are known diff failures complex?

[Neels J Hofmeyr <ne...@elego.de>]

Hi all,

I'm planning to use the diff framework to determine whether two directories
are identical.

Since I've now seen a case where diff actually fails (and needs to be fixed
if I'm going to use it), I fear that many more will come up and that I
shouldn't endeavour fixing diff to become precise enough to depend on it.

Does anyone reading this know of a diff error that is inherently hard to fix
within a short timespan?

Thanks
~Neels

-- 
Neels Hofmeyr -- elego Software Solutions GmbH
Gustav-Meyer-Allee 25 / Gebäude 12, 13355 Berlin, Germany
phone: +49 30 23458696  mobile: +49 177 2345869  fax: +49 30 23458695
http://www.elegosoft.com | Geschäftsführer: Olaf Wagner | Sitz: Berlin
Handelsreg: Amtsgericht Charlottenburg HRB 77719 | USt-IdNr: DE163214194

Re: are known diff failures complex?

[Julian Foad <ju...@btopenworld.com>]

On Thu, 2008-10-09 at 01:50 +0200, Neels J Hofmeyr wrote:
> Hi all,
> 
> I'm planning to use the diff framework to determine whether two directories
> are identical.
> 
> Since I've now seen a case where diff actually fails (and needs to be fixed
> if I'm going to use it), I fear that many more will come up and that I
> shouldn't endeavour fixing diff to become precise enough to depend on it.
> 
> Does anyone reading this know of a diff error that is inherently hard to fix
> within a short timespan?

I can't answer this really, but I have just replied to your message from
yesterday on this topic. Part of my reply says: let's just start using
the diff framework, and then fix it.

The same reply also indicates a starting point for reducing complexity
(in the normal-versus-summarizing dimension) which should help to reduce
bugs if it is feasible.

The most difficult problems I can think of in the diff code are
extensibility and generality - such as making a diff between two
separate local paths ("svn diff --old=wcpath1 --old=wcpath2") work. I'm
not aware that there are any bugs that are really difficult to fix,
though of course there might be.

- Julian



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org