r24148 - Actor movement out of sync - server crashed

[dast]

made a new build, started campaign - first mission map +farm

-map seemed not to be correctly aligned - visible lines between map blocks
-actor was able to move through the drop ship

in the console.log lots of s.b. but only after a short time in game (actor leaving starting block of map?)

Shutdown server: Server crashed.
********************
ERROR: LET_PathMove: Actor movement is out of sync: 139:115:0 should be 139:121:0 (step 1 of 1)
********************

Shutdown server: Server crashed.
Wrote config.cfg.
Shutdown server: Server quit.
Shutdown gametype 'Campaign mode'
Cvar 'mn_base_id' wasn't found
Shutdown server: Quitting server.
Wrote keys.cfg

[voteslave]

similar thing here on Ubuntu 9.04


********************
ERROR: LET_PathMove: Actor movement is out of sync: 134:132:0 should be 141:134:0 (step 1 of 1)
********************
Shutdown server: Server crashed.
==== ShutdownGame ====
Server crashed.
./ufo[0x818cce9]
./ufo[0x818cd49]
[0xb7f40400]
./ufo[0x8116e51]
./ufo[0x810758a]
./ufo[0x81075ad]
./ufo[0x81075ad]
./ufo(MN_Draw+0xfd)[0x8107899]
./ufo(SCR_UpdateScreen+0xbe)[0x809886b]
./ufo(CL_Frame+0xd8)[0x8090003]
./ufo[0x813c470]
./ufo(Qcommon_Frame+0x9a)[0x813c739]
./ufo(main+0x66)[0x818ba51]
/lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe5)[0xb7774775]
./ufo[0x8069571]
Error: Received signal 11.



did the following before running the game - this is the first time i updated via svn - is that all i need do? type in make maps?

voteslave@quad:~/ufoai/trunk$ svn up
U    src/client/menu/node/m_node_text.c
A    contrib/dlls/libtheora-0.dll
A    contrib/dlls/libtheoradec-1.dll
A    contrib/dlls/libtheoraenc-1.dll
Updated to revision 24329.
voteslave@quad:~/ufoai/trunk$ make maps
make: Nothing to be done for `maps'.

[Duke]

@voteslave:
On which map did you get the crash ?

[voteslave]

sorry - can't recall

but i was worried about whether i had recompiled maps

prior to the 'svn up' info i posted below i had done another one which i think updated a maps - but nothing happened then either with 'make maps'

so after any svn up i think all i need do is simply:
voteslave@quad:~/ufoai/trunk$ make maps

or am i missing something - like do i need do ./configure and make as well???

[voteslave]

just tried again - the map was africa A terror mission)

ERROR: LET_PathMove: Actor movement is out of sync: 119:122:0 should be 123:121:0 (step 3 of 3)
********************
Shutdown server: Server crashed.
==== ShutdownGame ====
./ufo[0x818cce9]
./ufo[0x818cd49]
[0xb80c2400]
./ufo(HUD_ActorUpdateCvars+0x45)[0x807d242]
./ufo(CL_SlowFrame+0x10)[0x8090029]
./ufo[0x813c470]
./ufo(Qcommon_Frame+0x9a)[0x813c739]
./ufo(main+0x66)[0x818ba51]
/lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe5)[0xb78f6775]
./ufo[0x8069571]
Error: Received signal 11.

[voteslave]

just done it again on 24347

i also noticed that it seems to happen after the game switches from full screen to windowed  - it does this itself for no apparent reason

[Duke]

Hmm. Too bad. +Africa is a random map.
We're looking for someone who gets this error on a non-random map.

[Farmazon]

Rev 24322. i have this bug repeating all time at any map. It happens when your soldier move then game switching to geomap you can change time speed and if hold esc button game start rapidly swithings between geomap and battlearea. Game runing at Vista (SP1 not activated) CPU q4590  M/B Asus p5q deluxe Video Sappfire radeon 4850

[Kaz]

Don't know if dam is non random, but I just got the same error message (with somewhat different numbers):

ERROR: LET_PathMove: Actor movement is out of sync: 137:132:1 should be 135:135:1 (step 3 of 3)

The error occurred while trying to move a sniper up the tower. On one round, the sniper made it just to the top of the first flight of stairs. As the game would not let me move the guy further into the second flight, I decided to try and crouch him in case there were some actor height issues. It seemed to work at first, since the grid turned green at the top floor, but the sniper did not move and the grid turned blue again. At this point I decided to try and move him just to the stairs and not beyond when I found a spot there which gave me a green grid. As I clicked to make the guy move, he started his crouched motion towards the stairs, then out the builduing through a wall, and back to the main menu (I started the map in skirmish mode). Checked thre console and the error message was there.

