Re: [Boost-users] filesystem statistics

From: Beman Dawes (bdawes_at_[hidden])
Date: 2005-06-22 20:45:17

"Thomas Matelich" <matelich_at_[hidden]> wrote in message 
news:3944d45805062213536c52a507_at_mail.gmail.com...
> On 6/22/05, Beman Dawes <bdawes_at_[hidden]> wrote:
>>
>> "Thomas Matelich" <matelich_at_[hidden]> wrote in message
>> news:3944d45805062113231fa767ff_at_mail.gmail.com...
>>
>> Probably a dumb question, but why did you use <sys/vfs.h> and statfs()
>> instead of the POSIX <sys/statvfs.h> and statvfs()? Were they not present 
>> on
>> some of the target systems?
>>
>> I would like to include the functionality in Boost.Filesystem, but am
>> concerned because POSIX lists it as optional.
>
> I started with our legacy HP implementation as a reference and didn't
> need to change it for linux, so I didn't even notice that there was a
> different header. :-)

Understood. Does the legacy HP implementation also supply <sys/statvfs.h> 
and statvfs()?

> I'd be fairly surprised if a platform supported by Filesystem didn't
> have a way to determine disk space.  Well, I guess not all that
> surprised.  Murphy's always waitin'.

Yep. But that isn't necessarily a showstopper. 1.34 will add a function to 
create hard links, and just document the fact that it will always fail on 
some file systems.

To me the key platforms are POSIX/Windows/Linux/Solaris/MacOS/HPux/AIX. If 
we get uniform behavior across those platforms, it is OK if a function can't 
succeed on more obscure platforms.

--Beman