[Box Backup] Diff Optimisations now in SVN

Martin Ebourne boxbackup at fluffy.co.uk
Mon Oct 17 14:59:39 BST 2005


Jonathan Morton <chromi at chromatix.uklinux.net> wrote:
> As of revision 17, a working version of my diff algorithm 
> optimisations is in the tree (at chromi/diffopt).  It passes all 
> tests on my G5 (powerpc64-unknown-linux-gnu), except for raidfile, 
> which seems to be completely broken on Linux/PPC anyway, and is 
> unrelated to my changes.

I've done a test merge of all my changes with your changes (just in my 
working copy, not checked in). It all merged without conflicts. I 
checked the merged files and they all looked sane.

It compiles and passes all the tests on i386-linux (though not first 
time - looks like bbstored processes aren't always exiting properly, 
causing subsequent tests to fail).

So pretty promising overall.

> Note that a side-effect of my using KATE is that there are a lot of 
> deltas referring to blank lines, where KATE has deleted trailing 
> whitespace characters from the entire file being edited.  I'm not 
> sure what to do about that, aside from trying to work out the right 
> GNU indent parameters and hoping nothing else breaks.  Perhaps 
> someone had better go through all the other trees and delete trailing 
> whitespace from them too.

I think the best thing to do here is 'diff -uw' your changes and apply 
that as a patch to a new copy from trunk. Otherwise this is going to 
cause excessive and unnecessary pain. That's exactly what I did for my 
merge above, and it worked well.

As to whether to strip trailing spaces, I prefer that. I have emacs set 
to do that by default, but I turn it off when working on box. KATE must 
have an option to turn it off too (I'd consider it a bug if not). But 
either way I don't much care.

If we did strip the spaces we should do it after the merges, not 
before, else it'll just cause even more pain.

Cheers,

Martin.




More information about the Boxbackup mailing list