I also recall seeing the same error yesterday with an alien who was coming over the dam towards the tower walked over the railing and into the water just before the game crashed back to the main menu.

I hope this description helps reproduce the issue.

Still, progress has been made with pathfinding. About a week ago I could not move soldiers into the tower building.

Cheers!

[voteslave]

just done svn up and recompile

i noted this time that it said do "make pk3" which i hadn't done before - not sure if this is something new -do the web page instructions need updating

anyway - it still crashed on the africa map - though slightly different error (negative TU costs)

********************
ERROR: Negative TU costs while walking
********************
Shutdown server: Server crashed.
==== ShutdownGame ====
Server crashed.
./ufo[0x818d061]
./ufo[0x818d0c1]
[0xb7f67400]
./ufo[0x81170a9]
./ufo[0x810760a]
./ufo[0x810762d]
./ufo[0x810762d]
./ufo(MN_Draw+0xfd)[0x8107919]
./ufo(SCR_UpdateScreen+0xbe)[0x80988eb]
./ufo(CL_Frame+0xd8)[0x8090083]
./ufo[0x813c7e8]
./ufo(Qcommon_Frame+0x9a)[0x813cab1]
./ufo(main+0x66)[0x818bdc9]
/lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe5)[0xb779b775]
./ufo[0x80695f1]
Error: Received signal 11.

voteslave@quad:~/ufoai/trunk$

[Duke]

@Kaz:
Great dam.map is non-random (in the base/maps folder, the *.bsp files are non-random maps while the folders are RMAs). The numbers in the msg are x/y/z coords btw.

Your report should enable me to track it down. Once I can compile again
Meanwhile, if you want to help even more, you could toy a bit with that situation and try to find the least complex way to reproduce the crash ie. further narrow it down.

btw that was a *very* good bug report Do you happen to be a coder yourself ?

@others:
Thanks so far. No more reports on this issue for now. One reproducible situation is usually enough. But you should take personal notes of places on other non-RMAs if you encounter the crash there. I might ask you for such.

[Mattn]

negative TUs might be due to the way CL_ActorDoMove and LET_PathMove are implemented.

if the server decides that an actor is going into STATE_CROUCH while walking the following happens:
event for starting the move
end event
event for going into STATE_CROUCH
end event
event for starting the move with the crouched steps
end event when target is reached

it might happen that the CL_ActorDoMove resp. LET_PathMove functions won't note this fact at the right time.

[Kaz]

@Kaz:
Great dam.map is non-random (in the base/maps folder, the *.bsp files are non-random maps while the folders are RMAs). The numbers in the msg are x/y/z coords btw.

Your report should enable me to track it down. Once I can compile again
Meanwhile, if you want to help even more, you could toy a bit with that situation and try to find the least complex way to reproduce the crash ie. further narrow it down.

btw that was a *very* good bug report Do you happen to be a coder yourself ?

I will try and get some more crashes of this type to send any info. Is there some way to enable more detailed debugging info I could send if/when I get more crashes? You read me right I am a coder, but it has been ages since I've done C/C++ stuff. At the moment part of my full time job description is QA testing systems, which really benefits from detailed bug reports. Another part is fixing bugs, which gives you an appreciation for detailed reports...

Cheers!

[Mattn]

i've also seen (but not thought that much about it as i'm lacking time right now) that the EV_START_MOVE event is send to a maybe outdated vis_mask. See the attached patch. it should (again - untested and unproved) calculate the new vis mask for the new position of the actor and send the EV_START_MOVE event to all players that can see the actor on the new position already.

it would be nice if someone could verify this.

[attachment deleted by admin]

[Duke]

Is there some way to enable more detailed debugging info I could send if/when I get more crashes?

...but it has been ages since I've done C/C++ stuff.

UFO has lots of ways to produce debugging info. I prefer to use the debugger. So just find the fastest way to produce the crash. Like in: go from x1/y1/Z1 to x2/y2/z2 => crash.
You may find setting
g_notu 1
debug_map 1,2 or 3
useful though.

I understand I shouldn't try to persuade you to join the dev team (atm), but some qualified testing is what you want to contribute ? Fine with me A qualified tester can be worth much more than a 'rusted' coder

@Mattn:
*Who* do you think is qualified to verify your patch ? (I hope you're not thinking of ME, coz I haven't even started to look at the event system here...)
Or do you think your patch might fix the above crash ?