coverage: Provide baseline support for Linux and QNX by castler · Pull Request #698 · eclipse-score/communication

castler · 2026-07-13T18:50:18Z

With this commit we provide baseline support for Linux and QNX coverage.
We can remove manual filtering based on patterns and ensure that only the files (but also all files) from the respective public libraries are covered.

LittleHuba · 2026-07-14T11:24:26Z

+  bazel build --config=qnx //quality/coverage:coverage_scope --output_groups=allowlist
+  ALLOWLIST_FILE="$(bazel info bazel-bin)/quality/coverage/coverage_scope_allowlist.txt"


This is not very portable. Better would be if this target is provided as input artifact via bazel.

It is actually as portable as possible - only better approach is to not do it in this preprocessing step. I have local changes for this refactoring, but this is right now not performant enough.

LittleHuba · 2026-07-14T11:25:56Z

+  BASELINE_LCOV="${TMPDIR_EXTRACT}/baseline_lcov.dat"
+  BASELINE_COUNT=0
+  LCOV_INPUT_LIST="${BUILD_WORKSPACE_DIRECTORY}/bazel-out/_coverage/lcov_files.tmp"
+  if [[ -f "${LCOV_INPUT_LIST}" ]]; then
+    while IFS= read -r lcov_input; do
+      if [[ "${lcov_input}" == *"/baseline_coverage.dat" ]]; then
+        if [[ -f "${lcov_input}" ]]; then
+          printf "\n" >> "${BASELINE_LCOV}"
+          cat "${lcov_input}" >> "${BASELINE_LCOV}"
+          printf "\n" >> "${BASELINE_LCOV}"
+          BASELINE_COUNT=$((BASELINE_COUNT + 1))
+        fi
+        continue
+      fi
+      if [[ "${lcov_input}" == *"/coverage.dat" ]]; then
+        baseline_input="${lcov_input%coverage.dat}baseline_coverage.dat"
+        if [[ -f "${baseline_input}" ]]; then
+          printf "\n" >> "${BASELINE_LCOV}"
+          cat "${baseline_input}" >> "${BASELINE_LCOV}"
+          printf "\n" >> "${BASELINE_LCOV}"
+          BASELINE_COUNT=$((BASELINE_COUNT + 1))
+        fi
+      fi
+    done < <(sort -u "${LCOV_INPUT_LIST}")
+  fi


This section would benefit from a comment that explains what its goal is.

LittleHuba · 2026-07-14T11:29:03Z

+    if filename.startswith("/proc/self/cwd/"):
+        return filename[len("/proc/self/cwd/"):]


Better not duplicate this string.

LittleHuba · 2026-07-14T11:30:18Z

+    if not allowlist:
+        return list(file_entries)


This does not fulfill the contract of the function. Either extend the description of the function or change this to return an empty list.

LittleHuba · 2026-07-14T11:40:24Z

+        line_number = incoming_line["line_number"]
+        incoming_copy = dict(incoming_line)
+        incoming_copy["branches"] = [dict(branch) for branch in incoming_line.get("branches", [])]


Code smell: Based on access-pattern in line 214 incoming_line is a dict. Based on this, the assignment in line 215 results in a shallow copy. This means, in line 216 you do not only modify incoming_copy but also incoming_line.

Possible options:

If the modification of incoming_line is non-destructive, just drop the shallow copy

If the modification of incoming_line is destructive use copy.deepcopy().

LittleHuba · 2026-07-14T11:46:18Z

+def _merge_single_file_entry(target: Dict[str, Any], incoming: Dict[str, Any]) -> None:
+    """Merge one file entry into another in-place."""
+    target_lines = {line["line_number"]: dict(line) for line in target["lines"]}
+    for incoming_line in incoming.get("lines", []):
+        line_number = incoming_line["line_number"]
+        incoming_copy = dict(incoming_line)
+        incoming_copy["branches"] = [dict(branch) for branch in incoming_line.get("branches", [])]
+
+        if line_number not in target_lines:
+            target_lines[line_number] = incoming_copy
+            continue
+
+        existing = target_lines[line_number]
+        existing["count"] = max(existing.get("count", 0), incoming_copy.get("count", 0))
+        existing["branches"] = _merge_branch_entries(
+            existing.get("branches", []),
+            incoming_copy.get("branches", []),
+        )
+
+    target["lines"] = sorted(target_lines.values(), key=lambda line: line["line_number"])
+
+    target_functions = {func["name"]: dict(func) for func in target["functions"]}
+    for incoming_func in incoming.get("functions", []):
+        name = incoming_func["name"]
+        incoming_lineno = incoming_func.get("lineno", 0)
+        incoming_execution_count = incoming_func.get("execution_count", 0)
+
+        if name not in target_functions:
+            target_functions[name] = dict(incoming_func)
+            continue
+
+        existing_func = target_functions[name]
+        if existing_func.get("lineno", 0) == 0 and incoming_lineno != 0:
+            existing_func["lineno"] = incoming_lineno
+        existing_func["execution_count"] = max(
+            existing_func.get("execution_count", 0),
+            incoming_execution_count,
+        )
+
+    target["functions"] = sorted(target_functions.values(), key=lambda func: func["name"])


This function would benefit from splitting into dedicated subfunctions for line and function coverage as well.

