2021-03-04 12:06 EST


View Issue Details Jump to Notes ]
IDProjectCategoryView StatusLast Update
0001654FSSCPgameplaypublic2013-03-26 00:26
ReporterARSPR 
Assigned To 
PrioritynormalSeveritytweakReproducibilityalways
StatusnewResolutionopen 
PlatformDELL Dimension 9200OSWindows VistaOS VersionUltimate 32-bit
Product Version 
Target VersionFixed in Version 
Summary0001654: Crazy movements at the begining of the autopilot run when not leading it.
DescriptionThis is a "spin-off" from 0001520 and I feel a tweak would be fine in autopilot code.

As the root problem of 0001520, Taylor has fixed the different object vs. wing order which caused pretty strange behaviours. Now, the object order in the fs2 file doesn't matter anymore.

Now the autopilot run is lead by the first wing ship (ie. by Alpha 1). But then a tweak is needed in old autopilot code, (not cinematic autopilot):

+ When player's ship doesn't lead the run, (the player is not Alpha 1 but Alpha n), it makes crazy movements before stabilizing itself in formation with the wing first ship. These movements should be dampened somehow.
Additional InformationYou can test with any of the two missions from 0001520. (Now they behave exactly identical).
TagsNo tags attached.
Attached Files

-Relationships
+Relationships

-Notes

~0010688

Woolie Wool (reporter)

Bumping this to call attention to the serious bugginess of autopilot in recent builds.

~0010852

chief1983 (administrator)

Here's some info on some builds which were around some possible autopilot-breaking commits:

http://www.freespacefaq.com/Misc-Downloads/Builds/Old/Team_Loadout_Build.7z (Jan 26 2008 Stable Branch, before a bunch of Feb autopilot commits)

http://fs2source.warpcore.org/exes/latest/20080312-Goober5000.7z (March 12 2008, I think stable branch, right before the below one if so)

http://fs2source.warpcore.org/exes/latest/C03162008.zip (March 16 2008, right after WMC committed some scripting stuff)

http://wwwcsif.cs.ucdavis.edu/~chos/fs2_open_trackir.zip (March 5, 2008, another one before the last one if the goober one turns out to be the HEAD branch or something)

There was also a commit on August 30, but that's since I started the nightly build system so you should be able to find a build from that time frame easily, before and after r4773

I sent this info to Woolie over IM but thought adding it to some mantis reports would be prudent.

If some testing with those builds could be done, it might help. They should all be trunk/stable branch, not Xt stuff, so if any of those do/don't work it will help more in tracking down the problem.

~0010856

taylor (administrator)

Have the commits from me on March 20, 2008 been ruled out yet (both with "1520" in the description)? There were several commits which were made to address all of these autopilot bugs. The code was tested by ARSPR before commit and I tested using original and cinematic autopilot missions as well. I don't know how something passed through so much testing on a single feature, but it's possible that something was b0rked in the commit (something missing from the Xt tree) or that so other commit broke it later on.

Also, the changes that I committed were NOT what I wanted to go with. There was a post in the SCP internal concerning the issue with special_ship and how the AI handled wing formations with the player not being the wing leader (if special_ship != 0). That original code was given the ok by ARSPR but then rewritten to be what eventually made it into SVN to fit with Goober's explanation of how it should work.

~0010857

chief1983 (administrator)

Actually, according to Woolie, the autopilot stuff he's using worked fine in the March 16 build. So I don't know what the commits on March 20 were suppposed to fix. But according to him, the mid-August build is broken, so something between March 16 and mid-August broke whatever he's been checking. So, if http://fs2source.warpcore.org/exes/latest/20080321-Goober5000.7z could be tested that would help eliminate that set of commits. That's one of the first released SVN trunk builds and should have what taylor is referring to.

~0010858

Woolie Wool (reporter)

The 3/21 build is broken. We may just have our culprit.

~0010859

taylor (administrator)

The code was seriously broken before, but depending on how the mission was set up it might not have been obvious.

The part dealing with this particular bug, the movement problem at wing formation start, is due to the AI code always assuming that the player ship is the wing leader. The formation code has always been wonky movement wise, but since the order bug was fixed it is now painfully obvious that it is really screwed up, because the player gets bunched in with the AI on formation rather than being the leader of said formation. This is the exact reason that I wanted the code to make the player always be the wing leader, the AI formation code simply wasn't set up to be "nice" otherwise.

~0010865

chief1983 (administrator)

I'm not sure whether it was this bug specifically being tested, but Woolie was trying the same test with several builds, and whatever he was testing worked fine in a pre-SVN March 16 build, that I think was mostly stable trunk code, but didn't work in one of the first SVN trunk builds from March 21. I was thinking that should help narrow down some problems somewhat.

~0012400

The_E (administrator)

Is this still current? Or was it fixed together with the other autopilot stuff?

~0012406

chief1983 (administrator)

Probably fixed, maybe Woolie could vouch for it.

~0012449

The_E (administrator)

Not fixed. The AI still does some pretty violent maneuvers when getting into position.

~0014839

Woolie Wool (reporter)

Last edited: 2013-03-26 00:26

View 3 revisions

Bumped to report even more autopilot weirdness. If you autopilot to an asteroid field and are NOT the wing leader, and are using cinematic autopilot, you will never be able to autopilot away as you will be instantly teleported back where you came from. This does not happen if you are wing leader.

Autopilot code seems to assume that the player is the wing leader. This is not a wise assumption to make.

Discovered this tonight while FREDing a mission for Twist of Fate. Will create proof of concept for regular FS2 tomorrow.

EDIT: This could really be changed to "autopilot does not work correctly if player is not wing leader" and bumped in priority because this can totally break a mission. It's not a minor thing.

+Notes

-Issue History
Date Modified Username Field Change
2008-04-03 14:19 ARSPR New Issue
2009-02-23 18:12 Woolie Wool Note Added: 0010688
2009-04-29 16:50 chief1983 Note Added: 0010852
2009-04-30 18:16 taylor Note Added: 0010856
2009-04-30 18:31 chief1983 Note Added: 0010857
2009-04-30 18:53 Woolie Wool Note Added: 0010858
2009-04-30 19:16 taylor Note Added: 0010859
2009-05-01 11:34 chief1983 Note Added: 0010865
2010-10-08 06:46 The_E Note Added: 0012400
2010-10-08 11:05 chief1983 Note Added: 0012406
2010-11-11 08:26 The_E Note Added: 0012449
2013-03-26 00:24 Woolie Wool Note Added: 0014839
2013-03-26 00:25 Woolie Wool Note Edited: 0014839 View Revisions
2013-03-26 00:26 Woolie Wool Note Edited: 0014839 View Revisions
+Issue History