Darkvater opened the ticket and wrote:

The problem is obvious, game crashes.
To reproduce: "bridge_crash.sav": any time any part of the train is under the bridge.
To reproduce: "bridge_crash2.sav" (with patch): only at that exact moment.

The original problem (been there for ages, in 0.4.0.1 as well) is that EnsureNoVehicleProcZ() is called and the tile is released as free-to-demolish when the v->z_pos and ti->z don't match. This is incomplete, as a train on a slope will have z values ranging from the bottom to the top. Eg from 8 to 16 if it goes from level 1 to 2. The patch fixes that.
The second problem (bridge_crash2.sav) still occurs however. This is because at the exact moment the train enters the tile, it's 'Z'-value is still 8; no change made. Currently there is no way to fix this without an overhaul of UP/DOWN detection (which is XXX-d and is crap), as there is not a single variable that tells us if the train is going up/down. These variables are cleared just a few lines before in AfterSetTrainPos:2603 in train_cmd.c.

A note: bridge_crash2.sav, while it does happen, does this VERY sporadically. To reproduce it faithfully I had to breakpoint the game at the exact moment the train enters the tile and pause it programmatically. Nevertheless it can happen.

Any thoughts on the patch? I think it is sufficient for the time (using > instead of >= will break deleting tracks under bridges); any better fixes need a lot more work. Also noted that the hash is totally braindead and doesn't really work as required.

Please test and give thoughs.

Attachments

• bridge_crash.diff (0.70 KiB)

Reported version: trunk
Operating system: All

This issue was imported from FlySpray: https://bugs.openttd.org/task/40