Schrödinger's bug

[Nick]

So (fingers crossed) I am now back to some sort of regular development work, and my first task is to try and finish tracking down this bug:

Bug: Items previously found and ignored being redisplayed as unknown objects after another treasure detection within their range. Note the two red asterisks to @'s left. These were previously ignored objects, but now show again after use of treasure detection.

I have identified that the problem is player knowledge of objects being wiped on closing and reopening the game. So I thought I would track down where the loss of information was happening by running in gdb and watching the appropriate known object. Result? The object I was watching remained known, but all others still disappeared.

While I'm slightly flattered by how important the game obviously considers me, it doesn't help fix the bug. Any suggestions? I think optimisation is already set to 0 (unless something is overriding what's in src/Makefile).

[debo]

Any suggestions?

Remove the ID game and monster knowledge and have everything known to the player from when they start the game?

[Nick]

Remove the ID game and monster knowledge and have everything known to the player from when they start the game?

Wouldn't help - this is whether the player remembers that individual found items on the current level even exist.

Your attempt to help is noted.

[Derakon]

Does it consistently work like that, that the item you keep an eye on remains known while others vanish?

Since you're closing and re-opening the game, I kind of assume you don't have a "reading memory that has been freed" type of situation here. What are the repro steps?

[Pete Mack]

There's another problem. After detecting once, detecting a second time does not remove items picked up by monsters

[t4nk]

I have identified that the problem is player knowledge of objects being wiped on closing and reopening the game. So I thought I would track down where the loss of information was happening by running in gdb and watching the appropriate known object. Result? The object I was watching remained known, but all others still disappeared.

While I'm slightly flattered by how important the game obviously considers me, it doesn't help fix the bug. Any suggestions? I think optimisation is already set to 0 (unless something is overriding what's in src/Makefile).

Well, here's one bug: use item with effect_handler_SENSE_OBJECT (e.g., scroll of treasure detection), detect some objects and close the game. Load it again and the red '*' are no longer there. As far as I can tell, that's because object_sense() in obj-knowledge.c creates new object without tval and sval (that is, with tval and sval 0, since it's mem_zalloc'ed). What is needed is tval 0 ("none") and sval 2 ("unknown item") (map_info() looks for that, specifically for obj->kind == unknown_item_kind). Objects with tval and sval 0 cannot be loaded (load.c, rd_item()):

	obj->kind = lookup_kind(obj->tval, obj->sval);

	/* Check we have a kind */
	if ((!obj->tval && !obj->sval) || !obj->kind) {
		object_delete(&obj);
		return NULL;
	}

[Nick]

Well, here's one bug: use item with effect_handler_SENSE_OBJECT (e.g., scroll of treasure detection), detect some objects and close the game. Load it again and the red '*' are no longer there. As far as I can tell, that's because object_sense() in obj-knowledge.c creates new object without tval and sval (that is, with tval and sval 0, since it's mem_zalloc'ed). What is needed is tval 0 ("none") and sval 2 ("unknown item") (map_info() looks for that, specifically for obj->kind == unknown_item_kind).

Awesome, problem solved