[libvirt] [PATCH v4 1/2] src/Makefile: move the new xen_xl_disk parser code at the correct place

[Daniel P. Berrange]

On Thu, Jan 08, 2015 at 04:43:36PM -0700, Jim Fehlig wrote:
> Eric Blake wrote:
> > On 01/08/2015 06:20 AM, Pavel Hrdina wrote:
> >   
> >> Signed-off-by: Pavel Hrdina <phrdina at redhat.com>
> >> ---
> >>  src/Makefile.am | 34 +++++++++++++++++-----------------
> >>  1 file changed, 17 insertions(+), 17 deletions(-)
> >>     
> >
> >   
> >>  
> >>  if WITH_XENCONFIG
> >> +AM_LFLAGS = -Pxl_disk_ --header-file=../$*.h
> >>     
> >
> > Uggh.  Not your fault (this patch was just code motion), but it
> > highlights a portability problem: RHEL 5 ships with flex 2.5.4, which
> > lacks --header-file, and which spells --outfile only as -o.  The build
> > is failing with:
> >
> > flex  -Pxl_disk_
> > --header-file=/home/dummy/libvirt/src/xenconfig/xen_xl_disk.h
> > --outfile=/home/dummy/libvirt/src/xenconfig/xen_xl_disk.c
> > xenconfig/xen_xl_disk.l
> > flex: unknown flag '-'.  For usage, try
> >         flex --help
> >   
> 
> Sigh...  Sorry for all the grief this has caused.  I wasted a huge
> amount of time trying to get all of this working right, including some
> of the approaches in the follow-up patches, but always hit some sort of
> problem.  I thought I finally had everything worked out with the V3
> patches that Michal ACKed.  But given all the problems I encountered
> along the way, I suppose I shouldn't be surprised by the problems folks
> are seeing.  So now I'm causing others to waste time :-(.  Again, I'm
> sorry for that.
> 
> And of course this has rippled to the Xen push gate tests too.  I
> responded to one such test failure today on xen-devel, to which Ian
> Campbell replied that the parsing code in Xen was placed in libxlutil to
> avoid libxl apps needing to create their own parsing code
> 
> http://lists.xen.org/archives/html/xen-devel/2015-01/msg00657.html
> 
> My immediate thought was cool, I can use that and remove all the flex
> nonsense before the next libvirt release.  But alas, the related header
> file (libxlutil.h) has never been installed and thus not included in any
> distro's xen-devel package
> 
> http://lists.xen.org/archives/html/xen-devel/2015-01/msg00690.html
> 
> I suppose it is possible to include some minimal header in libvirt (or
> an extern declaration?) for the xlu_disk_parse() function, and actually
> link against the libxlutil.so which is installed.  I can work on such an
> approach if folks think it is worthwhile.

To repeat what I said on IRC. I don't have a strong opinion one way or
the other. Whatever the libvirt xen maintainers think is the best
approach is fine by me . So it sounds like using libxlutil is probably
a winner right now, with a little #ifdef to cope with the missing
header file on old versions.

Regards,
Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|