BUG: Avoid encoding invalid XML chars in CTest

CTest encodes test and tool output in XML for dashboard submission.
This fixes the XML encoding implementation to not encode an invalid
character and instead put a human-readable tag in its place.
See issue #8647.

1 files changed, 11 insertions, 1 deletions

diff --git a/Source/cmXMLSafe.cxx b/Source/cmXMLSafe.cxx
index 717743b..eb22063 100644
--- a/Source/cmXMLSafe.cxx
+++ b/Source/cmXMLSafe.cxx
@@ -82,7 +82,17 @@ cmsys_ios::ostream& operator<<(cmsys_ios::ostream& os, cmXMLSafe const& self)
           // encoding.  Instead of escaping these bytes, we should
           // handle the current locale and its encoding.
           char buf[16];
-          sprintf(buf, "[bad-char-%hx]", static_cast<unsigned short>(c));
+          // http://www.w3.org/TR/REC-xml/#NT-Char
+          if(c >= 0x80)
+            {
+            sprintf(buf, "&#x%hx;", static_cast<unsigned short>(c));
+            }
+          else
+            {
+            // We cannot use "&#x%hx;" here because this value is not
+            // valid in XML.  Instead use a human-readable hex value.
+            sprintf(buf, "&lt;0x%hx&gt;", static_cast<unsigned short>(c));
+            }
           os << buf;
           }
         break;