Merge bitcoin/bitcoin#31901: contrib: Add deterministic-unittest-coverage

fa99c3b test: Exclude SeedStartup from coverage counts (MarcoFalke)
fa579d6 contrib: Add deterministic-unittest-coverage (MarcoFalke)
fa3940b contrib: deterministic-fuzz-coverage fixups (MarcoFalke)
faf905b doc: Remove unused -fPIC (MarcoFalke)
fa1e0a7 gitignore: target/ (MarcoFalke)

Pull request description:

  The `contrib/devtools/test_deterministic_coverage.sh` script is problematic:

  * It is written in bash. This can lead to issues when running with the ancient bash version shipped by macOS by default, or can lead to other compatibility issues, such as bitcoin/bitcoin#31588 (comment). Also, pipefail isn't set, so IO errors may be silently ignored.
  * It is based on gcov. This can lead to issues, such as bitcoin/bitcoin#31588 (review) (possibly due to prefix-map), or bitcoin/bitcoin#31588 (comment) (gcovr processing error), or bitcoin/bitcoin#31588 (review) (gcovr assertion error).
  * The script is severely outdated, with the last update to `NON_DETERMINISTIC_TESTS` being in the prior decade.

  Instead of patching around all issues one-by-one, just provide a fresh rewrite, based on the recently added `deterministic-fuzz-coverage` tool based on clang, llvm-cov, and llvm-profdata. (Initial feedback indicates that this is a more promising attempt: bitcoin/bitcoin#31588 (comment) and bitcoin/bitcoin#31588 (comment)).

  The new tool also sets `RANDOM_CTX_SEED=21` as suggested by hodlinator in bitcoin/bitcoin#31588 (comment).

ACKs for top commit:
  Prabhat1308:
    Concept ACK [`fa99c3b`](bitcoin/bitcoin@fa99c3b)
  hodlinator:
    re-ACK fa99c3b
  brunoerg:
    light ACK fa99c3b
  dergoegge:
    tACK fa99c3b
  janb84:
    Concept ACK [fa99c3b](bitcoin/bitcoin@fa99c3b)

Tree-SHA512: 491d5e6413d929395a5c7caea54817bdc1a0e00562c9728a374d4e92f2e2017dba4a770ecdb2e7317e049df9fdeb390d83c90dff9aa5709f97aa3f6a0e70cdb4

Author: fanquake

.gitignore

@@ -15,8 +15,8 @@
 # Previous releases
 /releases
 
-#build tests
-test/lint/test_runner/target/
+# cargo default target dir
+target/
 
 /guix-build-*
 

contrib/devtools/README.md

@@ -11,15 +11,39 @@ A tool to check for non-determinism in fuzz coverage. To get the help, run:
 RUST_BACKTRACE=1 cargo run --manifest-path ./contrib/devtools/deterministic-fuzz-coverage/Cargo.toml -- --help
 ```
 
-To execute the tool, compilation has to be done with the build options
-`-DCMAKE_C_COMPILER='clang' -DCMAKE_CXX_COMPILER='clang++'
--DBUILD_FOR_FUZZING=ON -DCMAKE_CXX_FLAGS='-fPIC -fprofile-instr-generate
--fcoverage-mapping'`. Both llvm-profdata and llvm-cov must be installed. Also,
-the qa-assets repository must have been cloned. Finally, a fuzz target has to
-be picked before running the tool:
+To execute the tool, compilation has to be done with the build options:
 
