Re: [libvirt] [PATCH 5/6] allow non-zero success values from a domain's MigratePerform
On 10/07/2009 03:59 PM, Daniel P. Berrange wrote:
This hack means that it would be ignoring*all* deserialization
errors,
which IMHO is just as bad.
These deserialization errors however do not include networking errors,
because libvirt calls xdr routines only after it has extracted the xdr
data from the reply packet. And the reply packet is also synthesized
via xdr, so basically the only errors that would be let through would be
bugs in xdr, or mismatches between daemon/dispatch.c and
src/remote_driver/remote_driver.c. Both cases would have likely showed
up long before the invocation of MigratePerform.
(I know it is a hack though).
Would a patch introducing MigratePerform2 (or 3, or 4??? not joking) be
appropriate? I hoped I wouldn't have to go down that way, but given I
have Chris's tunneling patch to follow the model, it shouldn't be hard.
Paolo