On 03/17/2010 08:04 PM, Joseph S. Myers wrote:
> 
> But you should be able to treat all ITEMs as if they have wide characters; 
> there should be no need to do something special for the case of no wide 
> characters.
> 
> The check still isn't appropriate.  The width might equal the number of 
> bytes even when wide characters are present.  I think the logic should be: 
> if NLS enabled and the relevant wide character functions available, use 
> get_wcs_aligned_len, otherwise use get_str_aligned_len.  This means only 
> one of those functions is used in any compiler configuration, so you can 
> just make them two different implementations of get_aligned_len depending 
> on the configuration, rather than having three function names involved.
> 
Thanks. I merge the algorithm of line break in function get_aligned_len and 
don't care if it includes wide characters. Yes, not only the code, but also the 
logic is clearer.
I add a separate conditional directive to get the aligned length when 
--diable-nls or the relevant interfaces are not available. The definition of 
function fake_ngettext is moved to the inclusions of ENABLE_NLS, not used a 
separate "ifndef ENABLE_NLS". That would use less conditionals at intl.c.

Tested on i686-pc-linux and enabled nls, the help output isn't changed when the 
locale is C. Disabled nls, the output isn't changed too.

Thanks
Pearly