correct behaviour of mbstowcs function

Hi!

It seems to be that the MSL lib implements the mbstowcs function
differently from Apple's implementation in case the first pointer is NULL.
Apple's implementation seems to return the length of the "in"-string
while MSL always returns 0. When looking at Apple's man pages even
(size_t) -1 should be returned.

Now, what is the right functionality?

Hartwig

PS: man pages from Apple to mbstowc

#include <stdlib.h>

size_t
mbstowcs(wchar_t * restrict wcstring, const char * restrict mbstring,
size_t nwchars);

#include <xlocale.h>

size_t
mbstowcs_l(wchar_t * restrict wcstring, const char * restrict
mbstring,
size_t nwchars, locale_t loc);

DESCRIPTION
The mbstowcs() function converts a multibyte character string mbstring
beginning in the initial conversion state into a wide character string
wcstring. No more than nwchars wide characters are stored. A
terminat-terminating
ing null wide character is appended if there is room.

While the mbstowcs() function uses the current locale, the
mbstowcs_l()
function may be passed a locale directly. See xlocale(3) for more
infor-information.
mation.

RETURN VALUES
The mbstowcs() function returns the number of wide characters
converted,
not counting any terminating null wide character, or -1 if an invalid
multibyte character was encountered.

ERRORS
The mbstowcs() function will fail if:

[EILSEQ] An invalid multibyte sequence was detected.

[EINVAL] The conversion state is invalid.

In article <43bd9ff7$0$31409$dbd4b001@news.wanadoo.nl>,
Hartwig Wiesmann <hartwig.wiesmann.TakeThisOut@wanadoo.nl> wrote:
>
> It seems to be that the MSL lib implements the mbstowcs function
> differently from Apple's implementation in case the first pointer is NULL.
> Apple's implementation seems to return the length of the "in"-string
> while MSL always returns 0. When looking at Apple's man pages even
> (size_t) -1 should be returned.

Where does it say that?

> Now, what is the right functionality?

I haven't got the C99 specification to hand, but this is the relevant
part of the Single Unix definition for the function:

> RETURN VALUE
> If an invalid character is encountered, mbstowcs() returns (size_t)-1 and may
> set errno to indicate the error. Otherwise, mbstowcs() returns the number of
> the array elements modified (or required if pwcs is null), not including a
> terminating 0 code, if any. The array will not be zero-terminated if the
> value returned is n.

<http://www.opengroup.org/onlinepubs/007908799/xsh/mbstowcs.html>

Note the words in brackets: 'or required if pwcs is null'.


Alwyn