[et-mgmt-tools] libvirt and query strings in EL 5.1 behaves differently than F9

[Pablo Iranzo Gómez]

	Hi

	Probably this one is more descriptive (koan generation, virsh test
and xml file for virsh dumped from koan.log)

	http://cobbler.pastebin.com/f50f3fa63

	The xml is the one dumped on ~/.koan/koan.log

	Regards
	Pablo




-- 
Pablo Iranzo Gómez
(http://Alufis35.uv.es/~iranzo/)
(PGPKey Available on http://www.uv.es/~iranzop/PGPKey.pgp)
                  --
Postulado de Boling sobre la Ley de Murphy:

Si se encuentra bien, no se preocupe. Se le pasará

On Tue, 22 Apr 2008, Michael DeHaan wrote:

> I was talking to Dan yesterday on OFTC about how to express query
> strings in libvirt.   Basically the new Cobbler wants to generate
> kickstarts on the fly, and also do some fun things about remote
> registration.   This already works for PXE but I am having trouble
> passing the kernel options line into libvirt still.   What I was trying was:
>
>     For instance:
>        ks=http://server.example.com/blah?op=ks&profile=blah   [OR]
>        ks=http://server.example.com/blah?op=ks;profile=blah
>
> Under F9, this works for me correctly when I use the former method and do
>
>     extra = extra.replace("&","&")
>
> However I have recieved reports today that the above does not work on
> EL-5.1, where it (also replicated with virsh) chokes -- giving a message
> that the end of the XML was unexpected (see pastebin below).   I am
> thinking the XML escaping built into virt-install/libvirt may not be the
> same in EL-5.1?   So I knew that query strings can also be delimited by
> colons, but that too fails.   If the form with the semicolon is easier,
> how would I express that with correct escaping so that calls to virtinst
> worked in both EL-5 and F-9?
>
> Here is the error we are seeing from the latter query string form on 5.1
>
> http://cobbler.pastebin.com/db0f8c7a
>
> Help?
>
> The last resort is to do something funky and make cobbler tolerate URLs
> that do not contain any form of "?", "=", "&", and ";", but I would like
> to stick to the RFCs for query string formatting if possible.   I do
> know that will work however, but if I need to do OS detection and make
> calls into libvirt differently that would be doable also (and maybe
> preferred).
>
> Thanks!
>
> --Michael
>
>
>
>
>
>
> _______________________________________________
> et-mgmt-tools mailing list
> et-mgmt-tools at redhat.com
> https://www.redhat.com/mailman/listinfo/et-mgmt-tools
>