On Wed, Jan 15, 2003 at 11:14:42AM +0000, Andrew Cheadle wrote: > Hi Chris, > > You are indeed correct, C=-1 is indeed a valid value in the domain and should > not have been pruned. This is a bug in the fd solver. [...] > Our development priority is on ic, so I'm afraid this bug will probably be > low priority. Indeed. Problems with multiplying negative numbers in FD were identified some time ago. Unfortunately then FD code is terribly complicated and mostly undocumented, which is one of the reasons we're replacing it with IC. Last time I looked, the problem seemed difficult to fix, with a high chance of introducing new bugs. Given that we're planning to throw the FD library away, fixing it was deemed to be a waste of our limited resources. Your two main options are to either adjust your model to avoid negative domains (or var-var multiplication, though that's harder ;) or to switch to IC. Please note that if you want to try IC, you are *strongly* recommended to upgrade to the latest ECLiPSe (5.5). Cheers, WarwickReceived on Wed Jan 15 11:27:28 2003
This archive was generated by hypermail 2.1.8 : Wed 16 Nov 2005 06:07:21 PM GMT GMT