Record OOPS for HWDB invalid submissions

On Tue, Aug 30, 2011 at 2:20 AM, Francis J. Lacoste
<email address hidden> wrote:
> Public bug reported:
>
> We should record OOPS for invalid HWDB submissions. Currently, they are

Huh? No. Why? We were getting thousands of these and I just nuked that code.

> Huh? No. Why? We were getting thousands of these and I just nuked that code.

Memo to self, never dash off a quick mail :)

So, There is a recent commit of mine that deletes oopses from HWDB because:
 - we do log them
 - we have no reason to believe they were server side issues that
should be looked at by a sysadmin

-> thus they should not be OOPSes at all.

Why do you think they should be?

Well actually there was something to be investigated in these thousands of OOPSes. Checkbox is broken (arguably) since Lucid (!!!) and submitting HW profile we couldn't parse. And there was something we could do about it server-side (that's the fix to bug 835103). Maybe we should only monitor the ratio of succesful to unsuccessful submissions to detect case like these. But here is a case where the OOPS were clearly a sign that something was wrong.

Perhaps. But they shouldn't be informational then :) note that the client is
told of the problem too... So it can inform e.g. via apport. I think this is
really a UFD situation not an oops one.

On 31/08/2011 6:53 AM, "Francis J. Lacoste" <email address hidden>
wrote:

Well actually there was something to be investigated in these thousands
of OOPSes. Checkbox is broken (arguably) since Lucid (!!!) and
submitting HW profile we couldn't parse. And there was something we
could do about it server-side (that's the fix to bug 835103). Maybe we
should only monitor the ratio of succesful to unsuccessful submissions
to detect case like these. But here is a case where the OOPS were
clearly a sign that something was wrong.

--
You received this bug notification because you are subscribed to
Launchpad Suite.
https://bugs....

Yes, UFD was what I had in mind.

Right. So for ufd we don.t oops it- which is why we shouldn't here: genuine
failures will auto oops as the other bug shows ufd style errors are reported
to the client to handle as it chooses

On 31/08/2011 7:55 AM, "Francis J. Lacoste" <email address hidden>
wrote:

Yes, UFD was what I had in mind.

--
You received this bug notification because you are subscribed to
Launchpad Suite.
https://bugs....

We've removed the hardware database, so this is no longer applicable.