Merge pull request #60 from go-git/conformance-tests

conformance: expand the upstream-test gate from t2008+t5308 to five tests / 42 cases

Author: pjbgf

.entire/settings.json

@@ -0,0 +1,10 @@
+{
+  "enabled": true,
+  "telemetry": true,
+  "strategy_options": {
+    "checkpoint_remote": {
+      "provider": "github",
+      "repo": "go-git/entire"
+    }
+  }
+}

.github/workflows/build.yml

@@ -58,6 +58,7 @@ jobs:
         shell: bash
         env:
           GO_GIT_REF: ${{ inputs.go_git_ref }}
+          CONFORMANCE_VERBOSE: "1"
         run: make conformance
       - name: Upload TAP results
         if: failure()

Makefile

@@ -10,7 +10,7 @@ $(GOLANGCI):
 
 .PHONY: build
 build:
-	go build -o build/ ./...
+	go build -o build/bin/ ./...
 
 validate: validate-lint validate-dirty
 

README.md

@@ -2,6 +2,24 @@
 
 This provides a Go Git CLI binary using `go-git`.
 
+## Purpose
+
+The CLI exposes key features from `go-git` so it can be assessed and tested
+against upstream Git on a like-for-like basis. Each subcommand is a thin
+wrapper over a `go-git` API, so the behaviour exercised by `gogit <command>`
+is the behaviour that `go-git` consumers will see.
+
+## Bridging gaps in `go-git`
+
+When upstream Git tests require a feature that `go-git` does not yet
+expose, the gap is filled temporarily in `internal/<package>/`. This
+directory is a staging area whose layout mirrors the eventual go-git
+home (e.g. `internal/plumbing/format/idxfile` is shaped for upstream's
+`plumbing/format/idxfile`). Not everything that lives here belongs in
+`go-git` — some bits are test-only shims — but the layering keeps
+upstream tests that exercise the bulk of `go-git` unblocked while the
+upstreaming conversation is in progress.
+
 ## Installation
 
 ```bash
@@ -14,7 +32,41 @@ go install github.com/go-git/cli/cmd/gogit
 gogit <command> [flags]
 ```
 
+## Conformance testing
+
+`make conformance` runs a curated set of upstream Git tests against
+`gogit`, using upstream's own `t/` harness. The runner
+(`conformance/run.sh`) builds `gogit`, symlinks it as `git` in a
+staging directory, and points `GIT_TEST_INSTALLED` there so the
+upstream framework picks our binary up unchanged.
+
+The upstream test source comes from either:
+
+- A local checkout pointed at by `$GIT_SRC` (e.g.
+  `GIT_SRC=$HOME/git/go-git/git make conformance`), or
+- A shallow clone of `github.com/git/git`'s default branch, re-fetched
+  on every run so upstream behaviour drift surfaces immediately.
+
+The set of curated tests is `conformance/tests.txt` — one filename per
+line, `#` comments are ignored:
+
+```
+t2008-checkout-subdir.sh
+t5308-pack-detect-duplicates.sh
+t5325-reverse-index.sh
+```
+
+Individual tests can be run directly:
+
+```bash
+./conformance/run.sh t2008-checkout-subdir.sh        # one test, all cases
+./conformance/run.sh t2008-checkout-subdir.sh 1,3    # one test, selected cases
+```
+
+`$GO_GIT_REF` (commit SHA, tag, or branch) overrides the `go.mod` pin
+for a single run so the gate can be evaluated against an in-flight
+`go-git` change.
+
 ## License
 
 This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
-

cmd/gogit-test-tool/delta.go

@@ -0,0 +1,38 @@
+package main
+
+import (
+	"errors"
+	"fmt"
+	"os"
+
+	"github.com/go-git/go-git/v6/plumbing/format/packfile"
+)
+
+// runDelta implements `test-tool delta {-d|-p} <base> <delta> <out>`. We only
+// need -p (apply); upstream's -d (compute) is not used by the curated tests.
+func runDelta(args []string) error {
+	if len(args) != 4 || args[0] != "-p" {
+		return errors.New("usage: delta -p <base> <delta> <output>")
+	}
+
+	base, err := os.ReadFile(args[1])
+	if err != nil {
+		return fmt.Errorf("read base %s: %w", args[1], err)
+	}
+
+	delta, err := os.ReadFile(args[2])
+	if err != nil {
+		return fmt.Errorf("read delta %s: %w", args[2], err)
+	}
+
+	result, err := packfile.PatchDelta(base, delta)
+	if err != nil {
+		return fmt.Errorf("apply delta: %w", err)
+	}
+
+	if err := os.WriteFile(args[3], result, 0o644); err != nil {
+		return fmt.Errorf("write %s: %w", args[3], err)
+	}
+
+	return nil
+}

cmd/gogit-test-tool/delta_test.go

@@ -0,0 +1,68 @@
+package main
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/go-git/go-git/v6/plumbing/format/packfile"
+)
+
+func TestDeltaApplyRoundTrip(t *testing.T) {
+	t.Parallel()
+	dir := t.TempDir()
+
+	src := []byte("hello, world\nthis is the base text\n")
+	tgt := []byte("hello, world\nthis is the target text with edits\n")
+
+	deltaBytes := packfile.DiffDelta(src, tgt)
+
+	srcPath := filepath.Join(dir, "src")
+	deltaPath := filepath.Join(dir, "delta")
+	outPath := filepath.Join(dir, "out")
+
+	if err := os.WriteFile(srcPath, src, 0o644); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := os.WriteFile(deltaPath, deltaBytes, 0o644); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := runDelta([]string{"-p", srcPath, deltaPath, outPath}); err != nil {
+		t.Fatalf("runDelta: %v", err)
+	}
+
+	got, err := os.ReadFile(outPath)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if string(got) != string(tgt) {
+		t.Fatalf("round-trip mismatch:\n got: %q\nwant: %q", got, tgt)
+	}
+}
+
+func TestDeltaApplyRejectsCorrupt(t *testing.T) {
+	t.Parallel()
+	dir := t.TempDir()
+
+	src := []byte("base")
+	corruptDelta := []byte{0xff, 0xff, 0xff} // not a valid delta header
+
+	srcPath := filepath.Join(dir, "src")
+	deltaPath := filepath.Join(dir, "delta")
+	outPath := filepath.Join(dir, "out")
+
+	if err := os.WriteFile(srcPath, src, 0o644); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := os.WriteFile(deltaPath, corruptDelta, 0o644); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := runDelta([]string{"-p", srcPath, deltaPath, outPath}); err == nil {
+		t.Fatal("expected non-nil error on corrupt delta")
+	}
+}

