Re: Who to contact to get a buildd retry on alpha?
On Nov 30, 2007 11:35 AM, Steve Langasek <firstname.lastname@example.org> wrote:
> > The test that failed is known to succeed most of the time but
> > occasionally fail for no apparent reason; if the package is retried it
> > should build fine. Who should I be bothering?
> No one; you should fix your package so that it builds reliably.
> The very test that's failing includes documentation of why the test is
> broken and unreliable. Please either fix the test or disable it in the
> build. Having to retry packages because they sometimes fail is obnoxious,
> especially when a security update is involved.
I think you may have misunderstood the comments in the test, which
detail all the *previous* race conditions that were found and fixed.
There is obviously still something wrong with the test, but the cause
is not known; given that it only fails about once every fifty builds,
fixing it has not been a high priority for upstream.
Unless upstream does get around to fixing it for sure, the next upload
will disable that test. However, I would prefer not to do yet another
upload of 0.37 just to do that. I think it is a more efficient use of
everyone's time if the package is retried on alpha as is.