Which bugzilla component?
Richard Hughes
hughsient at gmail.com
Tue Sep 21 10:09:58 UTC 2010
On 9 September 2010 17:39, John Poelstra <poelstra at redhat.com> wrote:
> Richard Hughes said the following on 09/08/2010 12:13 PM Pacific Time:
>> This is bad, both from a "your system" perspective, and from a
>> "PackageKit reporting errors to the user" perspective. File a bug
>> against PackageKit-command-not-found and we can debug there.
>>
>> Richard.
>
> I can't reproduce this any more. Still worth reporting?
I've found the problem, it was a race discovered when adding tests to
the PackageKit self test framework:
commit 2b3c13110a448ab1535716c0944e43a7e945ac98
Author: Richard Hughes <richard at hughsie.com>
Date: Tue Sep 21 11:06:58 2010 +0100
Ensure we set an error if the spawned backend gets cancelled
We can't assume the spawned backend gets a chance to set an error
when we SIGQUIT it
and so set a generic transaction cancelled error code when before
we send the signal.
This means we don't trigger the obnoxious internal error when
sometimes cancelling
transactions, e.g. '...This is a serious error as the spawned
backend did not...' and
we don't scare the user when there's nothing actually wrong.
Richard.
More information about the test
mailing list