While I was working on these changes, I realized I needed to make sure
that the BE logic was handled correctly on image-update. This led to a
cascade of discovered bugs. I'll be putting out a new webrev once I get
a handle on all of them.
Shawn Walker wrote:
On Jun 19, 2009, at 9:06 PM, Brock Pytlik wrote:
This wad fixes some of the current error hiding that happens. In the
test suite, if we start getting unexpected index rebuilds, the test
suite will fail that test. The error messaging to users has also been
improved so that if an exception happens during indexing which we
can't recover from, we notify the user of the error, but then inform
them that the install/uninstall/image-update succeeded despite the
error while updating the index.
Webrev:
http://cr.opensolaris.org/~bpytlik/ips-8948-v1/
Bugs:
8948 search errors should make it clear that the rest of the command
worked
http://defect.opensolaris.org/bz/show_bug.cgi?id=8948
7534 Need to check when auto-reindex happens in test suite
http://defect.opensolaris.org/bz/show_bug.cgi?id=7534
client.py:
lines 2596, 2597: doesn't this mean that setlocale() and
gettext.install() can happen multiple times in a wrapped error case?
It seems like these might belong in if __name__ == "main" with
appropriate try/except wrappers
I've moved it as suggested. I'm open to hearing what the except block
should contain. Previously, if gettext.install or misc.locale actually
raised an exception, we would have then raised a different exception in
the catch all block of the main exception handling since _() wouldn't be
defined, so I'm not seeing a regression here, and actually an improvement.
modules/client/api_errors.py:
lines 611, 628: missing ',' at end
line 629, 630: running or performing?
modules/client/imageplan.py:
line 116: should this be a public or private property? If nothing
needs access to this outside the imageplan class, please make it
private. Does it even need to be a property?
It needs to be a property to hang onto the state between the preexecute
phase and the execute phase.
One question that occurred to me when I looked over this code again is
how you want history to handle this case. I believe that if the main
operation reported success, that's what the history should record.
Brock
Cheers,
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
