main excludes checked with the defaults' case and fold; Latin-1 decoding memory; hardlink election documented

1 files changed, 16 insertions, 0 deletions

diff --git a/internal/extract/plain_test.go b/internal/extract/plain_test.go
index 5752bf8..c8e1008 100644
--- a/internal/extract/plain_test.go
+++ b/internal/extract/plain_test.go
@@ -3,6 +3,7 @@
 package extract
 
 import (
+	"bytes"
 	"context"
 	"errors"
 	"os"
@@ -228,3 +229,18 @@ func TestSniffRuneStraddlingSampleBoundary(t *testing.T) {
 		t.Errorf("got tail %q, want it to end in %q", got[len(got)-8:], want)
 	}
 }
+
+// TestLatin1DecodingMemory: decoding Latin-1 allocates about what the UTF-8
+// result needs (at most two bytes per input byte), not a four-byte rune per
+// input byte on top of it (triage 28d).
+func TestLatin1DecodingMemory(t *testing.T) {
+	data := bytes.Repeat([]byte("Gr\xfc\xdfe "), 16000)
+	r := testing.Benchmark(func(b *testing.B) {
+		for range b.N {
+			decodeLatin1(data)
+		}
+	})
+	if got, limit := r.AllocedBytesPerOp(), int64(3*len(data)); got > limit {
+		t.Errorf("decoding %d bytes allocates %d bytes, over %d", len(data), got, limit)
+	}
+}