Move test which requires modules to be installed to a separate target

.appveyor.yml

@@ -40,3 +40,4 @@ build_script:
     - c:\cygwin-root\bin\sh -lc "znc --version"
 test_script:
     - c:\cygwin-root\bin\sh -lc "cd $APPVEYOR_BUILD_FOLDER/build; make VERBOSE=1 test < /dev/null"
+    - c:\cygwin-root\bin\sh -lc "cd $APPVEYOR_BUILD_FOLDER/build; make VERBOSE=1 test2 < /dev/null"

.travis.yml

@@ -81,9 +81,9 @@ script:
     - ../configure --enable-perl --enable-python --enable-tcl --enable-cyrus --enable-charset $CFGFLAGS CXXFLAGS="$CXXFLAGS $MYCXXFLAGS" LDFLAGS="$LDFLAGS $MYLDFLAGS"
     - cat config.log
     - make V=1
-    - sudo make install
-      # test after install, because make test currently relies on modules to be installed
     - make V=1 test
+    - sudo make install
+    - make test2
     - cd ..
 after_success:
     - test -r .travis_after_all.py && python .travis_after_all.py || echo No .travis_after_all.py found

Makefile.in

@@ -222,10 +222,13 @@ uninstall:
 
 test: unittest
 	$(Q)./unittest
+
+test2:
 	# The following is only for coverage to see that we also have modules, which (theoretically) should be tested too.
 	# It exploits the fact that ZNC opens every module on startup to check whether it was installed properly or not.
 	# This should be replaced by a better integration test at some point.
 	# Also it uses files at <prefix>/lib/znc, which is less than ideal, especially from build scripts of distros.
+	# That's why it's a separate make target.
 	$(Q)./znc --makeconf --datadir $(shell pwd)/empty-dir-for-test < /dev/null || true
 
 -include $(wildcard .depend/*.dep)