RttR Crashs after quitting

[Spikeone]

Everytime I quit RttR now the game gives me a message "RttR Crashed would you like to send... "

Latest rev
Win7 x64 Ultimate

[stefson]

I just stepped into this with system-boost-1.58 - still testing with 1.56

crash from a replay https://bpaste.net/show/9e544ebf9370
crash from a normal game https://bpaste.net/show/c0ccef3b1051

[Flamefire]

Thanks. This helps a lot:

8 0x00007ffff62cabad in vfprintf () from /lib64/libc.so.6
9 0x00007ffff637a4cb in _vfprintf_chk () from /lib64/libc.so.6
10 0x00000000007a82fa in Log::vwrite(char const, va_list_tag) ()
11 0x00000000007a8375 in Log::lvprintf(char const, _va_list_tag) ()
12 0x00000000007a8407 in Log::lprintf(char const*, ...) ()
13 0x000000000059dce9 in GameClient::Stop() ()

I'd really would weed out all these *printf functions as they easily cause problems and worse...

[stefson]

system-boost-1.56 is not affected. but the crash only occurs when using the [x] to close the game or replay while running - not when going through the menus (you know what I mean?)

[stefson]

system-boost-1.55 is affected, errors are slightly different.

game crash log https://bpaste.net/show/038d1300bb69
replay crash log https://bpaste.net/show/6d42003a4d2b

[Flow86]

propably a bug inside the Logger, I will provide a better logger soon, I already have something better in mind

Ah - one idea: could it be that you can't write into the logging directory or the logging directory is wrong in your build?

[stefson]

uhm, which directory is the logging directory and where is it supposed to be found at? (Linux x64 multilib)

[Flow86]

~/.s25rttr/LOGS

[stefson]

I have read + write access to it. So this not the problem :)

[Flamefire]

Found the problem: The Log singleton is freed before the client is freed. This causes access to invalid values...

[Flow86]

how come that the singleton is freed before the client O.o

[Flamefire]

Better question: How come the client was freed before the log?
Both are singletons. Their creation and therefore destruction order are not specified. We should certainly avoid using them. I'm currently implementing a dead-reference check for the singletons...
Made the Log a Phoenix, but same problem with LobbyClient -.- We probably need priorities for those