[pca] Patch mismatch when analyzing

[Martin Paul]

Laurent Blume wrote:
> Maybe (only a suggestion here) add a new flag? RSBI, the latest being a
> warning when an incompatibility has been detected? Same as for the
> badpatches, it would be left to the sysadmin to analyze and decide what
> to do.

Yes, that might be worth a closer look. I've put it onto my TODO list. I 
guess I'll set the 'I' flag for a patch if the patch it conflicts with 
is either installed or listed in the specified revision or higher. We'll 
see if this actually produces useful output or only creates confusion.

> Oh, right, I managed to miss that Special part. But then, I guess it's
> the patchdiag.xref file that's wrong: If I understand correctly, the
> patch is said to be applied on SUNWjmail. So it applies as a patch to
> the package it's supposed to create?

You're right, I missed that. That's complete nonsense, then. The patch 
should be applied if SUNWjmail/SUNWjaf are not installed (yet), not the 
other way round. Instead of SUNWjmail/SUNWjaf, one of the Sun Java 
System Application Server packages should be listed to trigger the 
installation of the missing packages. That's what results in abusing the 
patching infrastructure for things it wasn't meant to be for.

I guess the conclusion is that the patches for the Portal/Application 
Server are a complete mess, and pca can't be trusted in that area - 
garbage in, garbage out.

Martin.