[Svnmerge] KeyError on svnmerge.py

Raman Gupta rocketraman at fastmail.fm
Sat Mar 8 16:07:55 PST 2008


Giovanni Bajo wrote:
> On Sat, 2008-03-08 at 11:04 -0500, Raman Gupta wrote:
> 
>>> One question: is using LC_MESSAGES=C enough to fix the regression? In
>>> that case, I would assume that LC_MESSAGES would only affect SVN's own
>>> strings and not log messages. It would be a win-win.
>> Good idea, but svn doesn't seem to respect LC_MESSAGES for the output
>> of svn info:
>>
>> # export LC_ALL=de_DE
>> # svn info | head -1
>> Pfad: .
>>
>> # LC_MESSAGES=C svn info | head -1
>> Pfad: .
>>
>> # LC_ALL=C svn info | head -1
>> Path: .
> 
> I see. I guess this changed in later SVN versions (given that
> svnmerge.sh used to use LC_MESSAGES), but never mind.

LC_MESSAGES is respected by svn, as long as LC_ALL is *not* set. This
is apparently correct POSIX behavior:

http://www.opengroup.org/onlinepubs/009695399/basedefs/xbd_chap08.html#tag_08_02

Also, Hernan has shown that svn *does* override the Windows regional
settings with the value of LC_MESSAGES. So the Windows regional
settings are roughly equivalent to the UNIX LANG variable.

Therefore, should the attached patch solve the issue?

I don't know if this will regress svnmerge.py on Romulo's problem. The
testCommitMessageEncoding test case passes on my machine, but the test
case is heavily dependent on the machine environment in which it is
run. It would be helpful if Romulo could test this but I've been
getting bounces on his email address.

Cheers,
Raman Gupta
-------------- next part --------------
A non-text attachment was scrubbed...
Name: svnmerge_simple-localization.patch
Type: text/x-patch
Size: 689 bytes
Desc: not available
URL: </pipermail/svnmerge/attachments/20080309/e77d8a83/attachment-0001.bin>


More information about the Svnmerge mailing list