On Wed, Jun 25, 2008 at 10:06:11AM -0400, David Golden wrote:
> 
> * Is flagging tester in an environment variable a good idea?  Bad idea?

Is this so the author can specifically target a tester who they don't
wish to test their distribution? If this is because of a bug, and the
bug gets fixed, that tester is still excluded.

What if there are several testers that the author wishes to
circumnavigate, this could skew the testing so the author can ensure
only testers who regular pass their distribution are the only ones to
test them. What if the tester changes their "ID"?

Following on from that, what is the "ID", their sender email? These
change regularly, even for regular testers. There are at least 10 or so
mails that appear for previously known testers every month.

I'm not really seeing the point for targeting testers to not test your
distribution forever for that particular version. 

> * AUTOMATED_TESTING or something else?

I have seen checks for AUTOMATED_TESTING == 1, so those would break if
you were to use something else.

> * Let smokers set their own ID for this?  (more work, less standard)
> Or use the email "From" line?  (automatic and consistent)

This might be an idea, but we already now have the problem that CPAN has
added the CPANPLUS_IS_RUNNING flag, and if CPANPLUS adds
CPAN_IS_RUNNING, who knows what's running?

But again, you are excluded anyone ever testing a specific distribution
with a particular toolchain. The result is that the smoker gets patched
to be the same as one that isn't excluded.

If there is a problem with a particular tester, ask them to exclude your
distributions until the problem is resolved.

> The latter has potential spam issues since the email addresses will
> wind up visible in the environmental variables section of the report.
> On the plus side, it makes it easier to see who sent a report in the
> NNTP web interface when the From line is obscured.

The stats site uses the From address to identify testers, and that
should be accessible from within the smoker tools. However, banning
testers, which this effectively is intending, seems a bit harsh to me.

Cheers,
Barbie.
-- 
Birmingham Perl Mongers <http://birmingham.pm.org>
Memoirs Of A Roadie <http://barbie.missbarbell.co.uk>


Reply via email to