Update CI and use nextest (#59)

This updates the CI definitions a bit:

- Invokes `rustup` directly instead of using an action
- Updates the `setup-python` action to the latest version
- Adds a separate `doctest` run to run doctests and emit `cargo doc` errors
- Switches to `nextest` for running test, which also has builtin junit output
- Always run tests with coverage

Author: Swatinem

.config/nextest.toml

@@ -0,0 +1,2 @@
+[profile.default.junit]
+path = "core-test-results.xml"

.github/workflows/ci.yml

@@ -6,6 +6,9 @@ on:
       - main
   pull_request:
 
+env:
+  RUSTFLAGS: -Dwarnings
+
 concurrency:
   group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
   cancel-in-progress: true
@@ -15,81 +18,58 @@ jobs:
     name: Lint (rustfmt + clippy)
     runs-on: ubuntu-latest
     steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-      - name: Install Rust
-        uses: dtolnay/rust-toolchain@nightly
-        with:
-          toolchain: nightly
-          components: clippy, rustfmt
-      - name: Run lint
-        run: make lint.rust
+      - uses: actions/checkout@v4
+
+      - run: rustup toolchain install nightly --profile minimal --component rustfmt --component clippy --no-self-update
+
+      - run: cargo fmt --all -- --check
+      - run: cargo clippy --all-features --workspace --tests --examples -- -D clippy::all
 
   lint-python:
     name: Lint Python (ruff, mypy)
     runs-on: ubuntu-latest
     steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-      - uses: actions/setup-python@v3
-      - name: Set up venv
-        run: make ci.setup_venv
-      - name: Build and run lint
-        run: |
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-python@v5
+      - run: make ci.setup_venv
+
+      - run: |
           maturin develop
           make lint.python
 
-  test:
-    name: Test
+  doctest:
+    name: Documentation (and Tests)
     runs-on: ubuntu-latest
-    strategy:
-      fail-fast: false
+    env:
+      RUSTDOCFLAGS: -Dwarnings
     steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-      - name: Install Rust
-        uses: dtolnay/rust-toolchain@nightly
-        with:
-          toolchain: nightly
-      - name: Run Rust tests
-        run: |
-          cargo test
+      - uses: actions/checkout@v4
 
-      - uses: actions/setup-python@v3
-      - name: Set up venv
-        run: make ci.setup_venv
-      - name: Run Python tests
-        run: |
-          maturin develop
-          pytest
-
-  # This job runs tests, generates coverage data, and generates JUnit test
-  # results in a single test invocation and then uploads it all to Codecov.
-  # However, it doesn't print test results to stdout. If Codecov's failed test
-  # reporting is solid and we never need to see the results in the CI logs, we
-  # can delete the "normal" test step and just use this.
-  test-for-codecov:
-    name: Test (Upload to Codecov)
+      - run: rustup toolchain install nightly --profile minimal --no-self-update
+
+      - run: cargo test --workspace --all-features --doc
+      - run: cargo doc --workspace --all-features --document-private-items --no-deps
+
+  test:
+    name: Test
     runs-on: ubuntu-latest
-    strategy:
-      fail-fast: false
     steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-      - name: Install Rust
-        uses: dtolnay/rust-toolchain@nightly
-        with:
-          toolchain: nightly
-      - name: Install `cargo llvm-cov`
-        uses: taiki-e/install-action@cargo-llvm-cov
-      - name: Run tests
-        run: |
-          cargo install cargo2junit
-          cargo llvm-cov --lcov --output-path core.lcov -- -Z unstable-options --format json --report-time | cargo2junit > core-test-results.xml
+      - uses: actions/checkout@v4
+
+      - run: rustup toolchain install nightly --profile minimal --no-self-update
+      - uses: taiki-e/install-action@cargo-llvm-cov
+      - uses: taiki-e/install-action@nextest
+
+      # FIXME(swatinem): We should pass `--all-targets` to also compile and tests benchmarks
+      # Though currently `divan` does not support all CLI arguments as used by `nextest`,
+      # and benchmarks are unbearably slow anyway, so its not feasible to run in debug builds.
+      - run: cargo llvm-cov nextest --lcov --output-path core.lcov --workspace --all-features
+      - run: mv target/nextest/default/core-test-results.xml .
+
+      - uses: actions/setup-python@v5
+      - run: make ci.setup_venv
 
