It was not a gcc bug, it was just gcc behaving differently (properly?) from VC on some very iffy code. This caused bad damage calculations. We believe that this then resulted in bad forces being calculated, and then physics code exploding the stack, a type of crash that is notorious for leaving little or misleading forensic evidence behind, and preventing all the usual error reporting from working. Plus, the code worked as intended on VC. We also fixed a problem at a global level to prevent NaNs or any numbers way out of expected range getting into the physics engine to begin with. I am hopeful, but not 100% confident, that this same issue was affecting the Dr. G weapons.
- Fletch From: hlds-boun...@list.valvesoftware.com [mailto:hlds-boun...@list.valvesoftware.com] On Behalf Of Saul Rennison Sent: Thursday, July 28, 2011 1:15 PM To: Half-Life dedicated Win32 server mailing list Subject: Re: [hlds] TF2 update coming this afternoon Good job all glad you fixed the problem. What was the GCC error by the way? It'd be interesting to share with us developers! On Thursday, 28 July 2011, Fletcher Dunn <fletch...@valvesoftware.com<mailto:fletch...@valvesoftware.com>> wrote: > We have fixed the crash related to the mantreads. (We also found a problem > with the damage calculation.) The problem was gcc-specific and thus only > affected Linux servers. We are mostly sure (though not 100% sure) that a > similar problem was happening with the Dr. G weapons. > > > > This update will probably be an optional update, since it isn't critical for > clients. But of course all the server operators will want to make sure and > get it. > > > > Further bulletins as events warrant. > > > > Your humble servant, > > Fletch > > -- Kind regards, Saul Rennison _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