cmd/gogit-test-tool/genrandom.go

@@ -0,0 +1,58 @@
+package main
+
+import (
+	"bufio"
+	"errors"
+	"fmt"
+	"io"
+	"strconv"
+)
+
+// genRandom emits length deterministic bytes derived from seed, byte-for-byte
+// equivalent to upstream's t/helper/test-genrandom.c. The seed loop in C is
+// `do { next = next*11 + *c; } while (*c++);` — the body runs once per byte
+// INCLUDING the terminating NUL. The PRNG is the POSIX.1-2001 rand() LCG:
+// next = next*1103515245 + 12345; output byte = (next>>16) & 0xff. All
+// arithmetic is uint64 to match `unsigned long` on 64-bit hosts.
+func genRandom(w io.Writer, seed string, length uint64) error {
+	var next uint64
+
+	for _, c := range []byte(seed) {
+		next = next*11 + uint64(c)
+	}
+	// Fold the implicit NUL terminator (matching C do/while semantics).
+	next *= 11
+
+	bw := bufio.NewWriter(w)
+	defer bw.Flush()
+
+	for range length {
+		next = next*1103515245 + 12345
+
+		if err := bw.WriteByte(byte((next >> 16) & 0xff)); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+func runGenRandom(args []string) error {
+	if len(args) < 1 || len(args) > 2 {
+		return errors.New("usage: genrandom <seed_string> [<size>]")
+	}
+
+	seed := args[0]
+	length := ^uint64(0) // ULONG_MAX equivalent
+
+	if len(args) == 2 {
+		v, err := strconv.ParseUint(args[1], 10, 64)
+		if err != nil {
+			return fmt.Errorf("cannot parse size %q: %w", args[1], err)
+		}
+
+		length = v
+	}
+
+	return genRandom(stdoutWriter(), seed, length)
+}

cmd/gogit-test-tool/genrandom_test.go

@@ -0,0 +1,46 @@
+package main
+
+import (
+	"bytes"
+	"testing"
+)
+
+func TestGenRandomMatchesUpstream(t *testing.T) {
+	t.Parallel()
+
+	var buf bytes.Buffer
+
+	if err := genRandom(&buf, "foo", 8); err != nil {
+		t.Fatalf("genRandom: %v", err)
+	}
+
+	// Golden bytes locked in from genRandom("foo", 8) output and verified
+	// against a hand-trace of the algorithm for the first two bytes.
+	want := []byte{0xd3, 0x1c, 0x75, 0x5b, 0xc4, 0x0f, 0x9d, 0xd0}
+
+	if !bytes.Equal(buf.Bytes(), want) {
+		t.Fatalf("genRandom(\"foo\", 8) = %x; want %x", buf.Bytes(), want)
+	}
+}
+
+func TestGenRandomDeterministic(t *testing.T) {
+	t.Parallel()
+
+	var a, b bytes.Buffer
+
+	if err := genRandom(&a, "seed", 64); err != nil {
+		t.Fatalf("first: %v", err)
+	}
+
+	if err := genRandom(&b, "seed", 64); err != nil {
+		t.Fatalf("second: %v", err)
+	}
+
+	if !bytes.Equal(a.Bytes(), b.Bytes()) {
+		t.Fatal("two invocations with same seed/length differ")
+	}
+
+	if a.Len() != 64 {
+		t.Fatalf("length = %d want 64", a.Len())
+	}
+}

cmd/gogit-test-tool/main.go

@@ -0,0 +1,51 @@
+// Package main is the gogit-test-tool helper: a drop-in for upstream's
+// t/helper/test-tool used during conformance runs. Subcommands implemented
+// here mirror the upstream subcommands the curated tests actually exercise.
+package main
+
+import (
+	"errors"
+	"fmt"
+	"io"
+	"os"
+)
+
+func main() {
+	if err := run(os.Args[1:]); err != nil {
+		fmt.Fprintln(os.Stderr, err)
+		os.Exit(1)
+	}
+}
+
+func run(args []string) error {
+	if len(args) == 0 {
+		return errors.New("usage: gogit-test-tool <subcommand> [args...]")
+	}
+
+	sub, rest := args[0], args[1:]
+
+	switch sub {
+	case "genrandom":
+		return runGenRandom(rest)
+	case "delta":
+		return runDelta(rest)
+	case "sha1":
+		return runSHA1(rest)
+	case "sha256":
+		return runSHA256(rest)
+	case "date":
+		return runDate(rest)
+	case "path-utils":
+		return runPathUtils(rest)
+	case "env-helper":
+		return runEnvHelper(rest)
+	default:
+		return fmt.Errorf("unimplemented subcommand: %s", sub)
+	}
+}
+
+// stdoutWriter returns the writer that subcommands emit raw output to.
+// Indirected so tests can drop a buffer in if needed (not done in v1).
+func stdoutWriter() io.Writer {
+	return os.Stdout
+}