[graphics/okular/release/26.08] core: fix(core): use atomic for FontExtractionThread::mGoOn to prevent data race

Git commit fbaec8fa4aea3becf033ddc6c2fe40f202ccef1c by Albert Astals Cid, on behalf of zhang shoucheng.
Committed on 19/07/2026 at 00:12.
Pushed by aacid into branch 'release/26.08'.

fix(core): use atomic for FontExtractionThread::mGoOn to prevent data race

mGoOn is written by the main thread in stopExtraction() and read
by the worker thread in run(). A plain bool across threads without
synchronization is a data race and undefined behavior per the C++
memory model.

Change mGoOn from bool to std::atomic<bool>. This establishes a
proper happens-before relationship between the stop signal and the
worker loop, and prevents the compiler from caching the value in a
register or hoisting the read out of the loop.

(cherry picked from commit 63342a4c2d90b66ff8e4b6a7a0eda4a9f84f298b)

M  +3    -1    core/generator_p.h

https://invent.kde.org/graphics/okular/-/commit/fbaec8fa4aea3becf033ddc6c2fe40f202ccef1c

diff --git a/core/generator_p.h b/core/generator_p.h
index 4b6ddf599..ba7a64c63 100644
--- a/core/generator_p.h
+++ b/core/generator_p.h
@@ -17,6 +17,8 @@
 #include <QSet>
 #include <QThread>
 
+#include <atomic>
+
 class QEventLoop;
 
 #include "generator.h"
@@ -180,7 +182,7 @@ protected:
 private:
     Generator *mGenerator;
     int mNumOfPages;
-    bool mGoOn;
+    std::atomic<bool> mGoOn;
 };
 
 }