--with-libxfs error on F23.

--with-libxfs error on F23.

[Ira Cooper]

Work is being done on Fedora 23.

checking xfs/xfs.h usability... no
checking xfs/xfs.h presence... yes
configure: WARNING: xfs/xfs.h: present but cannot be compiled
configure: WARNING: xfs/xfs.h:     check for missing prerequisite headers?
configure: WARNING: xfs/xfs.h: see the Autoconf documentation
configure: WARNING: xfs/xfs.h:     section "Present But Cannot Be Compiled"
configure: WARNING: xfs/xfs.h: proceeding with the compiler's result
configure: WARNING:     ## ----------------------------------------- ##
configure: WARNING:     ## Report this to ceph-devel@vger.kernel.org ##
configure: WARNING:     ## ----------------------------------------- ##

There has been a change in the XFS header files, where there's a
problem:

(from config.log)

configure:23389: gcc -c -g -O2  conftest.c >&5
In file included from /usr/include/xfs/xfs.h:37:0,
                 from conftest.c:86:
/usr/include/xfs/linux.h:145:9: error: unknown type name 'off64_t'
 typedef off64_t  xfs_off_t;

ira@ira-t430:/usr/include/xfs
[/dev/pts/1](1002/0)$ ag off64_t
linux.h
145:typedef off64_t		xfs_off_t;

off64_t isn't there by default.

Kernel/Headers: uname -a 
Linux ira-t430 4.4.6-300.fc23.x86_64 #1 SMP Wed Mar 16 22:10:37 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux

My research turned up:

http://oss.sgi.com/archives/xfs/2016-01/msg00385.html

Ironically:

http://oss.sgi.com/archives/xfs/2009-11/msg00151.html

So it is known there was issues with off64_t in the past.

A google search showed other projects having issues, alas. What is a
good next step?  I'd be glad to provide a revert patch for the kernel.

Cheers,

-Ira

Re: --with-libxfs error on F23.

[Christoph Hellwig]

On Tue, Mar 29, 2016 at 06:10:37PM -0400, Ira Cooper wrote:
> ira@ira-t430:/usr/include/xfs
> [/dev/pts/1](1002/0)$ ag off64_t
> linux.h
> 145:typedef off64_t		xfs_off_t;
> 
> off64_t isn't there by default.

off64_t requires _LARGEFILE64_SOURCE on Linux/glibc which still likes
to make applications not handle large file offsets and inodes by default,
something which 4.4BSD solved 20 years ago.

We have no intent in supporting non-large file applications in
xfsprogs, so I think the right thing is to ensure that your configure
checks defines _LARGEFILE64_SOURCE.  I'm pretty sure the applications
actually already do that, and it's just the configure machinery that's
lazy.