Re: [Monotone-devel] Re: Hash collisions resiliency

[Nathaniel Smith]

On Fri, Apr 15, 2005 at 10:30:02AM -0400, Frank Ch. Eigler wrote:
> Then perhaps the test suite could use a few additions, such as
> explicit byte-wise comparisons of expected vs. actual stored files.
> Instead of asserting that monotone return codes are zero, actually
> archive and compare large bodies of text before/after version control
> operations.  That is the primary observable from a user's point of
> view anyway.

Hmm, so you're suggesting having a test that, say, imports a dozen
files, and then reads them out again to make sure that nothing
untoward has happened?  I guess this would be useful for
double-checking that we don't accidentaly replace our SHA1
implementation with bzero() or the like :-).

(To be clear: I actually think this would be a somewhat useful test,
though I'm dubious we could make that heinous an error without causing
the rest of the testsuite to fall over anyway; but it's not obvious to
me whether that's actually what you meant?)

-- Nathaniel

-- 
"Of course, the entire effort is to put oneself
 Outside the ordinary range
 Of what are called statistics."
  -- Stephan Spender