fix(native): normalize otool line endings and strip plist JSON BOM (#533)

src/native/parsers/otool.ts

@@ -44,8 +44,19 @@ interface ParsedLoadCommands {
   readonly flags: Set<string>;
 }
 
+/**
+ * otool emits LF, but a capture replayed from another host or toolchain can
+ * carry CRLF. The section/command boundaries below match a bare `\n`, so a
+ * stray `\r` would swallow the terminator and silently empty `sections` while
+ * leaking section fields into the command. Normalizing once at the boundary
+ * keeps every parser below working on one line ending.
+ */
+const normalizeLineEndings = (output: string): string =>
+  output.replaceAll("\r\n", "\n");
+
 /** Parse stable fields from `otool -l`, preserving unknown command fields. */
-export const parseOtoolLoadCommands = (output: string) => {
+export const parseOtoolLoadCommands = (raw: string) => {
+  const output = normalizeLineEndings(raw);
   const headerTokens = parseHeaderTokens(output);
   const state = createLoadCommandState(headerTokens);
   for (const block of output.split(/(?=Load command \d+)/u))

src/native/parsers/plist.ts

@@ -4,7 +4,8 @@ const plistObject = z.record(z.string(), z.unknown());
 
 /** Parse plutil JSON output and project stable bundle metadata. */
 export const parsePlistJson = (output: string) => {
-  const value: unknown = JSON.parse(output);
+  // plutil output can carry a UTF-8 BOM, which JSON.parse rejects outright.
+  const value: unknown = JSON.parse(output.replace(/^\uFEFF/u, ""));
   const object = plistObject.safeParse(value);
   const field = (name: string): string | null => {
     if (!object.success) return null;

tests/boundary/providers/native/nativeTextRobustness.test.ts

@@ -0,0 +1,36 @@
+import { expect, it } from "vitest";
+
+import { parseOtoolLoadCommands } from "../../../../src/native/parsers/otool.js";
+import { parsePlistJson } from "../../../../src/native/parsers/plist.js";
+import { nativeFixture } from "../../../fixtures/nativeCommands.js";
+
+const toCrlf = (value: string): string => value.replaceAll("\n", "\r\n");
+
+it("keeps section blocks separate under CRLF captures", async () => {
+  const lf = parseOtoolLoadCommands(await nativeFixture("otool-load.txt"));
+  const crlf = parseOtoolLoadCommands(
+    toCrlf(await nativeFixture("otool-load.txt")),
+  );
+  expect(crlf.segments).toMatchObject(lf.segments);
+  const crlfText = crlf.segments.find(({ name }) => name === "__TEXT");
+  expect(crlfText?.sections.length).toBeGreaterThan(0);
+  expect(crlfText?.sections.map(({ name }) => name)).toContain("__text");
+});
+
+it("does not leak section fields into command fields under CRLF captures", async () => {
+  const crlf = parseOtoolLoadCommands(
+    toCrlf(await nativeFixture("otool-load.txt")),
+  );
+  const segment = crlf.commands.find(
+    (command) => command.kind === "LC_SEGMENT_64",
+  );
+  expect(segment?.fields.sectname).toBeUndefined();
+  expect(segment?.fields["time stamp"]).toBeUndefined();
+});
+
+it("accepts a byte-order mark before plist JSON", () => {
+  const parsed = parsePlistJson(
+    '\uFEFF{"CFBundleIdentifier":"com.owned.fixture"}',
+  );
+  expect(parsed.bundle.identifier).toBe("com.owned.fixture");
+});