fix: decode Latin-1 WHOIS replies

Replies were always decoded as UTF-8, so registries that answer in Latin-1
(.br, .fi) lost every accented character to U+FFFD ("N�cleo de Informação").
Fall back to windows-1252 when the reply isn't valid UTF-8; a character cut off
by a read timeout doesn't count against UTF-8. Checked live: registro.br and
traficom.fi now decode cleanly.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

src/whois/client.test.ts

@@ -50,3 +50,21 @@ describe("query transformation for non-English WHOIS servers", () => {
     expect(capturedQuery).toBe("hairtect.jp/e");
   });
 });
+
+describe("decodeWhoisReply", () => {
+  it("decodes UTF-8, and Latin-1 replies without replacement characters", async () => {
+    const { decodeWhoisReply } = await import("./client");
+    expect(decodeWhoisReply(Buffer.from("owner: Núcleo de Informação", "utf8"))).toBe(
+      "owner: Núcleo de Informação",
+    );
+    // .br and .fi send Latin-1
+    expect(decodeWhoisReply(Buffer.from("owner: Núcleo de Informação", "latin1"))).toBe(
+      "owner: Núcleo de Informação",
+    );
+    // UTF-8 cut off mid-character by a read timeout stays UTF-8
+    const cut = Buffer.from("name: Liikenne- ja viestintävirasto", "utf8").subarray(0, -1);
+    expect(decodeWhoisReply(Buffer.concat([cut, Buffer.from([0xc3])]))).toBe(
+      "name: Liikenne- ja viestintävirast",
+    );
+  });
+});

src/whois/client.ts

@@ -30,6 +30,19 @@ export interface WhoisTransportOptions {
   blockPrivateAddresses?: boolean;
 }
 
+/**
+ * Decode a WHOIS reply as UTF-8, or as Latin-1 (windows-1252, its WHATWG superset) when it isn't
+ * valid UTF-8: some registries (.br, .fi) answer in Latin-1, which UTF-8 decoding turns into U+FFFD
+ * for good. A character cut off at the end (a read timeout) doesn't count against UTF-8.
+ */
+export function decodeWhoisReply(bytes: Uint8Array): string {
+  try {
+    return new TextDecoder("utf-8", { fatal: true }).decode(bytes, { stream: true });
+  } catch {
+    return new TextDecoder("windows-1252").decode(bytes);
+  }
+}
+
 /**
  * Perform a WHOIS query against an RFC 3912 server over TCP 43.
  * Returns the raw text and the server used.
@@ -95,7 +108,7 @@ async function queryTcp(
     let done = false;
     let timer: ReturnType<typeof setTimeout> | undefined;
 
-    const text = () => Buffer.concat(chunks).toString("utf8");
+    const text = () => decodeWhoisReply(Buffer.concat(chunks));
     const finish = (settle: () => void) => {
       if (done) return;
       done = true;