[Svnmerge] Unicode in log messages

[Raman Gupta]

Benson Margulies wrote:
> No, the data is destroyed by the time it's written out to the file.
> 
> Consider. There is UTF-8 Arabic in the log. svnmerge.py reads it, and
> then writes it to a file as mac-roman. It is now destroyed, since Arabic
> cannot be represented in mac-roman. I will check the config.

I'm not sure why you are disagreeing since that is fully consistent
with my last message. To reiterate, svnmerge.py *thinks* that svn
*wants* mac-roman which is why it writes to the file in mac-roman.
However, clearly in this case that is incorrect -- it appears there
may be some strangeness in your terminal and/or python environment
causing svnmerge.py to think mac-roman is the encoding svn wants.

I see in your other message log-encoding is not set. Ok.

I believe rather than trying to guess which encoding svn wants to see
in the file, your approach of forcing svnmerge.py to write the commit
log file in UTF-8, and similarly forcing svn to read the commit log
file in UTF-8 via the --encoding parameter is a good one. That will
eliminate any dependence on differences in platforms and environments.
It will also eliminate the current TODO in svnmerge.py to handle the
log-encoding config value.

Patches are welcome.

Cheers,
Raman