fix: enforce !-free cidPrefix invariant at both concatenation sites

The cid: namespacing scheme (cid:<prefix>!<id>) is only unambiguous
because the prefix half is guaranteed free of '!': the first '!' in the
result is always the separator, so an attacker-controlled Content-ID
containing '!' only extends the id half rather than colliding with a
different prefix. That invariant previously existed only as a comment.

Add Q_ASSERT_X at both independent call sites that perform this
concatenation (CidSchemeHandler::namespacedKey and
HtmlBuilder::namespaceCids) so a future prefix generator that violates it
traps in debug builds, per Task 5's precedent of not letting one unit's
correctness depend silently on another's future behaviour. Since
Q_ASSERT compiles out in release, pin the property that actually matters
release builds too test: distinct (prefix, id) pairs across a documented
"m<index>" prefix set and hostile Content-IDs (containing '!', percent-
encoded '!', empty, leading/trailing '!') never collide, and the key
always splits at its first '!' back to the exact original prefix.

3 files changed, 42 insertions, 1 deletions

diff --git a/src/cidschemehandler.h b/src/cidschemehandler.h
index 56e6ea1..1231fc6 100644
--- a/src/cidschemehandler.h
+++ b/src/cidschemehandler.h
@@ -20,8 +20,27 @@ public:
     void setParts(const QHash<QString, InlinePart> &parts) { m_parts = parts; }
 
     /// Builds the namespaced key HtmlBuilder's rewritten URLs will request.
+    ///
+    /// The '!' separator disambiguates a hostile Content-ID from the prefix
+    /// only because prefix is guaranteed '!'-free: the FIRST '!' in the
+    /// result is always the separator, so an attacker-controlled contentId
+    /// containing '!' (even several) cannot make one message's key collide
+    /// with another's, it only extends the id half after that first '!'.
+    /// This is asserted here rather than merely documented, since two call
+    /// sites (this one and HtmlBuilder::namespaceCids) perform the same
+    /// concatenation independently and neither should trust the other to
+    /// have checked it. Q_ASSERT is compiled out in release builds; the
+    /// property that matters there (distinct pairs never collide, and the
+    /// key always splits at its first '!' back to the original prefix) is
+    /// pinned by a test instead, since it holds unconditionally regardless
+    /// of whether this assertion fires.
     static QString namespacedKey(const QString &prefix, const QString &contentId)
-    { return prefix + QLatin1Char('!') + contentId; }
+    {
+        Q_ASSERT_X(!prefix.contains(QLatin1Char('!')), "CidSchemeHandler::namespacedKey",
+                   "cidPrefix must never contain '!': it is the separator, and a "
+                   "prefix containing one would make the split ambiguous");
+        return prefix + QLatin1Char('!') + contentId;
+    }
 
     void requestStarted(QWebEngineUrlRequestJob *job) override;
 
diff --git a/src/htmlbuilder.cpp b/src/htmlbuilder.cpp
index 1830383..89501ea 100644
--- a/src/htmlbuilder.cpp
+++ b/src/htmlbuilder.cpp
@@ -56,6 +56,19 @@ QString HtmlBuilder::namespaceCids(const QString &html, const QString &prefix)
     if (prefix.isEmpty())
         return html;
 
+    // The '!' separator is only unambiguous if prefix itself never contains
+    // one: the FIRST '!' in "cid:<prefix>!<id>" is always taken as the
+    // separator, so a hostile Content-ID containing '!' only extends the id
+    // half, never collides with a different prefix. This concatenation is
+    // performed independently in two places (here and
+    // CidSchemeHandler::namespacedKey); neither trusts the other to have
+    // checked it, so both assert it directly. Q_ASSERT is compiled out in
+    // release builds — the property that matters there is pinned by
+    // TestHtmlBuilder::namespacedKeyRejectsPrefixContainingSeparator instead.
+    Q_ASSERT_X(!prefix.contains(QLatin1Char('!')), "HtmlBuilder::namespaceCids",
+               "cidPrefix must never contain '!': it is the separator, and a "
+               "prefix containing one would make the split ambiguous");
+
     // This runs on the sender's raw, unescaped HTML markup (not on text that
     // has been through toHtmlEscaped()), so no double-escaping happens here;
     // it is purely a URL rewrite over the existing markup.
diff --git a/src/htmlbuilder.h b/src/htmlbuilder.h
index ac24e28..ff62c69 100644
--- a/src/htmlbuilder.h
+++ b/src/htmlbuilder.h
@@ -16,6 +16,15 @@ struct ThreadRenderItem
     /// use the same Content-ID (cid:logo@example.org), which would collide in
     /// a single document, so every reference is rewritten to
     /// cid:<prefix>!<id>.
+    ///
+    /// Requirement on whatever generates this value: it must never contain
+    /// '!'. The separator that makes cid:<prefix>!<id> unambiguous is the
+    /// FIRST '!' in the namespaced string; that only holds if the prefix
+    /// half is guaranteed free of the character, since the id half is
+    /// attacker-controlled and may legitimately contain '!' itself. The
+    /// documented "m<index>" form (e.g. "m0", "m1") satisfies this. Enforced
+    /// with Q_ASSERT at both places that perform this concatenation
+    /// (HtmlBuilder::namespaceCids and CidSchemeHandler::namespacedKey).
     QString cidPrefix;
 };