Ephemeral COPR build failed. @containers/packit-build please check.

Output is visible at the end of https://cirrus-ci.com/task/6362001716281344.

@jankaluza Can we use just normal gcov so that we don't need to add a new package dependency and maybe adding a separate test-coverage target instead?

Something like this?

diff --git a/Makefile b/Makefile
index 90ff8f0..72b25a0 100644
--- a/Makefile
+++ b/Makefile
@@ -79,6 +79,15 @@ test-binary: bin/conmon
 test: bin/conmon
        CONMON_BINARY="$(MAKEFILE_PATH)bin/conmon" test/run-tests.sh
 
+.PHONY: test-coverage
+test-coverage: clean
+       $(MAKE) bin/conmon DEBUGFLAG="--coverage"
+       CONMON_BINARY="$(MAKEFILE_PATH)bin/conmon" test/run-tests.sh
+       @echo "=== Coverage Summary ==="
+       @for file in src/*.c; do \
+               gcov $$file 2>/dev/null | grep -E "(^File|^Lines executed)" | paste - - | sed 's/File '\''src\/\([^.]*\)\.c'\''.*Lines executed:\([0-9.]*%\).*/\1.c: \2/'; \
+       done | grep -E "\.c: [0-9]" | sort
+
 bin:
        mkdir -p bin
 
@@ -90,7 +99,7 @@ docs:
 
 .PHONY: clean
 clean:
-       rm -rf bin/ src/*.o
+       rm -rf bin/ src/*.o src/*.gcda src/*.gcno *.gcov
        $(MAKE) -C docs clean
 
 .PHONY: install install.bin install.crio install.podman podman crio

Separate test-coverage target is probably good idea.

For the plain gcov - it does not print the missing line numbers which is one of the reasons I wanted to add this test. I'm using that to find out untested parts of the code for which I should write a tests. I know this is not the only metric to use, but it gives a good overview.

The .gcov file-format is quite complex when it comes to uncovered lines. We could do some approximation of what gcovr is doing with:

	@echo "=== Missing coverage ==="
	@for file in *.gcov; do \
		awk -F':[ \t]*' '/#####|%%%%%/ { sub(/-.*/,"",$$2); print $$2 }' $$file | sort -nu | paste -sd, | awk -v f=$$file '{print f ": " $$0}'; \
	done | sort

But it does not handle some of the .gcov file-format features which are not trivial to handle without more complex code. I think gcovr is not that big to not justify its use. But I can also imagine using gcovr myself locally for my use-case and using just gcov in our tests for an overview.

@jnovy , there's a dedicated test-coverage target now and the code to fallback to gcov if gcovr is not found.

I believe the cri-o oldstable failure is not related? I will check it more to be sure.

LGTM now, thanks.