 ```
-RUST_BACKTRACE=1 cargo run --manifest-path ./contrib/devtools/deterministic-fuzz-coverage/Cargo.toml -- $PWD/build_dir $PWD/qa-assets/corpora-dir fuzz_target_name
+-DCMAKE_C_COMPILER='clang' -DCMAKE_CXX_COMPILER='clang++' -DBUILD_FOR_FUZZING=ON -DCMAKE_CXX_FLAGS='-fprofile-instr-generate -fcoverage-mapping'
+```
+
+Both llvm-profdata and llvm-cov must be installed. Also, the qa-assets
+repository must have been cloned. Finally, a fuzz target has to be picked
+before running the tool:
+
+```
+RUST_BACKTRACE=1 cargo run --manifest-path ./contrib/devtools/deterministic-fuzz-coverage/Cargo.toml -- $PWD/build_dir $PWD/qa-assets/fuzz_corpora fuzz_target_name
+```
+
+deterministic-unittest-coverage
+===========================
+
+A tool to check for non-determinism in unit-test coverage. To get the help, run:
+
+```
+RUST_BACKTRACE=1 cargo run --manifest-path ./contrib/devtools/deterministic-unittest-coverage/Cargo.toml -- --help
+```
+
+To execute the tool, compilation has to be done with the build options:
+
+```
+-DCMAKE_C_COMPILER='clang' -DCMAKE_CXX_COMPILER='clang++' -DCMAKE_CXX_FLAGS='-fprofile-instr-generate -fcoverage-mapping'
+```
+
+Both llvm-profdata and llvm-cov must be installed.
+
+```
+RUST_BACKTRACE=1 cargo run --manifest-path ./contrib/devtools/deterministic-unittest-coverage/Cargo.toml -- $PWD/build_dir <boost unittest filter>
 ```
 
 clang-format-diff.py

contrib/devtools/deterministic-fuzz-coverage/Cargo.lock

@@ -5,25 +5,25 @@
 use std::env;
 use std::fs::{read_dir, File};
 use std::path::Path;
-use std::process::{exit, Command, Stdio};
+use std::process::{exit, Command};
 use std::str;
 
 const LLVM_PROFDATA: &str = "llvm-profdata";
 const LLVM_COV: &str = "llvm-cov";
-const DIFF: &str = "diff";
+const GIT: &str = "git";
 
 fn exit_help(err: &str) -> ! {
     eprintln!("Error: {}", err);
     eprintln!();
-    eprintln!("Usage: program ./build_dir ./qa-assets-corpora-dir fuzz_target");
+    eprintln!("Usage: program ./build_dir ./qa-assets/fuzz_corpora fuzz_target_name");
     eprintln!();
     eprintln!("Refer to the devtools/README.md for more details.");
     exit(1)
 }
 
 fn sanity_check(corpora_dir: &Path, fuzz_exe: &Path) {
-    for tool in [LLVM_PROFDATA, LLVM_COV, DIFF] {
-        let output = Command::new(tool).arg("--version").output();
+    for tool in [LLVM_PROFDATA, LLVM_COV, GIT] {
+        let output = Command::new(tool).arg("--help").output();
         match output {
             Ok(output) if output.status.success() => {}
             _ => {
@@ -135,7 +135,7 @@ fn deterministic_coverage(
             .expect("merge failed")
             .success());
         let cov_file = File::create(&cov_txt_path).expect("Failed to create coverage txt file");
-        let passed = Command::new(LLVM_COV)
+        assert!(Command::new(LLVM_COV)
             .args([
                 "show",
                 "--show-line-counts-or-regions",
@@ -144,34 +144,31 @@ fn deterministic_coverage(
                 &format!("--instr-profile={}", profdata_file.display()),
             ])
             .arg(fuzz_exe)
-            .stdout(Stdio::from(cov_file))
+            .stdout(cov_file)
             .spawn()
             .expect("Failed to execute llvm-cov")
             .wait()
             .expect("Failed to execute llvm-cov")
-            .success();
-        if !passed {
-            panic!("Failed to execute llvm-profdata")
-        }
+            .success());
         cov_txt_path
     };
     let check_diff = |a: &Path, b: &Path, err: &str| {
-        let same = Command::new(DIFF)
-            .arg("--unified")
+        let same = Command::new(GIT)
+            .args(["--no-pager", "diff", "--no-index"])
             .arg(a)
             .arg(b)
             .status()
-            .expect("Failed to execute diff command")
+            .expect("Failed to execute git command")
             .success();
         if !same {
             eprintln!();
-            eprintln!("The coverage was not determinstic between runs.");
+            eprintln!("The coverage was not deterministic between runs.");
             eprintln!("{}", err);
             eprintln!("Exiting.");
             exit(1);
         }
     };
-    // First, check that each fuzz input is determinisic running by itself in a process.
+    // First, check that each fuzz input is deterministic running by itself in a process.
     //
