Re[4]: New printing behavior & console

[Arseny S. <am...@ic...>]

Hello Sam,

Tuesday, October 16, 2001, 8:01:03 AM, you wrote:

>> >> I changed win32 streams again, I believe it looks nicer now.
>> Sam> thanks - please re-submit the patch, removing the workaround for the
>> Sam> above bug.
>> I have 2 patches at the moment, 1st just makes a code look nicer (affects
>> only stream.d) and second (encoding.d, stream.d, win32.d, win32aux.d)
>> changes default terminal encoding for win32 to OEM code page so that
>> ANSI2OEM/OEM2ANSI not needed anymore (when with unicode). Unicode characters
>> (frame chars for example) may then being printed.

Sam> okay - could you please elaborate?
Sam> I just want to understand what you did.
Sam> Do I understand correctly that you are just saving the ANSI2OEM/OEM2ANSI
Sam> conversions for the terminal i/o on win32?
The point of initialization of terminal codepage to OEM:

  In the original charset translation of internal unicode character occur
  in two stages:
                   UC -->(internal en/decoding mechanism)-->
                   ANSI CHAR /Windows charset/ -->(ANSI2OEM)-->
                   OEM (screen char)

  (and back to front)

  There are some characters which cannot pass this scheme unharmed and
  can be succesfully translated to OEM charset directly through
  internal mechanism. So I changed (only in UNICODE and WIN32_NATIVE
  part) default value of *terminal-encoding* to OEM and removed any
  extra translations from window-stream and win32 lowlevel file
  routines (which checked itself if it is a console and if it is,
  translated the data to/from OEM). Patch touches only console and screen
  streams and only in win32.

  The OEM charset detection method is like iconv
  method:

  iconv:
     static char buf[2 + 10 + 1];
     sprintf (buf, "CP%u", GetACP ());
     codeset = buf;
     
  new function 'win_console_encoding':
     static char buf[2+10+1];
     sprintf(buf,"CP%u",GetOEMCP());
  
>> Diffs are not incremental (i.e. it needs only one zip for patching)
Sam> too bad.  I would rather commit the "functionality changes" and
Sam> "reformatting" separately.
Sam> Also, we need a single ChangeLog entry per commit.
Piece of cake.

Sam> Finally, the files in the zip files are weird: "encoding.151.a.dif"
Sam> should be called "encoding.d.diff" etc.
That's version numbers. I not assimilate that CVS branch stuff yet.

-- 
Best regards,
 Arseny                            mailto:am...@ic...