On Thu, Jan 28, 2010 at 08:42:25AM +0000, Gary Pennington wrote:
> > - looking at this code i'm concerned that it will only work if the gz is
> >   running the latest bits from the repo.  specifically, if "entire" is
> >   not installed in the gz, then you install SUNWcsd/SUNWcs but you don't
> >   put any version constraints on those packages.  so imagine if you had a
> >   system running snv_128, then you remove "entire" and add the on-ips
> >   publishers.  then you try to install a zone.  then what syncs
> >   SUNWcsd/SUNWcs with the global zone?
> >
> >   or another example.  if a system is already running on-ips nightly
> >   bits.  then the on-ips nightly repo is updated with new bits.  then if
> >   you try to install a zone, you'll happily install bits that aren't
> >   in sync with the global zone, right?
> >
> >   it seems to me that when installing zones you need to add explicit
> >   version syncs (which include timestamp granularity) for the following
> >   packages:
> >
> >     SUNWcsd/SUNWcs
> >     SUNWips
> >
> >   if you do that, then since those packages depend on their
> >   incorporations you'll also sync the ON and IPS gate incorporations.
> >
> >   doing this won't sync every ON/IPS package at a timestamp granularity
> >   (which is ultimately what we need), but it wouldn't be worse that what
> >   we have today.
> >
>
> These questions are (I think) answered by the changes that went into b130
> to address suggestion 1 from comment 0 in 12738.
>
> Specifically, I am relying on the fact that now all packages delivered by
> a consolidation have a dependency on the consolidation incorporation (to
> paraphrase the comment I reference).
>
> However, I'm really relying on the feeback that I have from Liane on the
> success of her testing of these changes with the new on ips code that she's
> working on.
>
> Perhaps Liane could provide some more detail to this answer?
>

the fact that you're unable to explain to me why this is not a problem
confirms in my mind that it is indeed a problem.

i see that you updated the bug with a description of this issue, but if
you putback this fix without addressing this issue, then another bug
will have to be opened to track this issue.  given that:
- we're trying to get this into b132
- putting this back as is doesn't make things any worse than they are today
- putting this bask as is lets us work with on-ips in the common case

i guess i'm ok with this issue being addressed as a follow on bug.

ed
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to