More Xconq bug fixes

More Xconq bug fixes

[Hans Ronne]

My latest bunch of bug fixes has been checked in. Most of them are
concerned with how the kernel and AI code deal with non-mobile units that
can fire. This kind of unit is not very common in existing games, and was
therefore not handled correctly.

One new feature is the ability to set maximal and minimal sizes for
advanced units (size-min and size-max). This is useful if you want to avoid
runaway unit growth or units starving to death.

I have not dealt with the problems that were recently discussed on the list
yet, except for some preliminary moves, such as adding code for unit view
validation and clearance (still unused).

Hans

More Xconq bug fixes

[Hans Ronne]

I have checked in a number of improvemets to the repair and resupply code.
I believe that most of the bugs in this code, some of which were serious,
have now been fixed. This means that units no longer get trapped in reserve
mode because of failed repair tasks etc.

Other recent improvements:

1. The ai now knows how to do explicit repair, at least for the same unit
(self-repair).

2. The ai has been improved, in particular its action reaction code and
responsiveness when switched on and off.

3. Colonizing units now behave a little more intelligently.

4. Several kernel bugs affecting combat model 1 games such as civ2 have
been fixed.

5. A new unit namer "in-order" picks names in order instead of randomly
from a name list. This is a useful game design feature that has so far been
missing.

6. A serious bug in the unit-wrecking code has been fixed.

7. Several bugs affecting self-units have been fixed, and the associated
messages have been improved.

8. Several bugs in the MacPPC interface have been fixed.

9. Most of the remaining graphics problems on Windows have been fixed.
Terrain transitions, overlays and unseen fuzz are now drawn also on
Windows, and the "black-is-transparent" and "translucent edge cell" bugs
have been fixed.

10. BWidget has been updated to 1.6.0.

11. The tcltk help window has been overhauled. The advances tree is now
drawn correctly also in the MacTCL interface.

12. The help node headings suggested by Richard Hunt have been implemented.

This burst of activity does not mean that I have given up on updating the
lib and doc files before making a pre-release. However, going through all
the games and testing them did reveal a number of more or less severe
kernel bugs that I was unaware of. Hence the above fixes.

Hans

More xconq bug fixes

[Hans Ronne]

I have now checked in some more bug fixes, this time to the unit view code.
One of them did have a rather sever effect on the AI. Thus, it prevented
the AI from using the normal combat code (do_hit_unit_task) in any game
where see-all was not true. The AI could still fight in these games, but
only by overrun actions. Basically, it would first try to attack an enemy
unit, always fail, and then stumble over the unit when trying to enter its
cell. Firing at a distance by the AI did not work at all when see-all was
false (which is how I discovered this bug).

So if you thought the AI was lame in non-see-all games, you should check
out the new code. If you try the Magnuszew game without see-all enabled you
will notice that the AI now can fire at a distance, just as in see-all
games.

Hans

P.S. I have also checked in Lincoln Peters' improved AI explorer code.

Hans Ronne

hronne@pp.sbbs.se

More xconq bug fixes

[Hans Ronne]

I checked in a couple of new bug fixes yesterday. One bug (just a typo, but
a significant one) would cause acp-independent units to build nothing but
siege units. Hence the myriads of catapults in the Third Age game.

Another bug in the kernel code for occupant-transport volume checks would
cause crashes in the build code. It dealt with the rather unusual (but
still possible) situation where a unit builds another unit around itself,
perhaps this is why it escaped detection until now.

Hans

Hans Ronne

hronne@pp.sbbs.se