[frameworks/kcodecs] /: [KEncodingProber] Fix GB18030 false positive

Git commit d3498fd573978b3cf929f360969b992c3460b43d by Stefan Brüns.
Committed on 31/07/2026 at 07:01.
Pushed by bruns into branch 'master'.

[KEncodingProber] Fix GB18030 false positive

Although it is unlikely a random sequence fits a 4 byte GB18030 sequence
(both 2nd and 4th byte must be in the range 0x30 to 0x39), a match is
no guarantee for the string to be GB18030 encoded.

UTF-16 strings with e.g. Katakana often match this sequence, and even
strings with most ISO-8859-x or Windows single byte encodings can match
it.

M  +4    -4    autotests/kencodingprobertest.cpp
M  +0    -4    src/probers/nsGB2312Prober.cpp
M  +1    -1    src/probers/nsMBCSSM.h

https://invent.kde.org/frameworks/kcodecs/-/commit/d3498fd573978b3cf929f360969b992c3460b43d

diff --git a/autotests/kencodingprobertest.cpp b/autotests/kencodingprobertest.cpp
index 67027ab..5bbc008 100644
--- a/autotests/kencodingprobertest.cpp
+++ b/autotests/kencodingprobertest.cpp
@@ -122,10 +122,9 @@ void KEncodingProberTest::testProbe()
     QEXPECT_FAIL("utf-8 Japanese Universal", "Too low UTF-8 confidence, too high Win-1252", Abort);
     QEXPECT_FAIL("Konnichiwa UTF-16LE", "Too low UTF-16LE confidence, too high Win-1252", Abort);
     QEXPECT_FAIL("Konnichiwa UTF-16BE", "Too low UTF-16BE confidence, too high Win-1252", Abort);
-    QEXPECT_FAIL("EnjoyPlasma Japanese UTF-16LE Universal", "GB18030 false positive", Abort);
-    QEXPECT_FAIL("EnjoyPlasma Japanese UTF-16LE", "GB18030 false positive", Abort);
-    QEXPECT_FAIL("EnjoyPlasma Japanese UTF-16BE Universal", "GB18030 false positive", Abort);
-    QEXPECT_FAIL("EnjoyPlasma Japanese UTF-16BE", "GB18030 false positive", Abort);
+    QEXPECT_FAIL("EnjoyPlasma Japanese UTF-16LE Universal", "Too low UTF-16LE confidence, too high Win-1252", Abort);
+    QEXPECT_FAIL("EnjoyPlasma Japanese UTF-16LE", "MBCS filter broken", Abort);
+    QEXPECT_FAIL("EnjoyPlasma Japanese UTF-16BE Universal", "Too low UTF-16BE confidence, too high Win-1252", Abort);
     QCOMPARE(ep.encoding().toLower(), encoding);
 
     QEXPECT_FAIL("UTF-16BE Unicode", "UTF-16 no confidence", Abort);
@@ -134,6 +133,7 @@ void KEncodingProberTest::testProbe()
     QEXPECT_FAIL("UTF-16BE Unicode definite 2", "UTF-16 zero confidence", Abort);
     QEXPECT_FAIL("UTF-16LE Unicode definite 2", "UTF-16 zero confidence", Abort);
     QEXPECT_FAIL("utf-8 Japanese", "Too low UTF-8 confidence", Abort);
+    QEXPECT_FAIL("EnjoyPlasma Japanese UTF-16BE", "UTF-16 zero confidence", Abort);
     QCOMPARE_GE(ep.confidence(), 0.2);
 }
 
diff --git a/src/probers/nsGB2312Prober.cpp b/src/probers/nsGB2312Prober.cpp
index 2a76543..42cf479 100644
--- a/src/probers/nsGB2312Prober.cpp
+++ b/src/probers/nsGB2312Prober.cpp
@@ -31,10 +31,6 @@ nsProbingState nsGB18030Prober::HandleData(const char *aBuf, unsigned int aLen)
             mState = eNotMe;
             break;
         }
-        if (codingState == eItsMe) {
-            mState = eFoundIt;
-            break;
-        }
         if (codingState == eStart) {
             unsigned int charLen = mCodingSM->GetCurrentCharLen();
 
diff --git a/src/probers/nsMBCSSM.h b/src/probers/nsMBCSSM.h
index 9a65fc5..058aab1 100644
--- a/src/probers/nsMBCSSM.h
+++ b/src/probers/nsMBCSSM.h
@@ -238,7 +238,7 @@ static constexpr std::array<const unsigned char, 42> GB18030_st{
     eItsMe, eItsMe, eItsMe, eItsMe, eItsMe, eItsMe, eItsMe, // eItsMe
     eError, eError, eStart,      4, eError, eStart, eStart, // 3
     eError, eError, eError, eError, eError, eError,      5, // 4
-    eError, eError, eError, eItsMe, eError, eError, eError, // 5
+    eError, eError, eError, eStart, eError, eError, eError, // 5
     // clang-format on
 };