[darcs-users] Delta Debugging

[Andrew Pimlott]

On Mon, May 24, 2004 at 03:59:01PM -0700, Kenneth Knowles wrote:
> On Mon, May 24, 2004 at 05:28:36PM -0400, Andrew Pimlott wrote:
> > Ok, I took a look at those papers, and from what I can tell, they aren't
> > very promising for trying to break patches into sub-patches.  The
> > problem as I said is that most combinations of sub-patches will be
> > broken, ie they will give an indeterminate result.  Take this example,
> > which I don't think is too contrived:  We have 8 sub-patches.  1, 2, 3
> > (sub-patch set a) go together, as do 1, 2, 4, 5, 7 (sub-patch set b).
> > sub-patch set b works, sub-patch set a has a bug, and the code only
> > builds if the sub-patch sets are either entirely present or absent.  I
> > don't think that any variant of their algorithm will identify patch-set
> > a as the minimal test case.  Instead, they will give the full set of
> > sub-patches.
> 
> This doesn't compute for me, because sub-patch set b will not build by
> your criteria - you must either add 3 or remove 1 and 2.

Oh dear, I meant that b is (4, 5, 6, 7, 8).  Sorry.  I mixed this up
with another example.

Andrew