michael schuster wrote:
Shawn Walker wrote:
Ivan Wang wrote:
michael schuster wrote:
<snipped>
it seems that 'pkg refresh --full' was necessary to
get things going again
Though it's still not good experience why a manual download of server
metadata is required. one wouldn't do that had he not know there is
an update on server in the first place.
A pkg refresh --full shouldn't have been necessary; a pkg refresh
should have been sufficient.
However, there are two possible causes here that I can think of quickly:
* proxy/ISP issue preventing incremental catalog updates from working
(this has been reported a few times at least that I know of personally)
* The implicit refresh check performed by the packaging system had
already been performed at least once within the last four hours, so an
explicit 'pkg refresh' was necessary to force a check for updates.
For performance and other reasons, the default is to only check for
updates every four hours.
it was neither - without "--full", I got an error about "Sys V package
... already in ..." (can't remember the wording), which I found was
related to some respin of the internal packaging server (again, from
memory). the WA given was to use --full, which worked for me.
Ah, then you've hit one of the rarer cases:
An unexpected Repository rollback or change prevents the client from
incrementally updating the catalog because state information has been
invalidated so the client's reference point for the last catalog change
is no longer valid.
There has been quite a bit of debate about what to do in this scenario,
as it means that the available list of packages has changed, possibly in
a way that the user might not like. My personal inclination is to
simply warn the user and then perform a full catalog retrieval, and
that's what pkg(5) will eventually do. Right now it exits with an
UpdateLog error.
Cheers,
--
Shawn Walker
_______________________________________________
indiana-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/indiana-discuss
