Re: [rdiff-backup-users] Native Windows Port

[Josh Nisly]

Patrick Nagel wrote:
> Hi Josh,
>
> Josh Nisly (Friday, 2008-04-04):
>  
> > I'm attempting to make a native Windows port of rdiff-backup. I've
> > succeeded so far in building librsync and rdiff-backup's cmodules, as
> > well as fixing rdiff-backup's fs_abilities checks. What I'm running up
> > against now is that there seems to be a lot of assumptions that the
> > directory separator is "/", which obviously doesn't work on windows. I
> > think that the correct solution is to memorize in either the Connection
> > or RPath objects what the correct separator is, and use that everywhere.
> > Does this seem like the correct solution? Has anybody else tried porting
> > to Windows recently? There's an old page on the wiki with some
> > high-level notes, but I couldn't find anything else.
> >     
>
> Did you have a look at 
> http://www.python.org/doc/2.3.5/lib/module-os.path.html ? Especially 
> 'normpath(path)'.
>   
Yes, I've seen these. You have to be careful, though, because you can't 
use the client's normpath for remote paths, and vice versa.
> In my opinion it would be best to have something like 'native 
> directory separators' that are just a 'translation' of the internally 
> used separators (UNIX style), like it's done in Qt 
> (http://doc.trolltech.com/4.3/qdir.html). So only each ported 
> rdiff-backup instance must know what it's running on, and nothing 
> would change in the communication.
>   
I think this makes sense. It at least makes the port easier, because all 
the patch knowledge can be contained inside of RPath, and everyone 
outside of that class can just use "/".
> You'll also have to deal with that strange 'drive' foo (mount point 
> designations, e.g. C:) - see 'splitdrive(path)' - that probably 
> entails modifications in also the non-ported rdiff-backup version.
>   
That's possible. I'll look into it.

Josh Nisly