Shandalar - Program Bugs

[Korath]

One bug per bug, please. Most of this is a duplicate of #531.

[Korath]

commit 6f2029cca23e8dd5c491c2bfe5f3bfdc53e73370
Author: Korath <dgk@Dirge.none>
Date: Tue Oct 20 21:41:04 2015 -0400

[TH] FIX #757: T-to-control, optional-untap cards use AI assessment to untap

Rather than just untapping if the opponent controls a creature with a higher
power than the currently-controlled creature (as in the MicroProse version and
previous Shandalar ones) or never untapping if currently controlling anything
(as in Manalink), go through an abbreviated AI assessment similar to, but even
further abbreviated from, what's done in the damage prevention step.

Algorithm:
1. Backup the current game state (via backup_data_for_ai_frontend_0())
2. Untap all controller's cards except the optionally-untapping one
3. Check state-based effects (via recalculate_all_cards_in_play())
4. Run the AI evaluation function to get a baseline rating; add 12 to it to
give a slight bias towards not untapping.
5. For each valid target,
a. Restore the game state
b. Untap all controller's cards (including the optional one)
c. Check state-based effects
d. Tap the optional card
e. Reap dying creatures (via regenerate_or_graveyard_triggers())
f. Check state-based effects again
g. Gain control of the target
h. Run the AI evaluation function
i. Stop if the result is greater than the baseline rating.

This is abbreviated compared to the damage-prevention AI assessment in that it
only runs once for each target, and so might not get the best result in the
presence of auxiliary triggers (e.g., Merieke Ri Berit destoying a previously-
controlled creature when it untaps in step 5b, and that creature having a when-
this-dies trigger requiring targeting). The damage-prevention AI, in turn, is
abbreviated compared to the general AI in that each run is stopped at the end
of the step, rather than running through the rest of the turn.

Since it doesn't run until end of turn, it won't see the adjustments some cards
make manually to ai_modifier (as is done extensively to simulate the next
turn's upkeep, and many other similar situations), nor effects from triggers
happening later in this turn. So it will still, for example, untap with the
idea of controlling a different creature that happens to be enchanted by
Unstable Mutation, without being aware that that creature will have another
-1/-1 counter by the time it can be controlled; and then cheerfully tap to
re-control the creature it just released.

Still, it's much closer to the assessment used when activating for real, and
much better than the Manalink behavior of retaining control of a Tundra Wolves
and ignoring the new Serra Angel the human plays afterwards.

Very cool!

[Korath]

Now doesn't untap at all if you don't control a creature, even if it's not controlling one.