[Freeciv-Dev] Re: generalized improvemants (PR#243)

[David Pfitzner <dwp@xxxxxxxxxxxxxx>]

Jeff Mallatt wrote:

> The problem is that a patch must be complete.  It can't leave the CVS code
> in a dysfunctional state.

Agreed.  And it should be possible for each of several steps to
still be functional (and ideally, give identical results for the 
standard rulesets, eg regression testing), though this may involve 
some extra work (including writing some transitional code which 
doesn't end up in the final product).

Actually, I'd be more confident with something like this as a
single big patch if it could pass regression tests, such as giving
identical results for all-AI games.  (Or does it?  I haven't actually 
tried, but suspect not.)  I found this very useful when doing the 
first generation ruleset patches.  (Though it also tends to uncover 
subtle bugs in the "current" codebase, which can be time consuming 
to track down.)

Regards,
-- David