Re: Thoughts on the wcwidth confusion

On 01/06/2026 18:12, Thomas Wolff via Cygwin wrote:
>
> Am 01.06.2026 um 17:59 schrieb Thomas Wolff via Cygwin:
>> Am 01.06.2026 um 17:34 schrieb Jakob Bohm via Cygwin:
>>> Dear list,
>>>
>>> Having read through the recent debate around the wcwidth() POSIX API,
>>> wchar_t definitions, gcc-16 and cygwin, I have an idea not
>>> mentioned in the list so far:
>>>
>>> Using C17 types char32_t and char16_t, the situation can be
>>> summarized as follows:
>>>
>>> - Many, but not all POSIX systems define wchar_t as char32_t and thus
>>> wint_t as uint_least32_t
>>>
>>> - Win32 and thus Cygwin defines wchar_t as char16_t and thus wint_t as
>>> uint_least16_t
>>>
>>> - All systems considered treat wchar_t as unicode, with Win32 
>>> supporting
>>>  UTF-16 since the NT 5.00 (Windows 2000).
>>>
>>> - For char16_t/UTF-16, wcwidth() should use the high surrogate to
>>>  determine the range of unicode symbols and return a width common to
>>>  that range, then return 0 for the low surrogates, thereby allowing
>>>  computation of string width without having to first assemble 
>>> surrogates
>>>  into full char32_t values.  Deciding if char32_t implementations 
>>> should
>>>  still lump groups of 4 Unicode rows for UTF-16 compatibility is up to
>>>  each implementation.
>> It's a neat idea to split the width calculation over the surrogates. 
>> Unfortunately it does not work this way because widthness does not 
>> change in full 1024-byte blocks. For example, U+1F4FC is Wide, 
>> U+1F4FD and U+1F4FE are narrow/Neutral (N), and U+1F4FF is W again.
>> As a variant of your idea, wcwidth could return width 1 for every 
>> high surrogate, remember it, and if the subsequent invocation is a 
>> low surrogate, determine the combined width and return either 1 or 0.
>> Not quite standard behaviour, I suspect, so maybe not a good idea for 
>> the purists, but maybe worth some discussion.
> On the other hand, there are also combining characters in the non-BMP, 
> so the only way this could work is width 0 for high surrogates, then 
> sum up to the actual width on the low surrogate. Leaving the question 
> how to handle an (errorneously) single high surrogate...
>
If using this "hidden state" concept, the big question is how to handle
a single or out-of-sync low surrogate in wcwidth().  For wcswidth(),
the full context is always available and lone surrogates will be no
different than other invalid chars such as U+1FFFFE .
>>>
>>> A practical solution would be for Cygwin/newlib to provide new 
>>> functions
>>> c16width(), c32width(), c16swidth() and c32swidth(), each being the
>>> explicit size equivalants of their wc and wcs similarly named 
>>> functions.
>>>
>>> Then wcwidth() can be a trivial inline alias of the explicit size
>>> equivalent for the compile target by having the newlib header 
>>> checking a
>>> compiler or standard define indicating the chosen size of wchar_t.
>>>
>>> // possible wchar.h snippet
>>> //
>>> // C17+ required
>>> // For C2Y+ this should go in uchar.h
>>> //
>>> int c16width(char16_t c);
>>> int c32width(char32_t c);
>>> int c16swidth(const char16_t *s, size_t n);
>>> int c32swidth(const char32_t *s, size_t n);
>>>
>>> // ...
>>>
>>> // This belongs in wchar.h for C1x- compat
>>> //
>>> #if SOMETHING_MEANING_16bit_WCHAR_T
>>> inline int wcwidth(wchar_t c) {
>>>   return c16width(c);
>>> }
>>> inline int wcswidth(const wchar_t *s, size_t n)
>>> {
>>>   return c16swidth(s, n);
>>> }
>>> #else
>>> inline int wcwidth(wchar_t c) {
>>>   return c32width(c);
>>> }
>>> inline int wcswidth(const wchar_t *s, size_t n)
>>> {
>>>   return c32swidth(s, n);
>>> }
>>> #endif
>>>
>>>
>>> Enjoy
>>>
>>> Jakob
>>
>>
>
>
Enjoy

Jakob
-- 
Jakob Bohm, CIO, Partner, WiseMo A/S.  https://www.wisemo.com
Transformervej 29, 2860 Søborg, Denmark.  Direct +45 31 13 16 10
This public discussion message is non-binding and may contain errors.
WiseMo - Remote Service Management for PCs, Phones and Embedded


-- 
Problem reports:      https://cygwin.com/problems.html
FAQ:                  https://cygwin.com/faq/
Documentation:        https://cygwin.com/docs.html
Unsubscribe info:     https://cygwin.com/ml/#unsubscribe-simple