On Wed, 25 Oct 2006, Yang Tse wrote:

> At this moment when a test suite result is shown on the autobuild table with
> tests that have failed to start the FTP, HTTP... servers they are considered
> as 'not failed' since they have not run. And the 'Test' column only reflects
> those that have run and failed or passed.
>
> The fact that a large number of tests, let's say at least 5, are not able to
> run because the FTP, HTTP... servers have failed to start, most likely needs
> attention by the one who's running the server, kill previous running test or
> something.
>
> It would be nice if by just looking at the autobuild table we were able to
> spot this condition in an easy way.

Yes, I agree completely. We don't necessarily have to mark the info with reds,
but perhaps a yellow or something that at least can bring your attention
without shouting error too much. I mean, the error is most likely in the
server or test suite and not actually in curl when that happens.

> One way could be counting as actual errors those tests that have not failed
> but have not started due to FTP, HTTP... servers not starting. In this way
> we would know for sure that the log file for that testsuite needs to be
> inspected.

I wouldn't like that. I'd prefer a differentiated output, but sure it could
get marked as an error but a server or test-suite related one.

> And another one could be something as showing a failed state as it is now
> done with 'configure' 'link' 'fail', maybe 'reboot', 'busy' or unavailable'
> ?? ;-)

I think I'll start with outputting some specific texts for the cases where a
test can't run due to the server failing to start or when an unknown server is
running on the test port, to be able to count those occurances and display on
the summary page.

-- 
  Commercial curl and libcurl Technical Support: http://haxx.se/curl.html