diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
index 2ff2614dc5c9652ab60434fece031bad25a8f044..031a9b8596976ad116de358c470db9ec6d226584 100644
--- a/.gitlab-ci.yml
+++ b/.gitlab-ci.yml
@@ -165,6 +165,7 @@ e-acsl-dev-tests:
   script:
     - ./nix/build-proxy.sh eva-$CONFIG-tests
   <<: *coverage_artifacts
+  # Do not add OUT here, the parallel rule in Eva domains overload variables
 
 eva-default-tests:
   variables:
@@ -353,7 +354,7 @@ internal:
   except:
     - schedules
 
-internal_nightly:
+internal-nightly:
   <<: *internal_template
   # The Opam target may affect this job
   timeout: 2h
diff --git a/nix/frama-c.nix b/nix/frama-c.nix
index 2aab3256ac2305aff99c5083713502e2d50bd60b..771bbff79686a397902de06ead1ccc16bd724c32 100644
--- a/nix/frama-c.nix
+++ b/nix/frama-c.nix
@@ -1,4 +1,10 @@
-# Nix
+# Frama-C build derivation
+# Input variables:
+# - `cover` (defaults to `true`) that indicates whether Frama-C should be
+#   compiled with coverage instrumentation,
+# - `release_mode` (defaults to `false`) that indicates whether Frama-C should
+#   be compiled in release mode.
+
 { lib
 , stdenvNoCC # for E-ACSL
 , fetchurl
diff --git a/nix/mk_tests.nix b/nix/mk_tests.nix
index 1754f291e51b51d27fbae62bca47c0477643e129..0a7b37dcc03a4c242e47d0e37968bd74c598eb95 100644
--- a/nix/mk_tests.nix
+++ b/nix/mk_tests.nix
@@ -17,6 +17,12 @@
 #   receives an additional build-input 'alt-ergo'. Furthermore, it configures
 #   Why3 before build phase and export the WP global cache. Note however that
 #   this cache is used only if the tests use the option '-wp-cache-env'
+#
+# - cover (optional, defaults to 'true')
+#   Indicates whether the tests should generate coverage files. BEWARE! If you
+#   disable this, make sure that you use a Frama-C version built without
+#   coverage instrumentation, else Frama-C might be build several times during
+#   the pipeline. See for example default-config-tests.nix
 
 { lib
 , alt-ergo