fix: file upload file name utf8 issues

backend/src/uploadMiddleware.ts

@@ -12,8 +12,17 @@ const diskStorage = multer.diskStorage({
   },
 });
 
+// Browsers send multipart filenames as raw UTF-8 bytes, but multer decodes them as Latin-1 by
+// default -- `哨子.3mf` would arrive as mojibake, and sanitizeFilename then strips the C1 bytes
+// needed to recover it.
+const defParamCharset = "utf8";
+
 /** Disk-backed multer instance for model file uploads (never memoryStorage — files can be large). */
-export const modelUpload = multer({ storage: diskStorage, limits: { fileSize: IMPORT_MAX_BYTES } });
+export const modelUpload = multer({ storage: diskStorage, limits: { fileSize: IMPORT_MAX_BYTES }, defParamCharset });
 
 /** Small in-memory multer instance for the 8MB-capped client-rendered thumbnail upload. */
-export const thumbnailUpload = multer({ storage: multer.memoryStorage(), limits: { fileSize: 8 * 1024 * 1024 + 1 } });
+export const thumbnailUpload = multer({
+  storage: multer.memoryStorage(),
+  limits: { fileSize: 8 * 1024 * 1024 + 1 },
+  defParamCharset,
+});

backend/src/utils/fileUtils.ts

@@ -28,7 +28,9 @@ export function sanitizeFilename(name: string | null | undefined): string {
  * alone instead of being mangled by the fix.
  */
 function decodeLatin1AsUtf8(value: string): string {
-  if (!/[\u0080-\u00ff]/.test(value)) return value;
+  // Only a pure byte string can be one; anything above U+00FF was already decoded properly, and
+  // Buffer.from(..., "latin1") would silently truncate it.
+  if (!/[\u0080-\u00ff]/.test(value) || /[\u0100-\uffff]/.test(value)) return value;
   const bytes = Buffer.from(value, "latin1");
   const decoded = bytes.toString("utf8");
   return Buffer.from(decoded, "utf8").equals(bytes) ? decoded : value;

backend/tests/importFilename.test.ts

@@ -1,4 +1,7 @@
+import express from "express";
+import request from "supertest";
 import { describe, expect, it } from "vitest";
+import { modelUpload } from "../src/uploadMiddleware";
 import { buildImportFilename, parseContentDisposition, sanitizeFilename } from "../src/utils/fileUtils";
 
 // HTTP header values are ISO-8859-1 (RFC 7230 section 3.2.4). A server that puts raw UTF-8 bytes
@@ -59,3 +62,24 @@ describe("buildImportFilename", () => {
     expect(/[\u0000-\u001f\u007f-\u009f]/.test(name)).toBe(false);
   });
 });
+
+describe("parseContentDisposition with already-decoded input", () => {
+  it("leaves a string with characters above U+00FF alone", () => {
+    // Not a byte string, so there is nothing to re-decode -- and latin1 would truncate `哨`.
+    expect(parseContentDisposition('attachment; filename="café哨.stl"')).toBe("café哨.stl");
+  });
+});
+
+describe("modelUpload", () => {
+  it("decodes a browser's raw UTF-8 multipart filename as UTF-8, not Latin-1", async () => {
+    const app = express();
+    app.post("/upload", modelUpload.single("file"), (req, res) => {
+      res.json({ name: sanitizeFilename(req.file?.originalname) });
+    });
+    const res = await request(app)
+      .post("/upload")
+      .attach("file", Buffer.from("solid x\nendsolid x\n"), { filename: "哨子.stl", contentType: "model/stl" });
+    expect(res.status).toBe(200);
+    expect(res.body.name).toBe("哨子.stl");
+  });
+});