Add recent changes to README.

Add a note to the manual to explain why --fullmultibyte may be an issue.
I am not completely satisfied of the position of this note, as it is symbology specific comaction, but it interferes at least with ECI described below.

README

@@ -132,7 +132,15 @@ Bugs:
 
 Version 2.7.2 Not released jet
 Changes:
-(New) Tests for auspost, codablock, composite, dotcode, general, telepen, upcean
+(New) Tests for auspost, codablock, composite, dotcode, general, telepen, upcean,
+  all output formats, 
+- QR, Han Xin, Grid Matrix: the multi byte compaction schemes (ex: Kanji)
+  are used by some decoders as codepage information (Ex: GB2312) and output
+  may be translated to UTF-8. This may destroy the data in a not controlable
+  manner.
+  In consequence, multibyte compaction schemes are disabled by default.
+  The new option --fullmultibyte (option_3 = ZINT_FULL_MULTIBYTE) enables this
+  optimisation.
 Bugs:
 - Ticket 181 penetration test found many bugs:
   - Auspost: null bytes in content caused segfault
@@ -144,12 +152,13 @@ Bugs:
   - EANUCC: buffer overflow on multiple + (multiple extension bars)
   - Maxicode: index overrun on numeric compaction
   - CodeOne: Simple i indexing not sp + i in C1_ASCII numeric latch loop.
-  - Aztec: free memory, 
+  - Aztec: free memory
 - Ticket 183: Databar stacked separator correction
 - Ticket 182: Output bitmap type was char, where some targets assigned 0
   if pixel colour >127 (e.g. negative). API change to use unsigned char.
 - HanXin: wrong codepage, gb2312 instead gb18030.
 - PDF417: corrected alloced sizes to max characters
+- Ticket 189: FNC1 encodation fixed (FLG(0) missing after FLG(n))
 
 CONTACT US
 ----------

docs/manual.txt

@@ -428,15 +428,26 @@ HIBC data may also be encoded in the symbologies Code 39, Code128, Codablock-F,
 Data Matrix, QR Code, PDF417 and Aztec Code. Within this mode, the leading '+'
 and the check character are automatically added.
 
-The --binary option prevents Zint from performing any convertion of the data
+The --binary option encodes the input data as given. Automatic code page
-before placing in the barcode symbol and should be used if you are encoding raw
+translations to ECI pages is disabled. This may be used for raw binary
-binary or encrypted data. For an example involving QR Code and UTF-8 encoding,
+or binary encrypted data.
-see Ex3 below.
+This switch plays together with the build-in ECI logic and examples may
+be found in that section.
 
 The --fullmultibyte option uses the multibyte modes of QR Code, Micro QR Code,
 Rectangular Micro QR Code, Han Xin Code and Grid Matrix for binary and Latin
-data, maximizing density. You should check your barcode reader supports this use
+data, maximizing density.
-before enabling it.
+Please check your reader device for not having the following
+non-standard property: this option may switch to multibyte (ex. Kanji)
+compression within the code to obtain a smaller code matrix. The Kanji mode is
+a compaction mode and does not inform about contained Kanji characters.
+Nevertheless, some decoders take blocks which are encoded in Kanji
+compaction and output them as Kanji characters, typically be applying a
+transformation to UTF8. The result is unpredictable, as zint
+--fullmultibyte chooses the compaction type only on the criteria to
+obtain smaller codes and not to inform about the presence of Kanji
+characters.
+An example decoder with this property is the zxing decoder app.
 
 If your data contains non ISO-Latin-1 characters, you may encode it using an
 ECI-aware symbology and an ECI value from the table below.