-      - uses: actions/setup-python@v3
-      - name: Set up venv
-        run: make ci.setup_venv
       - name: Run Python tests
         run: |
           # Clear prior profile data

.gitignore

@@ -28,3 +28,5 @@ __pycache__/
 
 .git
 lcov.info
+/*test-results.xml
+/*.lcov

Cargo.toml

@@ -1,5 +1,4 @@
 [workspace]
-
 resolver = "2"
 members = ["bindings", "core", "test_utils"]
 default-members = ["core"]

core/benches/pyreport.rs

@@ -2,7 +2,7 @@ use std::collections::HashMap;
 
 use codecov_rs::{
     parsers::pyreport::{chunks, report_json},
-    report::test::{TestReport, TestReportBuilder},
+    test_utils::test_report::{TestReport, TestReportBuilder},
 };
 use divan::Bencher;
 use test_utils::fixtures::{read_fixture, FixtureFormat::Pyreport, FixtureSize::Large};

core/src/lib.rs

@@ -7,4 +7,4 @@ pub mod parsers;
 pub mod error;
 
 #[cfg(any(test, feature = "testing"))]
-mod test_utils;
+pub mod test_utils;

core/src/parsers/pyreport/chunks.rs

@@ -20,6 +20,8 @@ use super::{
     },
     utils,
 };
+#[cfg(doc)]
+use crate::report::models;
 use crate::report::{
     pyreport::{types::*, CHUNKS_FILE_END_OF_CHUNK, CHUNKS_FILE_HEADER_TERMINATOR},
     Report, ReportBuilder,
@@ -28,7 +30,7 @@ use crate::report::{
 #[derive(PartialEq, Debug)]
 pub struct ChunkCtx {
     /// The index of this chunk in the overall sequence of chunks tells us which
-    /// [`crate::report::models::SourceFile`] this chunk corresponds to.
+    /// [`SourceFile`](models::SourceFile) this chunk corresponds to.
     pub index: usize,
 
     /// Each line in a chunk corresponds to a line in the source file.
@@ -44,7 +46,7 @@ pub struct ParseCtx<R: Report, B: ReportBuilder<R>> {
 
     /// Tracks the labels that we've already added to the report. The key is the
     /// identifier for the label inside the chunks file and the value is the
-    /// ID of the [`crate::report::models::Context`] we created for it in
+    /// ID of the [`Context`](models::Context) we created for it in
     /// the output. If a `"labels_index"` key is present in the chunks file
     /// header, this is populated all at once and the key is a numeric ID.
     /// Otherwise, this is populated as new labels are encountered and the key
@@ -55,12 +57,12 @@ pub struct ParseCtx<R: Report, B: ReportBuilder<R>> {
     pub chunk: ChunkCtx,
 
     /// The output of the report JSON parser includes a map from `chunk_index`
-    /// to the ID of the [`crate::report::models::SourceFile`] that the
+    /// to the ID of the [`SourceFile`](models::SourceFile) that the
     /// chunk corresponds to.
     pub report_json_files: HashMap<usize, i64>,
 
     /// The output of the report JSON parser includes a map from `session_id` to
-    /// the ID of the [`crate::report::models::Context`] that the session
+    /// the ID of the [`Context`](models::Context) that the session
     /// corresponds to.
     pub report_json_sessions: HashMap<usize, i64>,
 }
@@ -121,8 +123,8 @@ pub fn coverage<S: StrStream, R: Report, B: ReportBuilder<R>>(
     .parse_next(buf)
 }
 
-/// Parses the coverage type described by a `ReportLine`. Beware: this field may
-/// be inaccurate.
+/// Parses the coverage type described by a [`ReportLine`]. Beware: this field
+/// may be inaccurate.
 ///
 /// For example, in a chunks file for a Go project, the "coverage type" field is
 /// always `null` when some of the values in the "coverage" field indicate the
@@ -141,7 +143,8 @@ pub fn coverage_type<S: StrStream, R: Report, B: ReportBuilder<R>>(
     .parse_next(buf)
 }
 
-/// Parses value of the "complexity" field in a `ReportLine` or `LineSession`.
+/// Parses value of the "complexity" field in a [`ReportLine`] or
+/// [`LineSession`].
 ///
 /// Examples: `1`, `3`, `[0, 1]`, `[2, 2]`
 pub fn complexity<S: StrStream, R: Report, B: ReportBuilder<R>>(
@@ -218,9 +221,9 @@ where
     .parse_next(buf)
 }
 
