Bug#907720: cminpack: autopkgtest fails with glibc 2.28 due to precision changes

[Adam Conrad]

Package: cminpack
Version: 1.3.6-2
Severity: normal

In Ubuntu, as we've moved on the glibc 2.28, we've seen the cminpack[1]
test suite "fail", due to what appears to be improved precision in the
results.

While this can be fixed by adjusting the ref files (as upstream has done
more than once in the past, it looks like), I question if this test as
it's executed is actually an appropriate autopkgtest.  It fails on most
architectures due to slightly different precision, and it will "regress"
on amd64 every time upstream improves precision there.

It might be a more appropriate test if it had a filter between running
the examples and examining the results that allowed for jitter tolerance
rather than just a plain diff looking for identical reference results.

If that's not feasible, maybe a more lightweight test that just makes
sure the package isn't entirely broken, but doesn't run the examples and
look for specific outputs might be a saner autopkgtest.

... Adam

[1] http://autopkgtest.ubuntu.com/packages/c/cminpack

-- System Information:
Debian Release: buster/sid
  APT prefers cosmic
  APT policy: (500, 'cosmic')
Architecture: amd64 (x86_64)
Foreign Architectures: i386

Kernel: Linux 4.17.0-7-lowlatency (SMP w/4 CPU cores; PREEMPT)
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8), LANGUAGE=en_CA:en (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Init: systemd (via /run/systemd/system)
LSM: AppArmor: enabled