On 07/29/2014 09:18 AM, Pádraig Brady wrote:
> On 07/29/2014 02:25 PM, Bernhard Voelker wrote:
>> On 07/26/2014 10:34 PM, Pádraig Brady wrote:
>>> On 07/26/2014 07:06 PM, Bernhard Voelker wrote:
>>>> * NEWS (seq invocation): Explicitly mention 'infinity' as possible
>>>> values for the program arguments.
>>
>>> Perhaps s/infinity/inf/ as per http://bugs.gnu.org/18103#27
>>
>> Hmm, or even better enhance
>>   @node Floating point
>> with more examples like "inf" which *might* be supported?
>> WDYT?
> 
> Sounds good.
> "inf" is supported on GNU/Linux, FreeBSD, NetBSD, Solaris at least

Case-insensitive "inf" and "infinity" are required by POSIX for all
strtod() implementations; and are required to map to floating point
infinities for all but ancient hardware that lacks such a number (in
which case it maps to HUGE_VAL instead).  Any libc that can't parse
"inf" is non-POSIX-compliant, and we could argue that gnulib's strtod
should be used to work around that broken implementation.  Therefore, I
think the documentation should be unconditional that both forms are
supported.

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to