-/// Parses values in the "partials" field of a `LineSession`. These values don't
-/// necessarily have to do with partial branch coverage; what they describe is
-/// the coverage status of different subspans of a single line.
+/// Parses values in the "partials" field of a [`LineSession`]. These values
+/// don't necessarily have to do with partial branch coverage; what they
+/// describe is the coverage status of different subspans of a single line.
 ///
 /// Examples:
 /// - `[null, 10, 0]`: This line was not covered from its start until column 10
@@ -254,7 +257,7 @@ pub fn partial_spans<S: StrStream, R: Report, B: ReportBuilder<R>>(
 }
 
 /// Parses a [`LineSession`]. Each [`LineSession`] corresponds to a
-/// [`crate::report::models::CoverageSample`] in the output report.
+/// [`CoverageSample`](models::CoverageSample) in the output report.
 ///
 /// A [`ReportLine`] has a [`LineSession`] for each upload ("session") sent to
 /// us for a commit. The [`LineSession`] contains the coverage measurements for
@@ -346,7 +349,7 @@ pub fn label<S: StrStream, R: Report, B: ReportBuilder<R>>(
 ///
 /// Technically `_coverage_type` is optional, but the way it gets serialized
 /// when it's missing is identical to the way we serialize
-/// [`crate::report::models::CoverageType::Line`] so there's no way to tell
+/// [`CoverageType::Line`] so there's no way to tell
 /// which it is when deserializing.
 pub fn coverage_datapoint<S: StrStream, R: Report, B: ReportBuilder<R>>(
     buf: &mut ReportOutputStream<S, R, B>,
@@ -369,11 +372,12 @@ pub fn coverage_datapoint<S: StrStream, R: Report, B: ReportBuilder<R>>(
 
 /// Parses a [`ReportLine`]. A [`ReportLine`] itself does not correspond to
 /// anything in the output, but it's an umbrella that includes all of the data
-/// tied to a line/[`CoverageSample`].
+/// tied to a line/[`CoverageSample`](models::CoverageSample).
 ///
 /// This parser performs all the writes it can to the output
-/// stream and only returns a `ReportLine` for tests. The `report_line_or_empty`
-/// parser which wraps this and supports empty lines returns `Ok(())`.
+/// stream and only returns a [`ReportLine`] for tests. The
+/// `report_line_or_empty` parser which wraps this and supports empty lines
+/// returns `Ok(())`.
 pub fn report_line<'a, S, R: Report, B: ReportBuilder<R>>(
     buf: &mut ReportOutputStream<S, R, B>,
 ) -> PResult<ReportLine>

core/src/parsers/pyreport/utils.rs

@@ -53,7 +53,8 @@ fn format_pyreport_branch(branch: &MissingBranch) -> (models::BranchFormat, Stri
 }
 
 /// Each [`LineSession`] corresponds to a single [`models::CoverageSample`].
-/// Each [`CoverageSample`] _may_ (but won't always) have:
+/// Each [`CoverageSample`](models::CoverageSample) _may_ (but won't always)
+/// have:
 /// - multiple related [`models::BranchesData`] records, one for each specific
 ///   branch path we have data for
 /// - a single related [`models::MethodData`] if the `LineSession` is for a

core/src/report/pyreport/mod.rs

@@ -11,9 +11,9 @@
  * [`shared/reports/types.py`](https://github.com/codecov/shared/blob/main/shared/reports/types.py),
  * and [`shared/utils/sessions.py`](https://github.com/codecov/shared/blob/main/shared/utils/sessions.py).
  *
- * Parsers that will build a [`SQLiteReport`] from these parts live in
+ * Parsers that will build a [`SqliteReport`] from these parts live in
  * [`crate::parsers::pyreport`] but code that will convert a
- * [`SQLiteReport`] back into a Pyreport lives here.
+ * [`SqliteReport`] back into a Pyreport lives here.
  *
  * # Report JSON
  *

core/src/report/pyreport/types.rs

@@ -2,6 +2,8 @@ use std::collections::HashMap;
 
 pub use super::super::models::CoverageType;
 use crate::parsers::json::JsonVal;
+#[cfg(doc)]
+use crate::report::models;
 
 /// Enum representing the possible values of the "coverage" field in a
 /// ReportLine or LineSession object.
@@ -63,22 +65,22 @@ pub struct Partial {
 }
 
