Shawn Walker wrote:
Brock Pytlik wrote:
As I noted in the bug, I'm not convinced that the behavior described
in 8613 is a valid bug. I believe that's performing as expected. I'd
like to have that discussion before this goes back.
Everywhere in our documentation, we tell users that the preferred
publisher is what determines what source is used for package installs.
In addition, as I stated in the bug:
I considered this a bug as the behaviour was inconsistent with install
and with another case for image-update.
Specifically, install will upgrade a package using the preferred
publisher (always) while image-update will not.
image-update currently only upgrades a package using the preferred
publisher if that package was installed using the preferred publisher.
Consider the following case for image-update:
* Assume publisher test1 has [email protected] and publisher test2 has [email protected]
pkg set-publisher -P test1
pkg install [email protected]
pkg set-publisher -P test2
pkg image-update
You changed the preferred auth, so you reset the pointer. I'm glad we
both agree this works correctly.
In the above case (today) image-update would happily upgrade to
[email protected] from test2 despite the fact that [email protected] was installed from
test1.
However, in the below cases:
pkg set-publisher -P test2
pkg install [email protected]
pkg image-update
pkg set-publisher -P test2
pkg install [email protected]
pkg set-publisher -P test1
pkg set-publisher -P test2
pkg image-update
...image-update will not upgrade to [email protected] even though test2 is the
preferred publisher. It didn't make any sense to me that install and
one case for image-update would update a package despite whatever
publisher was originally used for installation while another case
would not.
I'm fine with that working as specified as long as if I'd said pkg
install pkg://test1/[email protected], the image-update didn't take me off test1.
If your change continues that behavior, I don't object to it.
Finally, as I also mentioned in the bug, without this change, there is
(possibly) no way for a user to override opensolaris.org and use their
own packages from their own repository if they want to.
I'd consider that a fairly important bit of functionality.
I agree. I always assumed that if I wanted to install packages from my
own repository, I'd either a) make it my preferred auth, or b) use pkg
install pkg://myrepo/foo
Cheers,
Brock
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