LittleHuba · 2026-07-14T11:51:08Z

+
+
+def _merge_single_file_entry(target: Dict[str, Any], incoming: Dict[str, Any]) -> None:
+    """Merge one file entry into another in-place."""


I think it would help to describe the merge strategy here, since it is not straightforward.
Instead of actually merging the counts, you perform a selection of the maximum value between the two files.
This results in completely different counts than a user would expect.

Should this be a concious choice, I would prefer an explaination here.
If this was not concious, I would prefer to actually add the counts (considering overflows) so that the numbers are as a user would expect them.

Replace static filename-regex filtering with a Bazel-derived coverage scope: - add `coverage_scope` rule/aspect to collect dependable-element source files - generate allowlist + baseline object manifest for the reporter - add `reporter_wrapper` rule to wire scope artifacts into reporter invocation - update llvm-cov reporter to consume allowlist/baseline objects and support baseline-only coverage via `--empty-profile` - remove obsolete `filter_regexes.txt` and update coverage docs/config comments

Switch the QNX gcovr report path away from regex-based file filtering to using the Bazel-collected coverage allowlist from coverage_scope. Changes: - generate_coverage_html.sh: Build //quality/coverage:coverage_scope in default config, resolve its allowlist.txt and pass it to lcov_to_html.py via the new --allowlist flag. Remove the dependency on quality/coverage/llvm_cov/filter_regexes.txt for QNX. - lcov_to_html.py: Add --allowlist argument; load the allowlist, normalize LCOV SF: paths to workspace-relative form, and filter entries against the allowlist before handing off to gcovr. Fail fast if an explicit allowlist resolves to an empty set. - lcov_to_html_test.py: Add regression tests for allowlist loading, path normalization/filtering, and the empty-allowlist guard. - BUILD: Add py_test for lcov_to_html_test. - README.md: Document that QNX scope is now driven by the collected allowlist, not regex filters. The coverage_scope target is built without --config=qnx because the dependable element indices it depends on are Linux-only; the resulting workspace-relative allowlist is equally valid for QNX report scoping. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

github-project-automation Bot added this to COM - Communication FT Jul 13, 2026

github-project-automation Bot moved this to Backlog in COM - Communication FT Jul 13, 2026

github-code-quality Bot found potential problems Jul 13, 2026

View reviewed changes

Comment thread quality/coverage/llvm_cov/reporter.py Fixed

castler force-pushed the js_linux_baseline_coverage branch 2 times, most recently from e70b0fe to 0c953aa Compare July 14, 2026 09:09

castler marked this pull request as ready for review July 14, 2026 09:46

castler requested review from LittleHuba, bemerybmw, crimson11, hoe-jo and limdor as code owners July 14, 2026 09:46

LittleHuba reviewed Jul 14, 2026

View reviewed changes

castler force-pushed the js_linux_baseline_coverage branch from 0c953aa to a3ec8d6 Compare July 15, 2026 06:25

castler force-pushed the js_linux_baseline_coverage branch from a3ec8d6 to 26aaf8b Compare July 15, 2026 07:57

castler and others added 2 commits July 15, 2026 10:53

build(toolchain): bump LLVM toolchain to 22.1.7

26aaf8b

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

coverage: Provide baseline support for Linux and QNX#698

coverage: Provide baseline support for Linux and QNX#698
castler wants to merge 3 commits into
mainfrom
js_linux_baseline_coverage

castler commented Jul 13, 2026

Uh oh!

Uh oh!

Uh oh!

LittleHuba Jul 14, 2026

Uh oh!

castler Jul 14, 2026

Uh oh!

LittleHuba Jul 14, 2026

Uh oh!

LittleHuba Jul 14, 2026

Uh oh!

LittleHuba Jul 14, 2026

Uh oh!

LittleHuba Jul 14, 2026

Uh oh!

LittleHuba Jul 14, 2026

Uh oh!

LittleHuba Jul 14, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		bazel build --config=qnx //quality/coverage:coverage_scope --output_groups=allowlist
		ALLOWLIST_FILE="$(bazel info bazel-bin)/quality/coverage/coverage_scope_allowlist.txt"

		if filename.startswith("/proc/self/cwd/"):
		return filename[len("/proc/self/cwd/"):]



		def _merge_single_file_entry(target: Dict[str, Any], incoming: Dict[str, Any]) -> None:
		"""Merge one file entry into another in-place."""

Uh oh!

Conversation

castler commented Jul 13, 2026

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants