[3.5-dev] Reproducing chaos/polymorph crash

[molybdenum]

You can reproduce easily by summoning a chaos breather and a bunch (20+) of monsters. As long as the monster pack stays dense, a crash should occur. You can also just zapping a wand of polymorph at things until it happens.

The issue is actually in project_m(), which includes a notice at the top about polymorph crashes. In the first above case, this is what seems to be happening:
- chaos breather triggers random polymorph on another monster
- place_new_monster_one() (by way of place_new_monster()) can't place the monster in the location it wants to, so it returns false
- m_ptr is null, since cave_monster_at() doesn't return a monster
- mon_clear_timed() (only in the who > 0 branch) crashes because m_ptr is null

In the second case, the first three steps are the same. The crash just occurs when the check for dam > m_ptr->hp is made in the else branch of who > 0.

[Derakon]

Why can't place_new_monster_one() place a monster? Because the target monster is still in place?

Sorry, I shouldn't be asking questions when you've already provided a lot of helpful information. Thanks for the debugging!

[molybdenum]

Why can't place_new_monster_one() place a monster? Because the target monster is still in place?

Specifically, in my cases, it was because cave_isopen() returned false. I was doing this in fairly open rooms or in town, but I suppose there could have been something on the ground that I didn't pay attention to. Also, plenty of monsters were polymorphed successfully, so it's not happening all the time. I also think that I've had polymorph crash when IDing a wand in a corridor against a single monster.

[PowerWyrm]

You can reproduce easily by summoning a chaos breather and a bunch (20+) of monsters. As long as the monster pack stays dense, a crash should occur. You can also just zapping a wand of polymorph at things until it happens.

The issue is actually in project_m(), which includes a notice at the top about polymorph crashes. In the first above case, this is what seems to be happening:
- chaos breather triggers random polymorph on another monster
- place_new_monster_one() (by way of place_new_monster()) can't place the monster in the location it wants to, so it returns false
- m_ptr is null, since cave_monster_at() doesn't return a monster
- mon_clear_timed() (only in the who > 0 branch) crashes because m_ptr is null

In the second case, the first three steps are the same. The crash just occurs when the check for dam > m_ptr->hp is made in the else branch of who > 0.

The error comes from place_new_monster_one() where a check is made against cave_isempty() while it should be cave_isopen(). There is no reason not to place a monster on top of an object.

[molybdenum]

The error comes from place_new_monster_one() where a check is made against cave_isempty() while it should be cave_isopen(). There is no reason not to place a monster on top of an object.

Yeah, I actually did more testing this weekend and changed that. It's waiting in a pull request.