Re: [Monotone-devel] Re: Question for Tim - testsuite.lua giant list

[Daniel Carosone]

On Tue, Jul 04, 2006 at 12:04:07PM -0700, Graydon Hoare wrote:
>   2. Explicit lists aren't always a bad thing. Our makefile uses
>      them too. You trade one bug (executing something you didn't
>      intend) for another (failure to execute something you
>      expected but forgot to list).
> 
> I personally prefer big lists, but I won't argue the point strenuously.

Agreed, and neither will I, though there's a way to fill the gap.

What rounds this out is a way to ensure that the two forms are the
same: write a little tool (sh and etc) to generate the Big List from
the filesystem contents in the workspace. Have the generated list be
compared against the manual list - or simply compare diffs of the
generated list at commit time and use that.

The points about desiring stable ordering are also important, though
they could be addressed a number of ways, from simple sorting of file
paths through to more complex tsort-derived orderings from little
headers in the scripts.  Or the execution list could be maintained in
manual order, and sorted for order-neutral comparison against the
automatic one just to check for omissions/divergence.

Finally, you could just make this check itself be another test.

--
Dan.

pgpK3I4XusKK5.pgp
Description: PGP signature