[Bug middle-end/78703] -fprintf-return-value floating point handling incorrect in locales with a mulltibyte decimal point

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78703

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
For -fprintf-return-value, it is desirable to count with the worst case, but
I'm not sure it is the right behavior for the warning.  If we are aware of just
1 and 2 byte decimal points in the wild, forcing anyone that uses sprintf or
snprintf to change the lengths of their buffers through warnings to reserve
MB_LEN_MAX (e.g. for glibc 16) bytes for the decimal point just in case is IMHO
just too much.  It would be better to use (weak) maximum of 2 rather than 16.
Otherwise people will just disable this warning as counting something that is
totally off.  Even the counting of 2 rather than 1 byte for the decimal point
might upset people that always invoke *printf* in the C locale (on the other
side, for programs that invoke it in user selected locale might point to real
bugs that only show up in the Pashto locale).