Module Internationalisation.

1. Messages. Any text in the module body should be replaced by text coming from a messages file, this includes and single character strings which are referred to in the module. The only exception is Help and Syntax text which should remain in the module as well as being put in two separate messages files, the token used for the help and syntax messages should be the command itself.

   The messages file.

   The messages file should not be included in the module itself. It will be combined into a central Messages module which will provide all the messages files for all system modules. The path for the messages file for module X would normally be: Resources:$.Resources.X.Messages.

   Opening the messages file.

   The messages file should be opened on service PostInit, this service is issued immediately after all the modules have been initialised, and therefore ensures that both ResourceFS and the messages module have been initialised, and that the messages file can be found. This means that the messages file is not available during the initialisation of the module, If any text has to be printed during the initialisation of the module, you should do the following:

         Try and open the messages file.
               If you get an error message, print the text in English.
               Otherwise use the messages file to print text.
   

   At the moment opening the messages file will always fail, or open the English messages file which is in the ROM, but this enables us to replace the messages in the ROM with messages for another language, without making any further changes to the modules.

   The code to open the message file should look like this:

           ADR     r0, MessageFile_Block
           ADR     r1, message_filename
           MOV     r2, #0
           SWI     XMessageTrans_OpenFile
   

   Reading from the messages file:

   MessageTrans should be used for all access to the messages file. All text to be printed should be read from the messages file.

   Errors:

           At the moment the code for returning errors should look something
   

   like the follwoing:

                ADR        R0,ErrorBlock_AnError
                SETV
                Pull       "....,PC"
   

   To enable translation of the error text you should do the following:

   1. Change the text for the error to be the message token, and move the text to the messages file.
      2 When about to report the error:

          ADR          R0,ErrorBlock_AnError
          BL           CopyError
          Pull         "....,PC"
      

      Where CopyError is

              Push    "R1-R7,LR"
      

      ;       R0 -> Error block (Word aligned) containing error number , token, terminated by 0 or error number , "token:default"
      ;       R1 -> 4-word data structure passed to MessageTrans_LoadFile
      ;       R2 -> buffer to hold result (0 => Use internal buffer)
      ;       R3 = buffer size (if R2 non-0)
      ;       R4 -> parameter 0 (0 => don't substitute for "%0")
      ;       R5 -> parameter 1 (0 => don't substitute for "%1")
      ;       R6 -> parameter 2 (0 => don't substitute for "%2")
      ;       R7 -> parameter 3 (0 => don't substitute for "%3")
                                                                                
              ADR     R1,MessageFile_Block
              MOV     R2,#0                   ; Use internal buffer
              MOV     R4,#0                   ; Argument pointers.
              MOV     R5,#0
              MOV     R6,#0    
              MOV     R7,#0
              SWI     XMessageTrans_ErrorLookup    ; Get error text. (Sets V !)
              Pull    "R1-R7,PC"              ; error token.
      

      For errors which may be reported from IRQs you should pre_cache the translated error into a local buffer instead of translating it every time it is to be reported. In this case you should re-cache the error on Service_TerritoryStarted so that you get the new text if a new language comes into effect. Pre-caching should be done by calling MessageTrans_Lookup, and not MessageTrans_ErrorLookup.

      Sorting:

              All assumptions about the sorting order being the same as the ASCII
      

      order of characters should be removed.

            There is a new SWI provided Territory_Collate:
      

             On entry:
      

                      R1 -> String1  (0 Terminated)
                      R2 -> String2  (0 Terminated)
                      R3  Flags:
                                Bit 0 :  1 = Ignore case.
                                Bit 1 :  1 = Ignore accents.
             On exit:
                   R0   <0 if S1 < S2 
                        =0 if S1 = S2
                        >0 if S1 > S2 
                   Other registers preserved.
      

                   Z set if equal (EQ).
                   C set and Z clear if S1 > S2 (HI)
                   N set and V clear if S1 < S2 (LT)
      

                   V set if error.
                
            I.E.  flags set as for numeric compare you can use HI,LT, and EQ
      

      condition codes.

             Always use this call when sorting !
      

      Time:

            The concept of time zones has been introduced, the system 5 byte time
      

      is now concidered to be in UTC (GMT) whenever a 5 byte time is used it is assumed to be an absolute indication of the time, unaffected by changes in local time.

            All system calls which take a 5 byte time and produce text output will
      

      produce the output in local time, given a UTC 5 byte value.

            When setting the clock, a string giving the LOCAL time in the format
      

      day,dd mmm yyyy.hh:mm:ss should be given, note that this is the buffer order, so the day should be the first thing in the buffer even if the write direction is right to left.