"Zigbee" encoding is not a character encoding or an integer encoding.

It's a rule for interpretation of the length field of counted-string and
counted-octet-string fields.  This means it's 1) not a general rule for
interpreting integers and 2) not a character encoding, as it also
applies to octet strings and, even for character strings, it's
*orthogonal* to the character encoding.

Therefore, it should *not* be one of the character encoding values; it
should be a bit flag.

Make it so.  This means that

1) a character encoding can be specified for Zigbee Cluster Library
strings (they appear to have multiple character encodings possible);

2) the test of it that tested it as if it were a flag will no longer get
confused by character encodings that set one or more of the bits in the
old encoding value;

3) you don't have to special-case the encoding value passed to
get_uint_value().

Put in a comment emphasizing that values that aren't character encodings
should *not* be placed in the set of character encodings.

Change-Id: I8f50aaee8ca60b0781044287e9b38111de38c81f
Reviewed-on: https://code.wireshark.org/review/33341
Petri-Dish: Guy Harris <guy@alum.mit.edu>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <guy@alum.mit.edu>

1 files changed, 14 insertions, 2 deletions

diff --git a/epan/proto.h b/epan/proto.h
index 60a82de6ae..5974df9b32 100644
--- a/epan/proto.h
+++ b/epan/proto.h
@@ -415,6 +415,16 @@ void proto_report_dissector_bug(const char *format, ...)
 #define ENC_TIME_MSEC_NTP      0x00000022
 
 /*
+ * This is a modifier for FT_UINT_STRING and FT_UINT_BYTES values;
+ * it indicates that the length field should be interpreted as per
+ * sections 2.5.2.11 Octet String through 2.5.2.14 Long Character
+ * String of the ZigBee Cluster Library Specification, where if all
+ * bits are set in the length field, the string has an invalid value,
+ * and the number of octets in the value is 0.
+ */
+#define ENC_ZIGBEE               0x40000000
+
+/*
  * Historically, the only place the representation mattered for strings
  * was with FT_UINT_STRINGs, where we had FALSE for the string length
  * being big-endian and TRUE for it being little-endian.
@@ -440,8 +450,10 @@ void proto_report_dissector_bug(const char *format, ...)
  * For display, perhaps we should also map control characters to the
  * Unicode glyphs showing the name of the control character in small
  * caps, diagonally.  (Unfortunately, those only exist for C0, not C1.)
+ *
+ * *DO NOT* add anything to this set that is not a character encoding!
  */
-#define ENC_CHARENCODING_MASK    0x7FFFFFFE  /* mask out byte-order bits */
+#define ENC_CHARENCODING_MASK    0x3FFFFFFE  /* mask out byte-order bits and Zigbee bits */
 #define ENC_ASCII                0x00000000
 #define ENC_UTF_8                0x00000002
 #define ENC_UTF_16               0x00000004
@@ -471,7 +483,7 @@ void proto_report_dissector_bug(const char *format, ...)
 #define ENC_ASCII_7BITS          0x00000034
 #define ENC_T61                  0x00000036
 #define ENC_EBCDIC_CP037         0x00000038
-#define ENC_ZIGBEE               0x0000003A
+/* reserved                      0x0000003A */
 #define ENC_WINDOWS_1251         0x0000003C
 #define ENC_CP855                0x0000003E
 #define ENC_CP866                0x00000040