Re: [gnugo-devel] breakin and readconnect fixes

[Paul Pogonyshev]

Arend wrote:
> > > Btw, Paul, I tried to get on good terms with your change of the connection
> > > queue logic (only storing the origin instead of each stone of a string)
> > > and failed badly. So I hope you won't be too upset in case I revert that
> > > change soon...
> 
> I still don't like it very much (as it transforms each loop over every
> intersection in the connection queue into a double loop) but...

Yet I still don't understand why you would need to loop over all stones in
the first place.  Any string is connected so strongly that there is no good
reason to distinguish between its stones.

In any case, you can use the convenience function expand_connection_queue().

> [...]
> 
> So instead of reverting your change, here is a patch with 3 fixes, two
> of them resolving issues caused by your patch:
> 
> 1. Ladder attack against "secondary target" in break-in reading only
> makes sense in block_off reading (unrelated).
> 
> 2. add_to_start_queue must be called with string origin.
> 
> 3. Allow for strings to be partially in the break-in goal.

Btw, were there any regression changes?

Paul