[Svnmerge] Unicode in log messages

[Raman Gupta]

Benson Margulies wrote:
> It's still mac-roman after running the other python you sent.

Ok.

> If decoding with sys.stdout.... works, and that is UTF-8, then it cannot
> be true that svn itself is disgorging in the defaultlocale, since the
> Arabic would be corrupted. It really seems that svn log is delivering
> sys.stdout.encoding.

It looks like your issue is not reading the data output by svn log,
but rather when svn reads the data that svnmerge has created for the
merge commit log message.

svnmerge appears to think that the encoding of that file needs to be
mac-roman, while svn seems to expect utf-8.

Can you check in your subversion config whether you have the
log-encoding config value specified?

> So, the problem happens when this is different from the defaultlocale
> encoding. Thus my desire to be able to control/specify the file encoding.

I'd rather setup svnmerge.py to automatically select the right
encoding if possible.

Cheers,
Raman