strtold does not set errno when it should

Bruno Haible bruno@clisp.org
Mon Dec 16 10:05:00 GMT 2019


Hi Corinna,

> The code for strtold is almost verbatim taken from
> 
>   https://github.com/jwiegley/gdtoa.git
> 
> There haven't been any patches upstream.

Probably that's because the code is good enough for ISO C compliance;
however, POSIX [1] adds the "ERANGE upon underflow" requirement,
whereas ISO C only has an "ERANGE upon overflow" requirement.

[1] <https://pubs.opengroup.org/onlinepubs/9699919799/functions/strtod.html>

> diff --git a/newlib/libc/stdlib/strtodg.c b/newlib/libc/stdlib/strtodg.c
> index 013315946c1b..d6fb26ad3b45 100644
> --- a/newlib/libc/stdlib/strtodg.c
> +++ b/newlib/libc/stdlib/strtodg.c
> @@ -1091,6 +1091,10 @@ _strtodg_l (struct _reent *p, const char *s00, char **se, FPI *fpi, Long *exp,
>  				irv |= STRTOG_Underflow;
>  			}
>  		}
> +#ifndef NO_ERRNO
> +	if (irv & STRTOG_Underflow)
> +		errno = ERANGE;
> +#endif
>  	if (se)
>  		*se = (char *)s;
>  	if (sign)
> 
> which seems to do the trick.  But, does it make sense?  If not, I'd
> really appreciate a patch.

I think it goes in the right direction.

It can be improved a bit, though: The existing code for overflow is careful
to do the errno stuff only when STRTOG_Overflow is being set/added. I.e.
keep this overflow/underflow handling out of the main path, in order not
to slow down the 99.99% of the cases that don't need it.

My attempt to do this would look like the attached patch. Untested.

Bruno
-------------- next part --------------
A non-text attachment was scrubbed...
Name: strtodg.c.diff
Type: text/x-patch
Size: 1911 bytes
Desc: not available
URL: <http://sourceware.org/pipermail/newlib/attachments/20191216/2895a62a/attachment.bin>


More information about the Newlib mailing list