test: add end-to-end CLI coverage

.github/workflows/release.yml

@@ -75,6 +75,12 @@ jobs:
           mkdir -p dist
           go build -ldflags "-X main.Version=$VERSION" -o dist/compiledb ./cmd/compiledb
           ./dist/compiledb --help
+          mkdir -p dist/smoke
+          printf 'int main(void) { return 0; }\n' > dist/smoke/main.c
+          printf 'cc -c main.c -o main.o\n' > dist/smoke/build.log
+          (cd dist/smoke && ../compiledb --parse build.log)
+          grep -F '"file": "main.c"' dist/smoke/compile_commands.json
+          rm -rf dist/smoke
           tar cJf dist/compiledb-linux-amd64.txz -C dist compiledb
 
       - uses: actions/upload-artifact@v4
@@ -100,6 +106,12 @@ jobs:
           mkdir -p dist
           go build -ldflags "-X main.Version=$VERSION" -o dist/compiledb ./cmd/compiledb
           ./dist/compiledb --help
+          mkdir -p dist/smoke
+          printf 'int main(void) { return 0; }\n' > dist/smoke/main.c
+          printf 'cc -c main.c -o main.o\n' > dist/smoke/build.log
+          (cd dist/smoke && ../compiledb --parse build.log)
+          grep -F '"file": "main.c"' dist/smoke/compile_commands.json
+          rm -rf dist/smoke
           tar cJf dist/compiledb-linux-arm64.txz -C dist compiledb
 
       - uses: actions/upload-artifact@v4
@@ -128,6 +140,12 @@ jobs:
           mkdir -p dist
           go build -ldflags "-X main.Version=$VERSION" -o dist/compiledb.exe ./cmd/compiledb
           ./dist/compiledb.exe --help
+          mkdir -p dist/smoke
+          printf 'int main(void) { return 0; }\n' > dist/smoke/main.c
+          printf 'cc -c main.c -o main.o\n' > dist/smoke/build.log
+          (cd dist/smoke && ../compiledb.exe --parse build.log)
+          grep -F '"file": "main.c"' dist/smoke/compile_commands.json
+          rm -rf dist/smoke
           (cd dist && 7z a compiledb-windows-amd64.zip compiledb.exe)
 
       - uses: actions/upload-artifact@v4
@@ -153,6 +171,12 @@ jobs:
           mkdir -p dist
           go build -ldflags "-X main.Version=$VERSION" -o dist/compiledb ./cmd/compiledb
           ./dist/compiledb --help
+          mkdir -p dist/smoke
+          printf 'int main(void) { return 0; }\n' > dist/smoke/main.c
+          printf 'cc -c main.c -o main.o\n' > dist/smoke/build.log
+          (cd dist/smoke && ../compiledb --parse build.log)
+          grep -F '"file": "main.c"' dist/smoke/compile_commands.json
+          rm -rf dist/smoke
           tar cJf dist/compiledb-darwin-arm64.txz -C dist compiledb
 
       - uses: actions/upload-artifact@v4

README.md

@@ -364,40 +364,22 @@ response files remain opaque.
 The parser applies fixed limits to individual physical lines and response-file
 expansion:
 
-- A physical build-log line is limited to 100 MiB.
+- A physical build-log line is limited to 100 MiB. An oversized line is
+  reported and skipped without stopping later lines from being parsed.
 - One response file is limited to 8 MiB.
 - Aggregate response-file input for one compiler command is limited to 32 MiB.
 - Response-file nesting is limited to 32 levels and 256 files.
 - Expanded compiler arguments are limited to 100000 elements.
 - Expanded arguments and their copies across multi-source entries have a
   64 MiB budget.
 
-These are not whole-process resource limits. The total build-log size and the
-output produced by programs in backtick expressions do not have fixed limits;
-large inputs or unbounded command output can consume substantial memory.
-
 ## Related Projects
 
 - [nickdiego/compiledb][python-compiledb] is the Python project that originally
   inspired this Go rewrite.
 - [clangd][clangd] and [clang-tidy][clang-tidy] are common consumers of the
   generated `compile_commands.json` file.
 
