Hi, On Sun, Jan 27, 2008 at 03:05:32PM -0800, Cameron Dale wrote: > I think using --diff-cmd is better, as Otavio suggests. The best way
okay, so there are two opinions for this solution. So I will polish my patch and provide it to the devscripts maintainers when it is ready. > would be to specify to svn to use the internal diff on the > command-line, something like "svn diff --diff-cmd internal", thereby > overriding any config settings. This doesn't seem possible though, as > I just asked on the #svn chat on IRC and no one was able to help, so > using an external diff command might be the only way. Well, I don't think they see a sense for it, because --diff-cmd is normally used to configure something *different* from the internal diff code. > Have you checked the effects of a user setting git's color.diff option > to 'always' in their .gitconfig file? I suspect the same error would > occur, but I don't have a git repository with a package in it to > check. No, I have not checked that. But the problem is not limited to subversion, because it is not related to the VCS at all. The problem is quiet simple: The color codes that are used are not visible to you (well, they are as the colorized output), but it is visible to perl that tries to parse the lines with regular expressions. Therefore the line with the changes isn't '+ foobar', but 'x+ foobar' where x is to be replaced by a color code (which is more complicated and not so easy to filter as x would be :). Thats related to why I think stripping the sequences from the diff would be the better way of solving this if it is possible: it is a more common solution to *one* problem; i.e. it is not needed to add further diversity to the solution just to address the six or seven different VCSes. Despite that fact, my opinion is that a clever solution would be to strip *all* special sequences from the output, which would also solve the problem for eventual upcoming diff replacements to do blinky things. Best Regards, Patrick -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
