Join GitHub today
GitHub is home to over 50 million developers working together to host and review code, manage projects, and build software together.
Sign upFix #7670: Cache the origin tile to prevent recurring calls to the road pathfinder when a vehicle is blocked by another #7822
Conversation
SamuXarick
force-pushed the
SamuXarick:blocked-road-vehicle-yapf-fix
branch
from
fc4d6b1
to
dee9710
| Trackdir origin_td = INVALID_TRACKDIR; | ||
| TileIndex origin_tile = INVALID_TILE; |
This comment has been minimized.
This comment has been minimized.
nielsmh
Feb 9, 2020
Contributor
The RoadVehPathCache structure is saved in the savegame, and these two new members need to be stored as well.
You need to add them to the saveload data and add a SL-version bump.
This comment has been minimized.
This comment has been minimized.
SamuXarick
Mar 4, 2020
•
Author
Contributor
I did as you told me, but I don't think that in this particular situation, saving is required.
This comment has been minimized.
This comment has been minimized.
SamuXarick
Mar 4, 2020
•
Author
Contributor
From IRC:
but yeah, i'm more convinced there is a chance for a desync, just not sure how to properly set this up
let me think. Server has the origin tile cached the moment a client joins. Client doesn't have it cached, but 1 tick later it will, and be in sync with server
what would happen if...
in that tick
i remove the road
how am i gonna remove a road in the exact tick
pause command
remove road, make client join, unpause
now I should be getting a desync, let's see
confirmed!
vehicle went different ways
You're right. Desync confirmed. Save is required indeed.
https://i.imgur.com/gUmDN4J.png
SamuXarick
force-pushed the
SamuXarick:blocked-road-vehicle-yapf-fix
branch
from
dee9710
to
d9c7987
SamuXarick
force-pushed the
SamuXarick:blocked-road-vehicle-yapf-fix
branch
from
d9c7987
to
bb267f8
…ad pathfinder when a vehicle is blocked by another
SamuXarick
force-pushed the
SamuXarick:blocked-road-vehicle-yapf-fix
branch
from
bb267f8
to
ca05e3c
SamuXarick commentedNov 3, 2019
•
edited
This is a tentative fix, but not fully tested. I thought of caching the origin of where the pathfind starts its search. It seems to solve the issue, but I'm not sure this is the right approach.