gitlab-ci runners report a job as correctly done even though exit code was non-zero

Description of the problem

When compiling a large project, I occasionally get successful runs of the test suite even if the test suite failed (run using .gitlab-ci.yml). An example can be seen in the build: https://gitlab.com/gnutls/gnutls/pipelines/5149934

While on the overview page all looks ok, clicking on the libvirt build we see: https://gitlab.com/gnutls/gnutls/builds/6693360

FAILED
FAIL virnettlssessiontest (exit status: 1)

+ exit 1
error: Bad exit status from /var/tmp/rpm-tmp.kj5l20 (%check)


RPM build errors:
    Bad exit status from /var/tmp/rpm-tmp.kj5l20 (%check)

However the runner detects:

Build succeeded

The actual .gitlab-ci.yml script contains (the full is on this link):

Fedora/apps/libvirt:
  stage: compile
  image: $CI_REGISTRY/$PROJECT_GROUP:$FEDORA_BUILD
  script:
  - dnf install fedpkg numactl-devel sheepdog xen-devel numad -y
  - dnf builddep libvirt -y
  - make autoreconf && mkdir -p build && cd build &&
    ../configure --disable-non-suiteb-curves --prefix=/usr --libdir=/usr/lib64 --disable-cxx --disable-do
    make -j$(nproc) && make install
  - fedpkg clone -a libvirt
  - cd libvirt && fedpkg local
  tags:
  - shared
  dependencies:
  - build/Fedora24
  except:
  - tags

and the command that failed is the last one, (cd libvirt && fedpkg local), which should have failed since fedpkg's exit code is 1.

Is that a limitation of the .gitlab-ci.yml format for the script directive (e.g., in the documentation only one script command is mentioned rather than the multiple used in that build.

Which Group/Project (with full path) is experiencing the issue?

gnutls/gnutls

Approximate date/time when the error occurred.

2016-11-27