irregular segfaults during game play

[LynxAbraxas]

Observed irregular segfaults that happen after playing a few up to many rounds (master branch run under linux, sometimes not occurring for hours) no obviously related to some specific game state. Probably needs gdb debugging during playing.

[LynxAbraxas]

Tested if SDL surface locking is causing this with the changes from (LynxAbraxas@6a96bd2) and (LynxAbraxas@a6998b7) the crashes seem to be replaced by dead-locks of the game.

[MartinGuehmann]

Hey, I got a segfault while it was running on the debugger, must have something to do with the mouse and the blitter. Concerning that I changed the blitter thing on the Windows version, this does not seem so unsurprising.

#0 ??() at
#1 SDL_LowerBlit() at /usr/lib/x86_64-linux-gnu/libSDL-1.2.so.0
#2 SDL_UpperBlit() at /usr/lib/x86_64-linux-gnu/libSDL-1.2.so.0
#3 aui_SDLBlitter::Blt16To16(aui_Surface*, tagRECT*, aui_Surface*, tagRECT*, unsigned int)() at
#4 C3Blitter::Blt16To16(aui_Surface*, tagRECT*, aui_Surface*, tagRECT*, unsigned int)() at
#5 aui_Blitter::Blt(aui_Surface*, int, int, aui_Surface*, tagRECT*, unsigned int)() at
#6 aui_Mouse::BltDirtyRectInfoToPrimary()() at
#7 aui_UI::Draw()() at
#8 aui_UI::Process()() at
#9 CivApp::ProcessUI(unsigned int, unsigned int&)() at
#10 CivApp::Process()() at
#11 CivMain(int, char**)() at
#12 main() at

Unfortunately, I haven't figured out how to use the debugger in kdevelop. It shows me the frame stack but I can not click on the function and it sends me to the place in the source file if it is a cpp file. However, it does it if it is a c file.

The only thing I would do before really debugging is to merge in the changes from the Linux branch. It might be fixed there. Or the problem might be different. But I think that is could also be related to the mouse artifacts.

[LynxAbraxas]

Yes, that looks very much like the ones I got, see e.g. the commit message of a6998b7.

[LynxAbraxas]

Porting the game to SDL-2 (#92) might solve this issue, it might be better to spent time on #92 before loosing unnecessary time on this issue.

[LynxAbraxas]

While debugging with f54c83e caught this:

Thread 1 "ctp2" hit Catchpoint 1 (signal SIGSEGV), 0x00007ffff76d8740 in SDL_Flip () from /usr/lib/x86_64-linux-gnu/libSDL-1.2.so.0
#0  0x00007ffff76d8740 in SDL_Flip () from /usr/lib/x86_64-linux-gnu/libSDL-1.2.so.0
#1  0x0000555555845376 in aui_UI::Draw (this=0x5555578725bc) at ui/aui_common/aui_ui.cpp:1126
#2  0x0000555555846d6a in aui_UI::Process (this=0x5555578725bc) at ui/aui_common/aui_ui.cpp:1849
#3  0x00005555557809a4 in CivApp::ProcessUI (this=0x555556cb090c, target_milliseconds=30, used_milliseconds=@0x7fffffffeb0c: 0) at ctp/civapp.cpp:2397
#4  0x0000555555780e4a in CivApp::Process (this=0x555556cb090c) at ctp/civapp.cpp:2550
#5  0x0000555555776c68 in CivMain (iCmdShow=2, pSzCmdLine=0x7fffffffecb8) at ctp/civ3_main.cpp:1694
#6  0x0000555555776805 in main (argc=2, argv=0x7fffffffecb8) at ctp/civ3_main.cpp:1440

They have in common that they happen in aui_UI::Draw.

[LynxAbraxas]

Tested #298 for many turns now and did not have this segfault any more. Since it did not appear often (sometimes not within weeks), I cannot say for sure that is solve with #298 but am confident that it might be.

[MartinGuehmann]

I would link this to #298, so that this is closed when #298. If #298 indeed fixes this issue then we can just forget about. If it doesn't then we can still reopen this issue or open a new one.

[LynxAbraxas]

I would link this to #298, so that this is closed when #298. If #298 indeed fixes this issue then we can just forget about. If it doesn't then we can still reopen this issue or open a new one.

Sure, I just can't do the linking;-)

[MartinGuehmann]

I see.