Remove coverage for now

Author: gleocadie

.github/workflows/coverage.yml

@@ -7,7 +7,7 @@ on:
       - main
 
 jobs:
-  coverage-linux:
+  coverage:
     runs-on: ubuntu-latest
     env:
       CARGO_TERM_COLOR: always
@@ -46,38 +46,4 @@ jobs:
         with:
           token: ${{ secrets.CODECOV_TOKEN }}
           files: lcov.info
-          flags: linux
-          fail_ci_if_error: true
-
-  coverage-windows:
-    runs-on: windows-latest
-    env:
-      CARGO_TERM_COLOR: always
-      CARGO_INCREMENTAL: 0
-    steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # 4.2.2
-      - name: Install Rust
-        run: rustup install nightly-2024-12-16 && rustup default nightly-2024-12-16
-      - name: Install cargo-llvm-cov
-        uses: taiki-e/install-action@2c41309d51ede152b6f2ee6bf3b71e6dc9a8b7df # 2.49.27
-        with:
-          tool: [email protected],[email protected]
-      - name: Cache [rust]
-        uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # 2.8.1
-        with:
-          cache-targets: true
-          cache-bin: true
-      - name: Generate code coverage for Windows tests
-        shell: bash
-        run: |
-          cargo llvm-cov --all-features --workspace --no-report nextest
-          cargo llvm-cov --all-features --workspace --no-report --doc
-          cargo llvm-cov report --doctests --lcov --output-path lcov-windows.info
-          cargo clean
-      - name: Upload Windows coverage to Codecov
-        uses: codecov/codecov-action@5a1091511ad55cbe89839c7260b706298ca349f7 # 5.5.1
-        with:
-          token: ${{ secrets.CODECOV_TOKEN }}
-          files: lcov-windows.info
-          flags: windows
           fail_ci_if_error: true

bin_tests/doc/WINDOWS_BIN_TESTS_ARCHITECTURE.md

@@ -22,13 +22,7 @@ Following WER's architecture, crash handling occurs in a separate process. This
 - Ability to use full system APIs in the handler
 - Realistic testing of production crash handling flow
 
-**3. Coverage Collection Support**
-The architecture is designed from the ground up to support code coverage collection through:
-- Environment variable propagation to spawned processes
-- Multiple process profraw file generation
-- Integration with cargo-llvm-cov workflow
-
-**4. Platform Parity with Unix Tests**
+**3. Platform Parity with Unix Tests**
 While implementation differs fundamentally, the testing API mirrors the Unix bin_tests structure to maintain consistency across platforms.
 
 ## Architectural Components
@@ -124,29 +118,6 @@ This approach mimics real WER behavior where the out-of-process handler reads cr
 - Support timeouts for fault tolerance
 - Atomic signaling with no race conditions
 
-### 5. Coverage Collection Architecture
-
-**Challenge**: Coverage from spawned processes must contribute to overall coverage report.
-
-**Solution Strategy**:
-
-**Build-Time Instrumentation**:
-When `cargo llvm-cov` builds test binaries, it sets `RUSTFLAGS` with `-C instrument-coverage`. The build system propagates these flags to spawned `cargo build` commands to ensure all test binaries are instrumented.
-
-**Runtime Coverage Propagation**:
-The `LLVM_PROFILE_FILE` environment variable contains patterns like `%p` (process ID) that are expanded by each process at runtime:
-- Main test process: `cargo-test-1000-abc.profraw`
-- Spawned crash binary: `cargo-test-1001-def.profraw`
-- WER simulator: `cargo-test-1002-ghi.profraw`
-
-**Workflow Integration**:
-Separate Windows coverage job runs in parallel with Linux:
-- Both upload coverage to Codecov
-- Codecov merges reports based on commit SHA
-- Flags allow filtering by platform
-
-**Design Rationale**: This approach requires no code changes in test binaries—coverage "just works" when environment variables are propagated correctly.
-
 ## Testing Workflow
 
 ### Standard Windows Test Flow
@@ -192,31 +163,6 @@ Separate Windows coverage job runs in parallel with Linux:
 - Remove temporary files
 - Clean up registry keys (if applicable)
 
-### Parallel Coverage Collection Flow
-
-**Coverage CI Workflow**:
-```
-coverage-linux (ubuntu-latest)
-  ├─ Build with instrumentation
-  ├─ Run Unix bin_tests
-  ├─ Spawned processes write profraw files
-  ├─ Generate lcov.info
-  └─ Upload to Codecov (flag: linux)
-
-coverage-windows (windows-latest)
-  ├─ Build with instrumentation
-  ├─ Run Windows bin_tests
-  ├─ Spawned processes write profraw files
-  ├─ WER simulator writes profraw file
-  ├─ Generate lcov-windows.info
-  └─ Upload to Codecov (flag: windows)
-
-Codecov Server:
-  ├─ Receives both uploads (same commit SHA)
-  ├─ Merges line-by-line coverage
-  └─ Presents unified report
-```
-
 ## Design Trade-offs
 
