Exception when debugged process is compiled without "-g" about cma HOT 4 CLOSED

Hi,

I am sorry for deal with this issue later because I just caught cold in last 2 days.
The cause of this issue is I forgot remove a temp "raise Exception".
I just committed a new commit that can handle this issue in my part. Could you help me do a test?

Thanks,
Hui

from cma.

I've tested it with this basic tool without -g, also stripped and also with PIE. It works but it seems to be quite slow backtracing all mallocs. I'll test with Linux games later on.
With my own game cheating tool ugtrain I first record all mallocs, use scanmem to find an unique malloc and this way I get the object size. In the second step I filter with it and need to backtrace only few mallocs.

from cma.

I have the plan for add filter and put more clear data in CSV file.
I think the cause of speed issue is when GDB do backtrace with a program that doesn't have enough debug info, it need parse the asm code to get its behavior.

PS, the current issue is fixed, right? Do you mind close it and open some new issues for the current troubles.

from cma.

Yes, let's close this as it works on Ubuntu 12.04 precise at least. There are a lot more issues. E.g. Ubuntu 14.04 trusty complains about syntax errors.

I also still don't get the advantage over using a C library hooking memory allocations. This is not stopping the CPU all the time and can work with parallel threads. Works so good that I saw interleaving and have to use flockfile(); fputs_unlocked(); funlockfile(); to write the memory information out with stream caching but without interleaving. Realtime stuff like OpenGL will just end the process if it detects that it's running too slow. Even just using GNU backtrace() at a depth of 3 to just get code addresses on the stack on all allocations is too slow.

from cma.