[svk-users] Problem with umlauts in commit messages

Marko Kaening mk362 at mch.osram.de
Thu Jun 26 09:28:01 EDT 2008


Hi,

I ran into another problem with umlauts... :(

I am using svk 2.0.2 binary install on WXP.

I had a directory called test in my local repo at //local/test containing 
a single text file called test.txt. Checking it out, manipulating the text 
with notepad and subsequent try to commit the changes lead to a problem 
with the commit message encoding:
---

C:\temp\svk>svk co //local/test test
Syncing //local/test(/local/test) in 
C:\temp\svk\test to 333.
A   test\test.txt

C:\temp\svk>cd test

C:\temp\svk\test>notepad test.txt

C:\temp\svk\test>svk stat
M   test.txt

C:\temp\svk\test>svk commit
Waiting for editor...
Commit message saved in svk-commit760lN.tmp.
Can't decode commit message as utf8.
try --encoding.

C:\temp\svk\test>type svk-commit760lN.tmp
umlaut 'ae' in commit message: 'õ'
---




OK, so I tried the same on the console using option "-m", but see, even 
here it does not work:
---
C:\temp\svk\test>svk commit -m "umlaut 'ae' in commit message using option -m: ä"
Can't decode commit message as utf8.
try --encoding.
C:\temp\svk\test>svk --encoding
Unknown option: encoding
---
What does this hint to use --encoding mean? As one can see there is no 
option like that available.



Cygwin's hexdump gave me for notepad's file content this:
---
$ hexdump.exe -C svk-commit760lN.tmp
00000000  75 6d 6c 61 75 74 20 27  61 65 27 20 69 6e 20 63  |umlaut 'ae' in c|
00000010  6f 6d 6d 69 74 20 6d 65  73 73 61 67 65 3a 20 27  |ommit message: '|
00000020  e4 27                                             |.'|
---
So, the umlaut is coded as single byte 0xE4. Console codepage ist cp850.


Is there no sane way to commit with umlauts? (I know that the repo is kept 
in utf-8, even on Windows...)

This all is related to my other thread called

"[svk-users] Non-existent charset transcoding of SVK not in-line with SVN"

but since nobody responded yet I conclude that this never happened to 
anyone. Developers seem to limit themselves in choice of path names due to 
possible problems of their tools with weird transcoding problems, as it 
got pointed out also on Mercurials mailing list, recently.

Anyway, since I do use SVN for keeping all my files versioned I hoped that 
SVK would behave like SVN concerning transcoding, but as I see now it 
globally does not.

Hope for contradicting responses! ;)

Regards,
Marko


More information about the svk-users mailing list