From b3781049ac45644e999b64fd476d1db6b47b658e Mon Sep 17 00:00:00 2001
From: arc-alex <ac@omnivox.dk>
Date: Sat, 10 Jan 2026 21:13:23 +0100
Subject: [PATCH] fix: preserver utf8 charset in filenames

---
 apps/client/src/common/api/db.ts | 11 ++++++-----
 apps/server/src/utils/upload.ts  | 30 +++++++++++++++++++++++++++++-
 2 files changed, 35 insertions(+), 6 deletions(-)

diff --git a/apps/client/src/common/api/db.ts b/apps/client/src/common/api/db.ts
index 28f3b9b333..83b2da7464 100644
--- a/apps/client/src/common/api/db.ts
+++ b/apps/client/src/common/api/db.ts
@@ -32,15 +32,16 @@ export async function downloadProject(fileName: string) {
 
 /**
  * HTTP request to upload project file
+ *
+ * Note: The browser's FormData API automatically encodes filenames according to
+ * RFC 2231/RFC 5987 standards with UTF-8 charset. The issue is server-side where
+ * Busboy (used by Multer) defaults to 'latin1' charset. The server-side fix in
+ * upload.ts handles the encoding conversion.
  */
 export async function uploadProjectFile(file: File): Promise<MessageResponse> {
   const formData = new FormData();
   formData.append('project', file);
-  const response = await axios.post(`${dbPath}/upload`, formData, {
-    headers: {
-      'Content-Type': 'multipart/form-data',
-    },
-  });
+  const response = await axios.post(`${dbPath}/upload`, formData);
   return response.data;
 }
 
diff --git a/apps/server/src/utils/upload.ts b/apps/server/src/utils/upload.ts
index 323dda1b8d..2a8948a491 100644
--- a/apps/server/src/utils/upload.ts
+++ b/apps/server/src/utils/upload.ts
@@ -30,6 +30,33 @@ function generateNewFileName(filePath: string, callback: (newName: string) => vo
   checkExistence(newPath);
 }
 
+/**
+ * Fixes encoding issues where UTF-8 bytes are incorrectly interpreted as Latin-1.
+ *
+ * Multer uses Busboy internally, which defaults to 'latin1' charset for parsing
+ * Content-Disposition header parameters (like filenames). This causes UTF-8
+ * characters to be misinterpreted.
+ *
+ * Example: 'ø' (UTF-8: 0xC3 0xB8) gets misinterpreted as 'ø' (Latin-1: 0xC3 0xB8)
+ *
+ * Solution: Convert the string back to bytes using Latin-1 (which preserves
+ * byte values), then re-interpret those bytes as UTF-8.
+ *
+ * Note: This is the standard workaround since Multer doesn't expose Busboy's
+ * defParamCharset option directly.
+ * @link https://github.com/expressjs/multer/issues/1104
+ */
+function fixFilenameEncoding(filename: string): string {
+  try {
+    // Convert the string back to bytes using Latin-1 (which preserves byte values),
+    // then re-interpret those bytes as UTF-8
+    return Buffer.from(filename, 'latin1').toString('utf8');
+  } catch (error) {
+    // If conversion fails, return the original filename
+    return filename;
+  }
+}
+
 // Define multer storage object
 export const storage = multer.diskStorage({
   destination: function (_req, file, cb) {
@@ -40,7 +67,8 @@ export const storage = multer.diskStorage({
 
     ensureDirectory(publicDir.uploadsDir);
 
-    const sanitisedName = sanitize(file.originalname);
+    const fixedFilename = fixFilenameEncoding(file.originalname);
+    const sanitisedName = sanitize(fixedFilename);
     const filePath = path.join(publicDir.uploadsDir, sanitisedName);
 
     // Check if file already exists