| Subject: | SVN-Notifiy loops forever if it encounters a non UTF-8 char in the pipe from svnlook |
Hi, first thanks for the great tool, it's really awesome!
The problem I encountered is that svnnotify goes into an endless loop
spitting out:
utf8 "\xFC" does not map to Unicode at
/usr/local/lib/perl5/site_perl/5.8.6/SVN/Notify/HTML/ColorDiff.pm line
117, <PIPE> line 70.
We run svnnotify with --charset utf-8. The reason for this is, that we
have text files with different encodings in our repository, most of them
are utf-8 encoded (xml, templates...) but some (java sources, properties
files) are iso8859-1 encoded. Now in this case there was a commit that
included a java file with a german umlaut character in it (which maps to
\xFC in ISO8859-1 and is not parseable as utf-8), so the error is
understandable and correct. What is really annoying is that it keeps
svnnotifiy from terminating and causes the commit to hang if it's run
from the post-commit hook.
For now I don't really care whether the umlaut is shown in the resulting
mail correctly or not, so just mapping any unreadable char from the
svnlook output to something like a '?' char would be fine.
I am currently trying to pipe svnlook output through some custom script
that 'sanitizes' the svnlook output to valid utf-8, but it would be nice
if svnnotify would handle this gracefully!
Cheers,
Tobias