 ### Process Isolation vs Simplicity
@@ -309,10 +255,6 @@ Codecov Server:
 - Unix: Simpler synchronization, immediate results
 - Windows: Complex synchronization, asynchronous validation
 
-**Coverage Collection**:
-- Unix: Main process + receiver process profraw files
-- Windows: Main process + crash binary + WER simulator profraw files
-
 ### Implications for Test Design
 
 **Test Timing**:
@@ -451,23 +393,12 @@ When tests fail, they should:
 
 **Implementation**: Rich error types, verbose logging, debug files written to temp directories.
 
-### Coverage as First-Class Concern
-
-Test infrastructure considers coverage collection from the start:
-- Environment variable propagation built-in
-- Multiple process profraw files expected
-- CI workflow includes coverage verification
-- Documentation explains coverage architecture
-
-**Rationale**: Without coverage, we can't know what the tests actually exercise.
-
 ## Conclusion
 
 The Windows bin_tests architecture balances several competing concerns:
 - **Realism**: Match production WER behavior as closely as possible
 - **Testability**: Enable testing without WER service in CI
 - **Safety**: Handle crashes in separate process like real WER
-- **Observability**: Collect coverage from all spawned processes
 - **Maintainability**: Clear abstractions and extension points
 
 The result is a comprehensive testing framework that validates Windows crash tracking in realistic scenarios while supporting both local development and CI environments. The architecture's flexibility allows future extension while maintaining strong type safety and clear separation of concerns.

bin_tests/src/lib.rs

@@ -195,43 +195,3 @@ macro_rules! timeit {
         res
     }};
 }
-
-/// Propagates coverage environment variables to spawned processes.
-///
-/// This function enables coverage collection from integration tests that spawn separate
-/// processes (like bin_tests spawning crashtracker_bin_test). It propagates the
-/// `LLVM_PROFILE_FILE` environment variable which contains patterns like `%p` (process ID)
-/// that the LLVM profiling runtime expands at runtime, ensuring each process writes to a
-/// unique coverage file.
-///
-/// # How It Works
-///
-/// 1. Parent propagates the pattern: `LLVM_PROFILE_FILE="path/cargo-test-%p-%m.profraw"`
-/// 2. Each process expands at runtime: Parent PID 1000 → `cargo-test-1000-abc.profraw`
-/// 3. Coverage report merges all `.profraw` files
-///
-/// # Platform Support
-///
-/// - **Unix/Linux**: Fully supported, used in CI coverage collection
-/// - **Windows**: Fully supported, enables Windows bin_tests coverage
-///
-/// # Arguments
-/// * `cmd` - The Command to add environment variables to (before spawning)
-///
-/// # Example
-/// ```no_run
-/// use bin_tests::propagate_coverage_env;
-/// use std::process::Command;
-///
-/// let mut cmd = Command::new("path/to/test_binary");
-/// propagate_coverage_env(&mut cmd);
-/// let child = cmd.spawn().expect("Failed to spawn");
-/// ```
-pub fn propagate_coverage_env(cmd: &mut process::Command) {
-    // LLVM_PROFILE_FILE tells the instrumented binary where to write coverage data.
-    // The LLVM profiling runtime expands patterns like %p (PID) and %m (module signature)
-    // at runtime to ensure each spawned process writes to a unique file.
-    if let Ok(profile_file) = env::var("LLVM_PROFILE_FILE") {
-        cmd.env("LLVM_PROFILE_FILE", profile_file);
-    }
-}

bin_tests/src/test_runner_windows.rs

@@ -275,9 +275,6 @@ where
         .arg(&crash_ready_event_name)
         .arg(&info_file);
 
-    // Pass coverage-related environment variables to spawned processes
-    crate::propagate_coverage_env(&mut cmd);
-
     eprintln!(
         "[TEST_RUNNER] Spawning crash binary: {}",
         binary_path.display()
@@ -351,9 +348,6 @@ where
         .arg(&simulator_ready_event_name)
         .arg(&fixtures.output_dir); // Pass output directory for debug files
 
-    // Pass coverage-related environment variables to spawned WER simulator
-    crate::propagate_coverage_env(&mut simulator_cmd);
-
     let simulator_process = simulator_cmd
         .spawn()
         .context("Failed to spawn WER simulator")?;