there is much ambiguity in the standard (at least when
compared to other areas). This is a glibc documentation issue
(the non-default _FORTIFY_SOURCE=2 mode is not standards-conforming).

Thanks. Indeed the glibc manual currently only states "If this macro is
defined to 1, security hardening is added to various library functions.
If defined to 2, even stricter checks are applied."

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

* Philipp Klaus Krause:

size_t mbstowcs(wchar_t *restrict pwcs, const char *restrict s, size_t n);

According to my interpretation of the specification of mbstowcs (and wcstombs) in the C standard, these functions are not allowed to write
beyond a terminating 0. The question was discussed in comp.lang.c today.
So far, it seems everyone agrees.

However, Ubuntu does not. On Ubuntu 18.04 with all GCC versions I tried (5.5.0, 6.5.0, 7.5.0, 8.4.0), these functions will write up to n bytes,
i.e. beyond a 0 in the string.
On the other hand GCC and LLVM on Debian, and LLVM on Ubuntu will not
write after a terminating 0.

The GNU/Linux manual page states
"size_t mbstowcs(wchar_t *dest, const char *src, size_t n); […]
The programmer must ensure that there is room for at least n wide
characters at dest."

So clearly, people interpret the specification of these C functions differently.

I guess we need a proposal to make WG14 decide and clarify?

I don't think there is much ambiguity in the standard (at least when
compared to other areas). This is a glibc documentation issue
(the non-default _FORTIFY_SOURCE=2 mode is not standards-conforming).

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)