Bug#710696: severity of 710696 is serious, because it's a FTBFS bug

[Santiago Vila]

On Thu, Sep 29, 2016 at 03:26:54PM +0200, Michael Biebl wrote:

> I don't see what we would gain by disabling the failing tests.
> They are there to catch regressions after all. So this would be a step
> backwards.

Fixing RC bugs have multiple benefits :-) We would avoid the package
from being autoremoved from testing, or having to ask release managers
for permission to use stretch-ignore.

Ok, you still don't think this is RC. Well, even in such case, we
would gain a package that always build. That's everything I ask.

Currently, for me, it is as if this package had an undeclared
"build-depends: buildd.debian.org". We should be able to build the
package in every autobuilder which is sane and not misconfigured.

> If your main objective is to build the package on your own system, you
> can omit the test-suite by setting DEB_BUILD_OPTIONS=nocheck (not su

That's exactly the issue and that's why I consider this to be RC:

If my autobuilder is not misconfigured, I should not have to
special-case this package by setting DEB_BUILD_OPTIONS at all.

My main objective was really to check for "dpkg-buildpackage -A",
but when packages FTBFS for other reasons not related to using -A,
my secondary objective is ensuring that packages build ok,
and this, naturally, include the tests, so I can't just skip them.

> Of course, the better option would be if you can investigate why it
> fails on your particular setup. This would be very much appreciated.
> I'm not saying, that there isn't a race somewhere. The point is that
> disabling is not the answer and it's not necessary.

Yes, of course, I will try to help you to debug this within my possibilities.

[ I'm going to provide the backtrace you asked in the other email ]

But you are making a balance which IMO is a little bit biased. You
don't see the need to disable the failing tests because you have
already downgraded the bug!

Please think about this bug as RC. Your options would be:

A. Invest time and effort to debug it.

B. Disable the test, possibly forwarding the bug upstream so that they
investigate, not you.

Since we are all volunteers, I can't really ask that you do A.

Everything I ask is that the package builds ok, and for that B would
be completely acceptable.

I hope this clarifies a previous comment I did that apparently you
didn't like. If you choose B over A for whatever reason, I will never
say it's wrong, or complain or antyhing like that.

Thanks.