Re: package builds crashing under fakeroot

[Ian Jackson <ijackson@chiark.greenend.org.uk>]

Samuel Thibault writes ("Re: package builds crashing under fakeroot"):
> Jakub Wilk, on Tue 04 Oct 2016 10:55:36 +0200, wrote:
> > >with no override_dh_auto_build and override_dh_auto_test rules. By default
> > >the tests are run under fakeroot,
> > 
> > No, they're not.
> 
> Err, I've seen them do. This actually triggered fixing some fakeroot
> bugs on hurd-i386 due to various package suddenly failing to build after
> some debhelper upgrade (I don't remember which versione exactly), just
> because the testsuite was failing to run inside fakeroot.

Clearly the different people in this thread have seen different
results.  This is most likely because different people aren't looking
at the same packages or considering the same kinds of tests.

I don't think anyone is disputing that running tests during the binary
targets is a bad idea.

We have reports that there are (or have been) packages where this
happenened and no-seems to be suggesting that a general bug of this
kind has been fixed anywhere.  So there are probably more cases
besides openmpi.

Although people have accused dh (or "some dh_* command") of doing
this, there has been little concrete evidence.  I think we are only
going to be able to fix this problem by actually debugging it.

Alastair, can you figure out why the openmpi tests are running during
rules binary rather than rules build ?

Ian.

-- 
Ian Jackson <ijackson@chiark.greenend.org.uk>   These opinions are my own.

If I emailed you from an address @fyvzl.net or @evade.org.uk, that is
a private address which bypasses my fierce spamfilter.