On 11/22/11 11:26, Shawn Walker wrote:
[snip]
The current logic is also broken because it doesn't take into account
whether the manifest is for a package that's installed. If it is, then
it doesn't matter if the manifest is different than what the server
has now, it can't be replaced as that would invalidate the already
installed package.
Actually, it does take into account if the package is installed. See the
code in imageplan, among other places. It's just that the knowledge of
whether it's installed might be located differently than you'd prefer.
No, it doesn't. imageplan for example only bothers to do the check if
it's under newfmri. However, this doesn't account for the cases where
the fmri hasn't actually changed, but will be the same for both
oldfmri and newfmri (think pkg fix, facets, variants, mediators).
Unless I'm missing some other piece of logic that's actually limiting
this to not-yet-installed packages.
While I wrote tests for those cases, I may have gotten them wrong. Had I
gotten this feedback a month ago when this was all in my head, I would
probably be more interested in starting over. I agree that if the code
handles those cases, how it does so is at best unclear. Since I no
longer have much state for this bug and I have no near term plans to go
back and work from first principles here, I'm dropping myself as the RE.
[snip]
Brock
-Shawn
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
