[Openais] Reproducing saftest results?

[Steven Dake]

On Thu, 2007-09-06 at 12:02 -0700, Drew Moseley wrote:
> Steven Dake wrote:
> > I run a patched version of saftest 0.6.1. The patches I use fix various
> > problems in saftest0.6.1 which can be understood by examination of the
> > patches.  the patches along with the script and html generator are
> > included as a tarball in this message.
> >   
> Hi Steve,
> 
> Finally got around to looking at this.  The patches that are applied
> remove the
> following two tests completely:
> 
>     src/checkpoint_mamagement/saCkptCheckpointOpen/5-4.test
> 
>     src/checkpoint_mamagement/saCkptCheckpointOpenAsync/6-4.test
> 
> Are those just bogus tests?
> 

They are valid for AIS B.01.01 but point at an errata in the AIS B.01.01
specification itself which was later corrected.  Because it was deemed
differing retentiontime should not be used as a reason for returning
SA_AIS_ERR_EXIST in later checkpoint specifications because applications
on other nodes may not know the exact retention duration timeout that is
desired by the creating node.  In effect requiring this would make the
API more difficult to use.

Later AIS specifications specifically identified this issue and we
deemed it was a good change for the ckpt service and hence matched it.

Regards
-steve



> Drew