-## Development
-
-Run the full local verification suite with:
-
-```sh
-test -z "$(gofmt -l .)"
-go vet ./...
-go test ./...
-go test -race ./...
-go build -o /tmp/compiledb-go ./cmd/compiledb
-```
-
-Some tests modify the current directory, standard streams, and process-global
-state, so those tests must not run in parallel.
-
 ## License
 
 [GNU General Public License v3.0](LICENSE).

internal/build_log_benchmark_test.go

@@ -32,10 +32,7 @@ func BenchmarkBuildLogScan(b *testing.B) {
 			b.ResetTimer()
 
 			for range b.N {
-				lines, err := scanBuildLog(data)
-				if err != nil {
-					b.Fatal(err)
-				}
+				lines := scanBuildLog(data)
 				runtime.KeepAlive(lines)
 			}
 		})
@@ -47,21 +44,15 @@ func BenchmarkBuildLogMergeLogicalLines(b *testing.B) {
 	for _, size := range benchmarkBuildLogSizes {
 		b.Run(size.name, func(b *testing.B) {
 			data := buildLogBenchmarkData(size.size, chunk)
-			lines, err := scanBuildLog(data)
-			if err != nil {
-				b.Fatal(err)
-			}
+			lines := scanBuildLog(data)
 			data = nil
 			runtime.GC()
 
 			b.SetBytes(size.size)
 			b.ReportAllocs()
 			b.ResetTimer()
 			for range b.N {
-				logicalLines, issues := mergeLogicalLines(lines)
-				if len(issues) != 0 {
-					b.Fatalf("unexpected logical line issues: %#v", issues)
-				}
+				logicalLines := mergeLogicalLines(lines)
 				runtime.KeepAlive(logicalLines)
 			}
 		})
@@ -127,11 +118,8 @@ func benchmarkDiscoveryOutput(b *testing.B, size int64, chunk []byte) {
 		if err := writeBuildLogBenchmarkData(&stdout, size, chunk); err != nil {
 			b.Fatal(err)
 		}
-		buildLog, err := scanBuildLog(stdout.Bytes())
-		if err != nil {
-			b.Fatal(err)
-		}
-		tool.Parse(buildLog)
+		buildLog := scanBuildLog(stdout.Bytes())
+		tool.parseBuildLog(buildLog)
 		if tool.StatusCode != 0 {
 			b.Fatalf("parser status: %d", tool.StatusCode)
 		}

internal/init.go

@@ -1,7 +1,6 @@
 package internal
 
 import (
-	"bufio"
 	"bytes"
 	"context"
 	"encoding/json"
@@ -40,6 +39,7 @@ type Tool struct {
 	predefinedMacros       map[string][]string
 	compilerCommand        func(name string, arg ...string) *exec.Cmd
 	compilerCommandContext func(context.Context, string, ...string) *exec.Cmd
+	buildLogLineLimit      int
 	makeDirectoryMarkers   bool
 }
 
@@ -60,6 +60,13 @@ func (t *Tool) operationContext() context.Context {
 	return context.Background()
 }
 
+func (t *Tool) physicalLineLimit() int {
+	if t.buildLogLineLimit > 0 {
+		return t.buildLogLineLimit
+	}
+	return maxBuildLogLineSize
+}
+
 func (t *Tool) compilationDatabaseBuildDir() string {
 	if t.Config.BuildDir != "" {
 		return compilationDatabaseBuildDir(t.Config.BuildDir)
@@ -368,23 +375,42 @@ func (t *Tool) Generate() {
 		t.Logger.Debugf("Build from stdin")
 	}
 
-	buildLog, err := scanBuildLog(data)
-	if err != nil {
-		t.Logger.Fatalf("read build log failed: %v", err)
+	if t.operationContext().Err() != nil {
+		t.StatusCode = contextExitCode(t.operationContext())
+		return
 	}
+	buildLog := scanBuildLogWithLimit(data, t.physicalLineLimit())
 	if t.operationContext().Err() != nil {
 		t.StatusCode = contextExitCode(t.operationContext())
 		return
 	}
-	t.Parse(buildLog)
+	t.parseBuildLog(buildLog)
 }
 
-func scanBuildLog(data []byte) ([]string, error) {
-	var lines []string
-	scanner := bufio.NewScanner(bytes.NewReader(data))
-	scanner.Buffer(make([]byte, 1024*1024), maxBuildLogLineSize)
-	for scanner.Scan() {
-		lines = append(lines, scanner.Text())
+func scanBuildLog(data []byte) []buildLogLine {
+	return scanBuildLogWithLimit(data, maxBuildLogLineSize)
+}
+
+func scanBuildLogWithLimit(data []byte, limit int) []buildLogLine {
+	lineCount := bytes.Count(data, []byte{'\n'})
+	if len(data) > 0 && data[len(data)-1] != '\n' {
+		lineCount++
+	}
+	lines := make([]buildLogLine, 0, lineCount)
+	for len(data) > 0 {
+		line := data
+		if newline := bytes.IndexByte(data, '\n'); newline >= 0 {
+			line = data[:newline]
+			data = data[newline+1:]
+		} else {
+			data = nil
+		}
+		line = bytes.TrimSuffix(line, []byte{'\r'})
+		if len(line) > limit {
+			lines = append(lines, buildLogLine{raw: line, oversized: true, limit: limit})
+			continue
+		}
+		lines = append(lines, buildLogLine{text: string(line)})
 	}
-	return lines, scanner.Err()
+	return lines
 }

internal/init_test.go

@@ -1,6 +1,7 @@
 package internal
 
 import (
+	"bytes"
 	"context"
 	"encoding/json"
 	"fmt"
@@ -54,6 +55,66 @@ func TestGenerateFromStdinDoesNotPanic(t *testing.T) {
 	tool.Generate()
 }
 
