From 2a272e97c076c8a0f13a1f0f5140ad5d72cbd890 Mon Sep 17 00:00:00 2001
From: hackademix <giorgio@maone.net>
Date: Mon, 11 Mar 2019 22:35:46 +0100
Subject: Correctly decode using the original document charset and force
 re-encoding via UTF-8 BOM only when needed.

---
 bg/ResponseMetaData.js  | 18 +-----------------
 bg/ResponseProcessor.js | 20 +++++++++-----------
 2 files changed, 10 insertions(+), 28 deletions(-)

(limited to 'bg')

diff --git a/bg/ResponseMetaData.js b/bg/ResponseMetaData.js
index 41d1fe9..5a75a3c 100644
--- a/bg/ResponseMetaData.js
+++ b/bg/ResponseMetaData.js
@@ -37,7 +37,6 @@ class ResponseMetaData {
         this.headers[propertyName] = h;
       }
     }
-    this.forcedUTF8 = false;
   }
 
   get charset() {
@@ -52,21 +51,6 @@ class ResponseMetaData {
     return charset;
   }
 
-  get isUTF8() {
-    return /^utf-?8$/i.test(this.charset);
-  }
-
-  forceUTF8() {
-    if (!(this.forcedUTF8 || this.isUTF8)) {
-      let h = this.headers.contentType;
-      if (h) {
-        h.value = h.value.replace(/;\s*charset\s*=.*|$/, "; charset=utf8");
-        this.forcedUTF8 = true;
-      } // if the header doesn't exist the browser should default to UTF-8 anyway
-    }
-    return this.forcedUTF8;
-  }
-
   createDecoder() {
     if (this.charset) {
       try {
@@ -75,7 +59,7 @@ class ResponseMetaData {
         console.error(e);
       }
     }
-    return new TextDecoder("utf-8");
+    return new TextDecoder("latin1");
   }
 };
 
diff --git a/bg/ResponseProcessor.js b/bg/ResponseProcessor.js
index 4443d90..d10d46d 100644
--- a/bg/ResponseProcessor.js
+++ b/bg/ResponseProcessor.js
@@ -90,8 +90,6 @@ class ResponseTextFilter {
     };
 
     filter.onstop = async event => {
-
-      let params = {stream: true};
       // concatenate chunks
       let size = buffer.reduce((sum, chunk, n) => sum + chunk.byteLength, 0)
       let allBytes = new Uint8Array(size);
@@ -119,19 +117,19 @@ class ResponseTextFilter {
       } catch(e) {
         console.error(e);
       }
-      if (editedText !== null &&
-        (metaData.forcedUTF8 && request.type !== "script" ||
-          response.text !== editedText)) {
-        // if we changed the charset, the text or both, let's re-encode
-        filter.write(new TextEncoder().encode(editedText));
-      } else {
-        // ... otherwise pass all the raw bytes through
-        filter.write(allBytes);
+      if (editedText !== null) {
+        // we changed the content, let's re-encode
+        let encoded = new TextEncoder().encode(editedText);
+        // pre-pending the UTF-8 BOM will force the charset per HTML 5 specs
+        allBytes = new Uint8Array(encoded.byteLength + 3);
+        allBytes.set(new Uint8Array([0xEF, 0xBB, 0xBF]), 0); // UTF-8 BOM
+        allBytes.set(encoded, 3);
       }
+      filter.write(allBytes);
       filter.close();
     }
 
-    return metaData.forceUTF8() ? {responseHeaders} : ResponseProcessor.ACCEPT;;
+    return ResponseProcessor.ACCEPT;
   }
 }
 
-- 
cgit v1.2.3