[pkg-go] Bug#976943: golang-github-seccomp-libseccomp-golang: FTBFS on ppc64el (arch:all-only src pkg): dh_auto_test: error: cd obj-powerpc64le-linux-gnu && go test -vet=off -v -p 160 github.com/seccomp/libseccomp-golang returned exit code 1

[Lucas Nussbaum]

On 14/12/20 at 14:36 -0500, Reinhard Tartler wrote:
> 
> 
> > === RUN   TestRuleAddAndLoad
> >     seccomp_test.go:588: Syscall should have returned error code!
> > --- FAIL: TestRuleAddAndLoad (0.00s)
> 
> 
> Source code is here: https://sources.debian.org/src/golang-github-seccomp-libseccomp-golang/0.9.1-2/seccomp_test.go/#L529-L589
> 
> This test is basically loading a seccomp rule and expects that the
> getpid() syscall fails with the rule loaded, but in that test it seems
> to pass. This might mean that seccomp isn't working on that box.
> 
> Lucas, can you elaborate a bit what kind of test system you were using?
> Do you have any explanation what might be going on with the test setup that
> would make seccomp not work or only under certain conditions?

Hi,

This is a system running plain debian stable with an unstable chroot
(using sbuild). I cannot think of something strange. I tried disabling
SMT (hyperthreading) because with it enabled, the system has 160 visible
cores, which breaks some builds. But it did not change anything.

# uname -a
Linux drac-6.grenoble.grid5000.fr 4.19.0-13-powerpc64le #1 SMP Debian 4.19.160-2 (2020-11-28) ppc64le GNU/Linux

I also tried with a Debian testing image and could reproduce it. Kernel
was linux-image-5.9.0-1-powerpc64le  5.9.1-1

Lucas