On 04/ 8/10 04:33 PM, [email protected] wrote:
On Thu, Apr 08, 2010 at 04:22:15PM -0500, Shawn Walker wrote:
On 04/ 8/10 04:12 PM, [email protected] wrote:
I can see the utility of listing
- the version installed if matching the pattern
- the latest version matching the pattern
- the latest version matching the pattern with current constraints applied
- all versions matching the pattern
- all versions matching the pattern with current constraints applied
I understand that the -f flag is to match all versions, but for the
remaining three in this list, what happens?
Danek said he was going to think about it, but the more immediate
issue was ensuring that at a minimum, -a honored pattern versions.
My assumption was that once he decided that, we'd open a separate
bugzilla entry to cover those changes.
Sorry, I'm trying to get some clarification on what the new behavior is.
Does this change do the first and third; first, second, and third; or
some other combination of the above?
It fixes the first, third, and fifth. -f is required in combination
with -a for now if you want to do the second and fourth.
BRCMbnx was renamed at 0.133 but list doesn't tell me what it was
renamed to, or if I have it installed. By looking at the pkg contents,
I'm able to see that it was renamed to driver/network/bnx. Should pkg
list or pkg list -a return something like this when a renamed package is
installed?
<...>
In general, the goal for -a is to show you packages that you could
install or that are already installed.
I'm fine with trying to change the behaviour in this area, but I'd
rather address that as a separate changeset if we want to change it.
No objections for covering this in a separate bug. The case that I'm
concerned about is when other software or organizational process has a
dependency on the old name, BRCMbnx for example, and wants to know
whether that has been installed or not. If we have the renamed version
of the package installed, it seems like the answer to this question
should be yes.
Bart's opinion in this area is that we don't want to do that, because
there's a difference between having the renamed package installed, and
having the original and the renamed installed (which was the case in
b133 as an example). In other words, it is believed that knowing if the
stub package for a renamed package is installed is valuable separately
from knowing if the renamed package is installed.
I suspect renaming was a largely one-time event for most of our
packages, so as time moves on, this will matter less and less
(especially once the new release is out).
I don't remember what all of his rationale was, but I know Danek talked
to him about it as well.
-Shawn
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
