Fix a regression introduced in 6a1852

Summary:
Bytes from 'Strings' of type GroupString and KeyString should not be
escaped because they are valid UTF-8. Only instances of ValueString
should be escaped.

This fixes the failing test KConfigTest::testEncoding

Test Plan: Ran `ninja test` and found no errors.

Reviewers: dfaure, arichardson, apol, aacid, ngraham

Reviewed By: apol

Subscribers: kde-frameworks-devel

Tags: #frameworks

Differential Revision: https://phabricator.kde.org/D17856

1 files changed, 6 insertions, 1 deletions

diff --git a/src/core/kconfigini.cpp b/src/core/kconfigini.cpp
index 39e59364..b6749731 100644
--- a/src/core/kconfigini.cpp
+++ b/src/core/kconfigini.cpp
@@ -673,7 +673,12 @@ QByteArray KConfigIniBackend::stringToPrintable(const QByteArray &aString, Strin
         switch (s[i]) {
         default:
             // The \n, \t, \r cases (all < 32) are handled below; we can ignore them here
-            if (((unsigned char)s[i]) < 32 || ((unsigned char)s[i]) >= 127) {
+            if (((unsigned char)s[i]) < 32) {
+                goto doEscape;
+            }
+            // GroupString and KeyString should be valid UTF-8, but ValueString
+            // can be a bytearray with non-UTF-8 bytes that should be escaped.
+            if (type == ValueString && ((unsigned char)s[i]) >= 127) {
                 goto doEscape;
             }
             *data++ = s[i];