Re: [PATCH bpf-next v7 06/11] selftests/bpf: Tolerate test file compilation failures

[sashiko-bot@kernel.org]

Sashiko AI review found 2 potential issue(s):
- [High] Tolerating test compilation failures breaks the build at the linker step.
- [High] Omitting test objects causes the test suite runner to abort at startup.
--

commit 3b7eac5b82169326acf885f820b425d42e131d71
Author: Ricardo B. Marlière <rbm@suse.com>

selftests/bpf: Tolerate test file compilation failures

This commit modifies the selftests build system to make per-test compilation
failures non-fatal when permitted. It removes the output object on failure
and skips the post-processing step for missing objects.

> Make the per-test compilation non-fatal: remove the output object on
> failure and print a SKIP-TEST marker to stderr. Guard the BTFIDS
> post-processing step so it is skipped when the object file is absent. The
> linker step will later ignore absent objects, allowing the remaining tests
> to build and run.

Is this description accurate regarding the linker step?

During linking, the Makefile rules pass the literal list of prerequisites to
the compiler. When a compilation fails and the object file is deleted, that
missing file is still passed to the linker.

This causes a fatal "No such file or directory" error that halts the build.

I notice this is addressed later in the series by the patch "selftests/bpf:
Allow test_progs to link with a partial object set", but it causes a build
regression in this specific commit.

> diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile
> index b3eaa4a93707c..5ad50d3ff2f8c 100644
> --- a/tools/testing/selftests/bpf/Makefile
> +++ b/tools/testing/selftests/bpf/Makefile
> @@ -708,11 +708,14 @@ $(TRUNNER_TEST_OBJS): $(TRUNNER_OUTPUT)/%.test.o:			\
>  		      $(TRUNNER_TESTS_DIR)/%.c				\
>  		      | $(TRUNNER_OUTPUT)/%.test.d
>  	$$(call msg,TEST-OBJ,$(TRUNNER_BINARY),$$@)
> -	$(Q)cd $$(@D) && $$(CC) -I. $$(CFLAGS) -MMD -MT $$@ -c $(CURDIR)/$$< $$(LDLIBS) -o $$(@F)
> +	$(Q)cd $$(@D) && $$(CC) -I. $$(CFLAGS) -MMD -MT $$@ -c $(CURDIR)/$$< $$(LDLIBS) -o $$(@F) $(if $(PERMISSIVE),|| \
> +		($(RM) $$(@F); printf '  %-12s %s\n' 'SKIP-TEST' '$$(notdir $$@)' 1>&2))

How does the test suite handle the missing objects at runtime?

The test framework generates the test header by scanning the source files
directly, so a test definition is emitted even if its object file was
skipped.

Because the compiled object is missing, the weak function pointers for that
test resolve to a null pointer during linking. When the test runner validates
these pointers at startup, it prints an error and aborts the entire test
framework setup.

This prevents the remaining tests from running.

I see this is resolved later in the series by the patch "selftests/bpf: Skip
tests whose objects were not built", but it introduces a runtime regression
here.

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260416-selftests-bpf_misconfig-v7-0-a078e18012e4@suse.com?part=6