testrepository-dev team mailing list archive

[Gary Poster <gary.poster@xxxxxxxxxxxxx>]

On 04/30/2012 03:52 PM, Robert Collins wrote:
> Gary and I caught up a little on IRC and are more fully synced now :).
>
> tl;dr (in fact, private chat, thats why noone else has read it :P)
>
>   - --subunit probably shouldn't ever be filtered
>   - doing the zope filtering etc is still desired for end users using
> things without buildbot
>   - the stuff that is muddy is what jml and I put in place in our
> refactorings back before the LP parallel test project
>   - the branch I have is old (from budapest) merely brought up to date
> with one commit in place to start shaping the aggregation of
> TestResult filters.
>
> Attached is a new image to show what I think will work:
>   -- subunit will be taken straight from the cloned mux stream and will
> always be what-came-in (modulo the MUX tagging which is intended to be
> non-destructive).
>   -- --full-results will bypass user transforms such as tagged test
> removals (we may want to tweak the name before releasing, but thats
> minor) (e.g. --no-transforms)
>
> A thing to note is that the existing CLITestResult object is doing
> everything reached from the UI box, so its kindof bloated; this
> diagram may help us clean that up in future, but it doesn't affect the
> short term work to do to make the branch landable - basically reglue
> the three options slightly, and the implement the tag based filtering
> (including special casing errors/failures IMO).

Cool, thanks Rob.

I'll put a card on the Yellow kanban board for someone to move this 
branch along in slack time, though there's no guarantees anyone will 
take it--don't let us stop anyone, please!

Rob, a question about the "time" box.  You said this in your first email.
"""
time the run (for the summary) - but this perhaps comes from the
repository by querying the last run, so arguably ignorable. Note that
this should not be 'sum of test times' but 'duration in stream' or
something similar (because with two streams we want the wall time, not
the sum of time per CPU).
"""
IOW, it should simply be the last time stamp minus the first time stamp? 
 I'm not sure why you put a question mark next to it--you questioned 
whether we would need a clone in order to accomplish it?

Thanks

Gary