 /// Represents the coverage measurements taken for a specific "session". Each
-/// `LineSession` will correspond to a
-/// [`crate::report::models::CoverageSample`].
+/// [`LineSession`] will correspond to a
+/// [`CoverageSample`](models::CoverageSample).
 ///
 /// Each upload to our system constitutes a "session" and may be tagged with
 /// flags or other context that we want to filter on when viewing report data.
 #[derive(Debug, PartialEq)]
 pub struct LineSession {
     /// This ID indicates which session the measurement was taken in. It can be
     /// used as a key in `buf.state.report_json_sessions` to get the ID of a
-    /// [`crate::report::models::Context`] in order to create a
-    /// [`crate::report::models::ContextAssoc`].
+    /// [`Context`](models::Context) in order to create a
+    /// [`ContextAssoc`](models::ContextAssoc).
     pub session_id: usize,
 
     /// The coverage measurement that was taken in this session. The
-    /// `CoverageType` is "inherited" from the [`ReportLine`] that this
-    /// `LineSession` is a part of.
+    /// [`CoverageType`] is "inherited" from the [`ReportLine`] that this
+    /// [`LineSession`] is a part of.
     pub coverage: PyreportCoverage,
 
     /// A list of specific branches/conditions stemming from this line that were
@@ -106,13 +108,13 @@ pub enum RawLabel {
     /// A numeric ID that was assigned to this label. The original label can be
     /// accessed in the `"labels_index"` key in the chunks file's header.
     /// For our parser's purposes, we can access the ID of the
-    /// [`crate::report::models::Context`] created for this label in
+    /// [`Context`](models::Context) created for this label in
     /// `buf.state.labels_index`.
     LabelId(u32),
 
     /// The name of the label. If we have encountered this label before, it
     /// should be in `buf.state.labels_index` pointing at the ID for a
-    /// [`crate::report::models::Context`]. Otherwise, we should create that
+    /// [`Context`](models::Context). Otherwise, we should create that
     /// `Context` + mapping ourselves.
     LabelName(String),
 }
@@ -124,20 +126,20 @@ pub enum RawLabel {
 pub struct CoverageDatapoint {
     /// This ID indicates which session the measurement was taken in. It can be
     /// used as a key in `buf.state.report_json_sessions` to get the ID of a
-    /// [`crate::report::models::Context`] in order to create a
-    /// [`crate::report::models::ContextAssoc`].
+    /// [`Context`](models::Context) in order to create a
+    /// [`ContextAssoc`](models::ContextAssoc).
     pub session_id: u32,
 
     /// A redundant copy of the coverage measurement for a session. We prefer
     /// the value from the [`LineSession`].
     pub _coverage: PyreportCoverage,
 
-    /// A redundant copy of the `CoverageType`. We use the value from the
+    /// A redundant copy of the [`CoverageType`]. We use the value from the
     /// [`ReportLine`].
     ///
     /// Technically this field is optional, but the way we serialize it when
     /// it's missing is identical to the way we serialize
-    /// [`crate::report::models::CoverageType::Line`] so there's
+    /// [`CoverageType::Line`] so there's
     /// no way to tell which it is when deserializing.
     pub _coverage_type: Option<CoverageType>,
 
@@ -167,7 +169,8 @@ pub struct ReportLine {
     pub coverage_type: CoverageType,
 
     /// The list of measurements taken for this line. Each of these corresponds
-    /// to a [`CoverageSample`] record in a `SqliteReport`.
+    /// to a [`CoverageSample`](models::CoverageSample) record in a
+    /// `SqliteReport`.
     pub sessions: Vec<LineSession>,
 
     /// Long forgotten field that takes up space.
@@ -177,9 +180,9 @@ pub struct ReportLine {
     /// `sessions`.
     pub _complexity: Option<Option<Complexity>>,
 
-    /// The list of [`CoverageDatapoint`]s for this line. `CoverageDatapoint` is
-    /// largely redundant but its `labels` field is the only place where label
-    /// data is recorded (e.g. which test case was running when this
+    /// The list of [`CoverageDatapoint`]s for this line. [`CoverageDatapoint`]
+    /// is largely redundant but its `labels` field is the only place where
+    /// label data is recorded (e.g. which test case was running when this
     /// measurement was collected).
     pub datapoints: Option<Option<HashMap<u32, CoverageDatapoint>>>,
 }