Gavin 'Beau' Baumanis wrote:
> None the less, the gui - dif tool in particular normally shows up
> the difference by displaying eol / tab characters as being the
> difference on a certain line.
> And normally it is a line difference, Ie I end up with a diff
> showing 100's of differences.
>
> Oddly from what I normally see, the diff is showing the entire file
> as simply being 1 complete difference - which when in this
> particular the case the file is 1200 lines, makes it a little difficult
> to pick out the coding differences.
Not sure I totally understand. It sounds like one diff program is showing hundreds of lines worth of differences, with each diff being a whitespace difference. The other diff program--the svn built-in, I guess--shows the entire file as being one big diff. Is that correct?
If so, then I think the two diff programs are displaying the differences at different levels of granularity: your gui-diff tool displays the differences at the level of individual characters, while the svn diff simply says "this line is different in the two versions."
If that's what's going on, then I expect that the svn diff is having trouble re-synching. That is, the numerous different lines are preventing it from recognizing that two identical lines (assuming there are identical lines, ignoring whitespace) are corresponding (as opposed, say, to being re-ordered with respect to some other sets of lines). You've run into the limitation in the algorithm (or more likely, the number of differences the implementation tolerates before it gives up; the algorithm could in principle wait millions of lines to re-synch, but that would take millions of seconds :-)). Have you tried the -d parameter (if your diff accepts this) to see if it will find a more minimal diff? (Of course, this hypothesis doesn't explain why the -w option doesn't give you a minimal diff.)
Mike Maxwell
CASL/ U MD
Received on 2008-09-07 00:48:54 CEST