     // This can catch issues and isolate where a single fuzz input triggers non-determinism, but
     // all other fuzz inputs are deterministic.

contrib/devtools/deterministic-fuzz-coverage/src/main.rs

@@ -0,0 +1,6 @@
+[package]
+name = "deterministic-unittest-coverage"
+version = "0.1.0"
+edition = "2021"
+
+[dependencies]

contrib/devtools/deterministic-unittest-coverage/Cargo.lock

@@ -0,0 +1,124 @@
+// Copyright (c) The Bitcoin Core developers
+// Distributed under the MIT software license, see the accompanying
+// file COPYING or https://opensource.org/license/mit/.
+
+use std::env;
+use std::fs::File;
+use std::path::Path;
+use std::process::{exit, Command};
+use std::str;
+
+const LLVM_PROFDATA: &str = "llvm-profdata";
+const LLVM_COV: &str = "llvm-cov";
+const GIT: &str = "git";
+
+fn exit_help(err: &str) -> ! {
+    eprintln!("Error: {}", err);
+    eprintln!();
+    eprintln!("Usage: program ./build_dir boost_unittest_filter");
+    eprintln!();
+    eprintln!("Refer to the devtools/README.md for more details.");
+    exit(1)
+}
+
+fn sanity_check(test_exe: &Path) {
+    for tool in [LLVM_PROFDATA, LLVM_COV, GIT] {
+        let output = Command::new(tool).arg("--help").output();
+        match output {
+            Ok(output) if output.status.success() => {}
+            _ => {
+                exit_help(&format!("The tool {} is not installed", tool));
+            }
+        }
+    }
+    if !test_exe.exists() {
+        exit_help(&format!(
+            "Test executable ({}) not found",
+            test_exe.display()
+        ));
+    }
+}
+
+fn main() {
+    // Parse args
+    let args = env::args().collect::<Vec<_>>();
+    let build_dir = args
+        .get(1)
+        .unwrap_or_else(|| exit_help("Must set build dir"));
+    if build_dir == "--help" {
+        exit_help("--help requested")
+    }
+    let filter = args
+        .get(2)
+        // Require filter for now. In the future it could be optional and the tool could provide a
+        // default filter.
+        .unwrap_or_else(|| exit_help("Must set boost test filter"));
+    if args.get(3).is_some() {
+        exit_help("Too many args")
+    }
+
+    let build_dir = Path::new(build_dir);
+    let test_exe = build_dir.join("src/test/test_bitcoin");
+
+    sanity_check(&test_exe);
+
+    deterministic_coverage(build_dir, &test_exe, filter);
+}
+
+fn deterministic_coverage(build_dir: &Path, test_exe: &Path, filter: &str) {
+    let profraw_file = build_dir.join("test_det_cov.profraw");
+    let profdata_file = build_dir.join("test_det_cov.profdata");
+    let run_single = |run_id: u8| {
+        let cov_txt_path = build_dir.join(format!("test_det_cov.show.{run_id}.txt"));
+        assert!(Command::new(test_exe)
+            .env("LLVM_PROFILE_FILE", &profraw_file)
+            .env("BOOST_TEST_RUN_FILTERS", filter)
+            .env("RANDOM_CTX_SEED", "21")
+            .status()
+            .expect("test failed")
+            .success());
+        assert!(Command::new(LLVM_PROFDATA)
+            .arg("merge")
+            .arg("--sparse")
+            .arg(&profraw_file)
+            .arg("-o")
+            .arg(&profdata_file)
+            .status()
+            .expect("merge failed")
+            .success());
+        let cov_file = File::create(&cov_txt_path).expect("Failed to create coverage txt file");
+        assert!(Command::new(LLVM_COV)
+            .args([
+                "show",
+                "--show-line-counts-or-regions",
+                "--show-branches=count",
+                "--show-expansions",
+                &format!("--instr-profile={}", profdata_file.display()),
+            ])
+            .arg(test_exe)
+            .stdout(cov_file)
+            .status()
+            .expect("llvm-cov failed")
+            .success());
+        cov_txt_path
+    };
+    let check_diff = |a: &Path, b: &Path| {
+        let same = Command::new(GIT)
+            .args(["--no-pager", "diff", "--no-index"])
+            .arg(a)
+            .arg(b)
+            .status()
+            .expect("Failed to execute git command")
+            .success();
+        if !same {
+            eprintln!();
+            eprintln!("The coverage was not deterministic between runs.");
+            eprintln!("Exiting.");
+            exit(1);
+        }
+    };
+    let r0 = run_single(0);
+    let r1 = run_single(1);
+    check_diff(&r0, &r1);
+    println!("The coverage was deterministic across two runs.");
+}