Personal tools

Author Topic: R25602 - a crash (NOT actor still moving) and some unrelated feedback...  (Read 893 times)

Offline frr

  • Cannon Fodder
  • **
  • Posts: 6
    • View Profile
Today when I started a game by loading a savegame taken very early in the game (first ufo recovery), with "setdeveloper DEBUG_EVENTSYS" I managed to finish that recovery session. But the very next recovery session failed again, and it was NOT the "actor still moving" message. I'm attaching a log of ufoai's stdout+stderr, and the events.log.

My system configuration:
Fedora 10 64bit on a cheap all-Intel notebook
i965GM, integrated Intel graphics, C2D CPU, 2 GB RAM.

Speaking of my system configuration: about two months ago, I had a problem that seemd to stem from the Intel graphics driver. With "realtime lighting" and "GLSL shaders" both on, Battlescape initialization used to fail reliably with the three bargraphs overflowed left, and the system would freeze solid (i.e. a complete hang of the kernel). A normal user-space memory leak wouldn't have a chance to cause this effect :-) See also the JPG image attached. As I said, this does NOT seem to happen anymore. Now I have both "realtime lighting" and "GLSL shaders" turned on and I've been through two battlescape sessions without this problem. I'm not wasting time with this bug, because I know that the current Fedora release is r11 (while I have r10 installed), and the Intel graphics driver has just undergone a bit of revolution (between r10 and r11), the dust hasn't quite settled yet...

Actually I've tried loading a savegame once more, and with those graphics features turned on, I got stuck in the "Battlescape empty screen" once again (with the three bargraphs just empty). So maybe I'll just turn them off again, when I feel like zapping some more aliens...

I'm aware that this message probably doesn't help much in the way of precise debugging information :-)
I intend to provide this as just a bit of informal feedback, from my odd setup...
Overall I have to say that the 2.3-DEV series seems to be gradually improving - every time I compile a new version, some of the bugs are gone, the look and feel is better, there are new models, even new music... it's great :-)

Offline Mattn

  • CaveExpress
  • Administrator
  • PHALANX Commander
  • *****
  • Posts: 4830
    • View Profile
    • CaveExpress
it's the crouchingState assert - this should already be fixed in current trunk