Bug 1614796 - Body search: Strip soft line breaks in QP parts when assembling HTML body. r=benc

2020-02-23 08:59:56 +01:00 · 2020-02-23 08:59:56 +01:00 · 8ca5007982
--- a/mailnews/base/search/src/nsMsgBodyHandler.cpp
+++ b/mailnews/base/search/src/nsMsgBodyHandler.cpp
@ -274,7 +274,7 @@ int32_t nsMsgBodyHandler::ApplyTransformations(const nsCString &line,
        eatThisLine = true;
      } else {
        // It is wrong to call ApplyTransformations() here since this will
-        // lead to the buffer being doubled-up at |buf.Append(line.get());|
+        // lead to the buffer being doubled-up at |buf.Append(line);|
        // below. ApplyTransformations(buf, buf.Length(), eatThisLine, buf);
        // Avoid spurious failures
        eatThisLine = false;
@ -313,9 +313,18 @@ int32_t nsMsgBodyHandler::ApplyTransformations(const nsCString &line,

  // Accumulate base64 parts and HTML parts for later decoding or tag stripping.
  if (m_base64part || m_partIsHtml) {
-    if (m_partIsHtml && !m_base64part)  // Replace newline in HTML with a space.
-      buf.Append(' ');
-    buf.Append(line.get());
+    if (m_partIsHtml && !m_base64part) {
+      size_t bufLength = buf.Length();
+      if (!m_partIsQP || bufLength == 0 ||
+          !StringEndsWith(buf, NS_LITERAL_CSTRING("="))) {
+        // Replace newline in HTML with a space.
+        buf.Append(' ');
+      } else {
+        // Strip the soft line break.
+        buf.SetLength(bufLength - 1);
+      }
+    }
+    buf.Append(line);
    eatThisLine = true;
    return buf.Length();
  }
--- a/mailnews/base/search/src/nsMsgSearchTerm.cpp
+++ b/mailnews/base/search/src/nsMsgSearchTerm.cpp
@ -859,7 +859,7 @@ nsresult nsMsgSearchTerm::MatchBody(nsIMsgSearchScopeTerm *scope,
      compare.Append(buf);
      // If this line ends with a soft line break, loop around
      // and get the next line before looking for the search string.
-      // This assumes the message can't end on a QP soft-line break.
+      // This assumes the message can't end on a QP soft line break.
      // That seems like a pretty safe assumption.
      if (softLineBreak) continue;
      if (!compare.IsEmpty()) {
--- a/mailnews/base/test/unit/test_searchBody.js
+++ b/mailnews/base/test/unit/test_searchBody.js
@ -132,6 +132,9 @@ var Tests = [
  // Message 22 contains Καλημέρα (good morning in Greek).
  { value: "Καλημέρα", op: Contains, count: 1 },

+  // Messages 21, 23 and 24 contain "softbreak" broken by a soft line break.
+  { value: "softbreak", op: Contains, count: 3 },
+
  // Messages 16, 17, 18, 20 contain "hähä" in the plaintext part.
  { value: "hähä", op: Contains, count: 4 },

--- a/mailnews/test/data/21-plaintext.eml
+++ b/mailnews/test/data/21-plaintext.eml
@ -11,4 +11,6 @@ Content-Transfer-Encoding: quoted-printable
 Content-Language: en-GB

 Search for h=C3=B6h=C3=B6
+Test that we ignore a soft=
+break correctly.

--- a/mailnews/test/data/23-HTML.eml
+++ b/mailnews/test/data/23-HTML.eml
@ -10,5 +10,9 @@ Content-Type: text/html; charset="utf-8"; format=flowed
 Content-Transfer-Encoding: quoted-printable
 Content-Language: en-GB

-<body>Search for <b>h=C3=B6h=C3=B6</b></body>
+<body>Search for <b>h=C3=B6h=C3=B6</b>
+Test that we assemble HTML bodies correctly when there is a
+<a href=3D"https://www.example.com">soft=
+break</a> involved.
+</body>

--- a/mailnews/test/data/24-HTML+attachment.eml
+++ b/mailnews/test/data/24-HTML+attachment.eml
@ -15,7 +15,10 @@ This is a multi-part message in MIME format.
 Content-Type: text/html; charset="utf-8"; format=flowed
 Content-Transfer-Encoding: quoted-printable

-<body>Search for <b>h=C3=B6h=C3=B6</b></body>
+<body>Search for <b>h=C3=B6h=C3=B6</b>
+Test that we ignore a soft=
+break correctly.
+</body>

 --------------BC006DD22051247571F398E0
 Content-Type: image/png;