+func TestScanBuildLogPhysicalLineLimit(t *testing.T) {
+	if reason := buildLogLineLimitReason(maxBuildLogLineSize); reason != "physical line exceeds 100 MiB limit" {
+		t.Fatalf("unexpected production line-limit diagnostic: %q", reason)
+	}
+
+	const limit = 8
+	for name, input := range map[string]string{
+		"LF":   strings.Repeat("x", limit) + "\n",
+		"CRLF": strings.Repeat("x", limit) + "\r\n",
+		"EOF":  strings.Repeat("x", limit),
+	} {
+		t.Run(name, func(t *testing.T) {
+			lines := scanBuildLogWithLimit([]byte(input), limit)
+			if len(lines) != 1 || lines[0].oversized || lines[0].text != strings.Repeat("x", limit) {
+				t.Fatalf("line at limit was not preserved: %#v", lines)
+			}
+		})
+	}
+
+	lines := scanBuildLogWithLimit([]byte(strings.Repeat("x", limit+1)+"\r\nok\n"), limit)
+	if len(lines) != 2 || !lines[0].oversized || lines[0].limit != limit || lines[0].text != "" ||
+		lines[1].oversized || lines[1].text != "ok" {
+		t.Fatalf("oversized line was not discarded cleanly: %#v", lines)
+	}
+}
+
+func TestGenerateSkipsOversizedPhysicalLine(t *testing.T) {
+	projectDir := t.TempDir()
+	buildLog := filepath.Join(projectDir, "build.log")
+	outputFile := filepath.Join(projectDir, "compile_commands.json")
+	contents := strings.Repeat("x", 32) + "\\\ncc -c joined.c\ncc -c valid.c\n"
+	if err := os.WriteFile(buildLog, []byte(contents), 0o644); err != nil {
+		t.Fatalf("write build log failed: %v", err)
+	}
+
+	var logs bytes.Buffer
+	tool := newTestTool(t, Config{
+		InputFile:    buildLog,
+		OutputFile:   outputFile,
+		RegexCompile: RegexCompile,
+		RegexFile:    RegexFile,
+		NoStrict:     true,
+	})
+	tool.buildLogLineLimit = 32
+	tool.Logger.SetLevel(log.ErrorLevel)
+	tool.Logger.SetOutput(&logs)
+	tool.Generate()
+
+	commands := readCompilerTestCommands(t, outputFile)
+	if tool.StatusCode != 0 || len(commands) != 1 || commands[0].File != "valid.c" {
+		t.Fatalf("oversized build-log line stopped Generate: status=%d commands=%#v", tool.StatusCode, commands)
+	}
+	diagnostic := logs.String()
+	for _, want := range []string{"build log line 1", "cwd", "physical line exceeds 32 byte limit", "at byte 32"} {
+		if !strings.Contains(diagnostic, want) {
+			t.Fatalf("physical line diagnostic lacks %q: %q", want, diagnostic)
+		}
+	}
+}
+
 func TestGenerateResolvesResponseFileFromBuildLogDirectory(t *testing.T) {
 	projectDir := t.TempDir()
 	buildLog := filepath.Join(projectDir, "build.log")

internal/make_wrap.go

@@ -290,18 +290,12 @@ func (t *Tool) runDiscoveryMake(cmd *exec.Cmd) discoveryResult {
 		return result
 	}
 
-	buildLog, scanErr := scanBuildLog(stdoutBuf.Bytes())
-	if scanErr != nil {
-		result.status = 1
-		t.Logger.Errorf("read dry-run output failed: %v", scanErr)
-		return result
-	}
 	clone := *t
 	clone.makeDirectoryMarkers = true
 	if !t.Config.NoBuild {
 		clone.Logger = loggerAtLevel(t.Logger, logrus.ErrorLevel)
 	}
-	clone.Parse(buildLog)
+	clone.parseBuildLog(scanBuildLogWithLimit(stdoutBuf.Bytes(), clone.physicalLineLimit()))
 	result.parserStatus = clone.StatusCode
 	return result
 }