[libvirt] PATCH: Add a tool virt-xml-validate

[Daniel P. Berrange]

On Thu, Apr 16, 2009 at 02:42:08PM +0200, Daniel Veillard wrote:
> On Thu, Apr 16, 2009 at 12:53:06PM +0100, Daniel P. Berrange wrote:
> > While we install our RNG schemas into the filesystem, and users could
> > use xmllint to validate them, it is a little tedious to use this way.
> > 
> > This patch adds a nice simple command
> > 
> >   virt-xml-validate XML-FILE [SCHEMA-NAME]
> > 
> > You can use it on any XML file format that libvirt knows about
> > 
> >   # virt-xml-validate ~/rhel5xen.xml 
> >   /home/berrange/rhel5xen.xml validates
> >   #$ echo $?
> >   0
> > 
> > I normally figures out the SCHEMA-NAME from the top level element name,
> > but you can override it if desired
> > 
> >   # virt-xml-validate ~/bug.xml
> >   virt-xml-validate: cannot determine schema type for /home/berrange/bug.xml
> >   # virt-xml-validate ~/bug.xml domain
> >   /home/berrange/bug.txt:1: parser error : Start tag expected, '<' not found
> >   From bugzilla at redhat.com Fri Feb 27 01:12:11 2009
> >   ^
> 
>   Sounds good, I'm wondering a bit why it's better to separate it
>   rather than make it a virsh command/

Well its not part of any libvirt API, and I don't want to require a 
libvirt connection in order to use it. We've got alot of other virt-XXX
commands so I think its nicer to keep it separate. 

> > The 'src/' directory is getting over large, so I decided we should have 
> > a separate 'tools' directory for command line programs included in the
> > main libvirt source. We could consider moving virsh here perhaps.
> 
>   I would wait after 0.6.3 for this, i.e. the next flow of patches,
> but yes we should start to clean up src/

Yeah, works for me.

> [...]
> > +++ b/tools/virt-xml-validate.in	Thu Apr 16 12:44:17 2009 +0100
> > @@ -0,0 +1,151 @@
> [...]
> > +if [ -z "$TYPE" ]; then
> > +  ROOT=`xmllint --format $XMLFILE 2>/dev/null| head -2 | tail -1`
> > +  case $ROOT in
> > +     *domain*)
> > +        TYPE="domain"
> 
>  Hum, that doesn't really work in the general case, you can have
> plenty of stuff on the two first lines like the xml declaration
> or comments, or stylesheet PI ...

I figured no one ever uses that stuff  ;-P

>   Instead I would use the streaming debug output to get the
> information whithout guess from the parser itself:
> 
>     xmllint --stream --debug  $XMLFILE 2>/dev/null | grep "^0 1 " | awk
>     '{ print $3 }'
> 
> the root top level element will show up as 0 1 foo ... and there should
> be only one per document :-)

Will change it to do that instead.

Daniel
-- 
|: Red Hat, Engineering, London   -o-   http://people.redhat.com/berrange/ :|
|: http://libvirt.org  -o-  http://virt-manager.org  -o-  http://ovirt.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: GnuPG: 7D3B9505  -o-  F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|