Windows UTF-8 printed with chcp 65001 - characters

Question

Here is one thing I can't get my head around:

I am using Windows 7 and Strawberry Perl 5.20, and I want to write UTF-8 to the console (cmd.exe) with chcp 65001.

The UTF-8 characters themselves are coming out fine, even >255, but there is a mysterious duplication of some caracters (this only happens if I don't redirect into a file)

EDIT: I now have seen another post that had essentially the same problem at last-octet-repeated-when-my-perl-program-outputs-a-utf-8 -- the solution is to inject a binmode(STDOUT, 'unix:encoding(utf8):crlf') into the perl program -- I have tested and it works fine now

Thanks to everybody who looked into this weird problem.

In a nutshell, I am writing a UTF-8 string (chr(300) x 3).chr(301)."UVW\x{0D}\x{0A}", when I redirect into a flat file and then print the flat file, everything is fine.

However, when I print directly to the console, some characters are mysteriously duplicated (I am talking about the characters "VW" in the seperate line), and I don't know why

Here is my test-output

Page de codes active : 65001

Redirected into a file:
-----------------------
ĬĬĬĭUVW

Printed directly:
-----------------
ĬĬĬĭUVW
VW

IO-Layers = (unix crlf)

C4ACC4ACC4ACC4AD5556570D0A

Here is my test program:

@echo off
chcp 65001
echo.

set H1=BEGIN{binmode(*STDIN); undef $/;
set HEXDUMP="%H1% print uc(unpack('H*',<STDIN>)), qq{\n}}"

set L1=my @l = PerlIO::get_layers(*STDOUT, output, 1);
set LAYERS="%L1% print {*STDERR} qq{IO-Layers = (@l)\n};"

set PROG="print chr(300) x 3, chr(301), qq{UVW\n};";

set TFILE=%TEMP%\tfile.txt

echo Redirected into a file:
echo -----------------------
perl -C6 -e%PROG% >%TFILE% && type %TFILE%
echo.

echo Printed directly:
echo -----------------
perl -C6 -e%PROG%

echo.
perl -e%LAYERS%
echo.

perl -e%HEXDUMP% <%TFILE%

echo.
pause

As I said, the characters themselves are printed correctly, but why is there this mysterious duplication ? ...and why * only * if not redirected into a file ?

Answer 1

As I suspected, this has been reported as a failure in Windows software:

This is caused by a bug in Windows. When writing to a console set to code page 65001, WriteFile() returns the number of characters written instead of the number of bytes.

I wasn't aware of a work-around, but if the :unix:encoding(utf8):crlf PerlIO stack works for you then it seems you have found one.