Jump to content

  • Log In with Google      Sign In   
  • Create Account

glibc: munmap_chunk(): invalid pointer


Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.

  • You cannot reply to this topic
9 replies to this topic

#1 Arppa   Members   -  Reputation: 122

Like
0Likes
Like

Posted 14 February 2008 - 04:32 AM

I'm working on a SDL+GL project and I encountered a following error and I can't figure out what's the problem. Code: Uint8* key = NULL; key = SDL_GetKeyState(NULL); // Error occures here if (key[SDLK_0]) // handle key ... and throws a following error *** glibc detected *** /home/arppa/projs/debrislinux/bin/Debug/debrislinux: munmap_chunk(): invalid pointer: 0x0816cdf8 *** + backtrace + memory map i'm using sdl version 1.2.12 Has someone seen this error before and knows what's wrong. I'm pretty sure it's not the code becouse it runs fine on windows. Thanks in advance.

Sponsor:

#2 keltar   Members   -  Reputation: 133

Like
0Likes
Like

Posted 15 February 2008 - 08:42 PM

Make your program with -O0 -ggdb -g3 flags (and if you have gcc with SSP extension, -fstack-protector-all). If program still fails, run it in debugger (on linux, i recommend gdb), and look on backtrace.


#3 truthsayer   Banned   -  Reputation: 100

Like
0Likes
Like

Posted 16 February 2008 - 12:29 AM

There's no way that that piece of code causes that error, unless SDL_GetKeyState has a bug in it, which is very unlikely. What's happening is that you are corrupting the heap somewhere else in your program and this is where it happens to manifest. Use Valgrind to track down where you are trashing memory.

#4 Arppa   Members   -  Reputation: 122

Like
0Likes
Like

Posted 16 February 2008 - 03:12 AM

I tried compiling with -O0 -ggdb -g3 -fstack-protector-all but with no success.
I've used exactly this same code on windows but I'm now compiling the code on linux and getting that error. When I comment out: key = SDL_GetKeyState(NULL); everything works, no errors, but also no key states :( . Of course I could just boot back to windows, but I'm curious what causes that error. I'm using glibc version 2.7

#5 truthsayer   Banned   -  Reputation: 100

Like
0Likes
Like

Posted 16 February 2008 - 03:15 AM

Quote:
Original post by Arppa
I tried compiling with -O0 -ggdb -g3 -fstack-protector-all but with no success.
I've used exactly this same code on windows but I'm now compiling the code on linux and getting that error. When I comment out: key = SDL_GetKeyState(NULL); everything works, no errors, but also no key states :( . Of course I could just boot back to windows, but I'm curious what causes that error. I'm using glibc version 2.7


Valgrind will tell you where the problem is instantly.

#6 Arppa   Members   -  Reputation: 122

Like
0Likes
Like

Posted 16 February 2008 - 03:51 AM

I tried Valgrind and for what I understand the output it seems not to be my coding error.
Full output: http://pastebin.com/m4895d6ec

#7 truthsayer   Banned   -  Reputation: 100

Like
0Likes
Like

Posted 16 February 2008 - 03:59 AM

Quote:
Original post by Arppa
I tried Valgrind and for what I understand the output it seems not to be my coding error.
Full output: http://pastebin.com/m4895d6ec


The first two entries might be issues with the system libraries. They're potential security problems, but we'll ignore them for now.


==20273== Invalid write of size 4
==20273== at 0x804B4AA: dSystem::dInputManager::LoadIdentityKeySet() (dInputManager.cpp:309)
==20273== by 0x804C43B: dSystem::dInputManager::dInputManager() (dInputManager.cpp:16)
==20273== by 0x805202F: main (main.cpp:73)
==20273== Address 0x65a4fdc is 4 bytes after a block of size 304 alloc'd
==20273== at 0x4021F14: operator new[](unsigned) (in /usr/lib/valgrind/x86-linux/vgpreload_memcheck.so)
==20273== by 0x804C3ED: dSystem::dInputManager::dInputManager() (dInputManager.cpp:8)
==20273== by 0x805202F: main (main.cpp:73)


This means you are allocating some memory at line 8 of dInputManager.cpp (in the dInputManager constructor), and later you are writing 4 bytes past the end of the chunk of memory. This overwrite happens in dInputManager.cpp at line 309. This is a real heap corruption bug and might be the cause of your problem.

The others are all saying that you're using memory that has not be initialized with a known value. For example, look at the code in dMath.cpp at line 23. You're passing a value to sqrt which has not been initialized (i.e. it contains random data). You need to look through your code to work out how this could happen and fix it up.

Also, you didn't paste the full log or it got truncated by the pastebin. I can't see the entries where Valgrind ends and reports any memory leaks, which also means I might not be seeing any other errors it reported that could be causing your problem.

#8 Arppa   Members   -  Reputation: 122

Like
0Likes
Like

Posted 16 February 2008 - 04:46 AM

That fixed the problem. I've to say that Valgrind is a quite good tool. Many thanks to you!

#9 songuke   Members   -  Reputation: 122

Like
0Likes
Like

Posted 06 March 2008 - 07:21 PM

Hi guys, I have this error, too.

My situation is much more strange. I created a window with some widgets using gtkmm. I also rewrite MFC's CArray for use in Linux. I found that the constructor of CArray is automatically called several times although I didn't declare any CArray in the main function. I even empty the main() function and remove all includes, like this:

int main() { return 0; }

and then run the program. It still throws out the munchunk() glibc error after calling several times the CArray constructor. Really strange, as the body of the main is empty!!!

I guess there's a problem with those linking libraries of gtkmm?

I'm using Eclipse on Ubuntu in joint with gtkmm library for my project.

Any hints? Thanks in advance.

#10 songuke   Members   -  Reputation: 122

Like
0Likes
Like

Posted 06 March 2008 - 08:26 PM

Phew, finally know what the reason is.

Actually I did allocate some elements in the array like T* elements = new T[size] and then only use "delete elements" in the destructor. This will leave some trash elements remaining in the heap so some further operations (which?, who knows!) will throw errors. I changed to "delete[] elements" to ensure every objects in the array is deleted properly and the error is gone.

Just wonder why those error is so terrible to figure out? (glibc throws exception although I didn't touch it)

And many thanks to the Valgrind tool since I would not have know the bug without it. :)




Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.



PARTNERS