Creating STL exception leads to segfault

[atomaki]

Hi everyone! This is my first message on the forum.

When I add std::runtime_error("err") expression into my code, I got segmentation fault at address 0 (before entering to main). If instead I use std::exception(), app runs fine.

How std::runtime_error() causes segfault and how to fix it? Thanks!

Code: Select all

 
#include <stdexcept>
#include <exception>
#include "irrlicht.h"
using namespace irr;
int main() {
    IrrlichtDevice* device = createDevice();
    if (!device)
        throw std::runtime_error("err");  //without this run fine
}
 

Compile:

Code: Select all

 
g++ -ggdb -Wall -Wextra -std=c++11 -o main main2.cc \
-lIrrlicht -lGL -lXxf86vm -lXext -lX11 -lXcursor
 

Backtrace:

Code: Select all

 
(gdb) bt full
#0  0x00000000 in ?? ()
No symbol table info available.
#1  0xb5d88fe1 in init () at dlerror.c:177
No locals.
#2  0xb5d8942e in _dlerror_run          (operate=operate@entry=0xb5d88e10 <dlsym_doit>,     args=args@entry=0xbffff160) at dlerror.c:129
    result = <optimized out>
#3  0xb5d88e98 in __dlsym (handle=0x0, name=0xb7f8afe5  "posix_memalign") at dlsym.c:70
    args = {handle = 0x0, name = 0xb7f8afe5 "posix_memalign", who = 0xb7f674bc, sym = 0xb5d65b20}
    result = <optimized out>
#4  0xb7f674bc in ?? () from /usr/lib/nvidia-   304/libGL.so.1
No symbol table info available.
Backtrace stopped: previous frame inner to this frame     (corrupt stack?)
 

LD debug

Code: Select all

 
LD_DEBUG=all LD_DEBUG_OUTPUT=log.txt ./main
tail log.txt.*
 
   3848:    symbol=__pthread_key_create;  lookup in file=/lib/i386-linux-gnu/libgcc_s.so.1 [0]
  3848: symbol=__pthread_key_create;  lookup in file=/lib/i386-linux-gnu/libc.so.6 [0]
  3848: symbol=__pthread_key_create;  lookup in file=/usr/lib/nvidia-304/tls/libnvidia-tls.so.304.131 [0]
  3848: symbol=__pthread_key_create;  lookup in file=/usr/lib/nvidia-304/libnvidia-glcore.so.304.131 [0]
  3848: symbol=__pthread_key_create;  lookup in file=/usr/lib/i386-linux-gnu/libXext.so.6 [0]
  3848: symbol=__pthread_key_create;  lookup in file=/lib/i386-linux-gnu/libdl.so.2 [0]
  3848: symbol=__pthread_key_create;  lookup in file=/usr/lib/i386-linux-gnu/libxcb.so.1 [0]
  3848: symbol=__pthread_key_create;  lookup in file=/lib/ld-linux.so.2 [0]
  3848: symbol=__pthread_key_create;  lookup in file=/usr/lib/i386-linux-gnu/libXau.so.6 [0]
  3848: symbol=__pthread_key_create;  lookup in file=/usr/lib/i386-linux-gnu/libXdmcp.so.6 [0]
 

[hendu]

It clearly says the crash comes from your Nvidia driver. Report to Nvidia.

[mongoose7]

It's not clear at all; the stack is corrupt. I would say it is not a Nvidia problem; the driver is written in C. Maybe you have a memory problem.

[CuteAlien]

I tried to reproduce it, but doesn't crash here. Sorry, not seen that specific problem before.

[atomaki]

I've checked RAM with memtest86 : no errors found.

[CuteAlien]

I suspect it's probably some kind of linker trouble causing this. Something wrong libs something...
Check if it also happens if you don't use any Irrlicht code at all. Or don't even link to Irrlicht.

[mongoose7]

I've checked RAM with memtest86 : no errors found.

Colloquially, a memory error is an error relating to allocating memory in a program. Errors occur where you a) free the same memory twice, b) free memory that was never allocated, or c) overwrite memory. Buffer overuns are the most common cause of memory issues.

Of course, it may also be a linking problem. C++ is pretty fragile regarding linking of dynamic objects (DLLs).

[CuteAlien]

It crashes before main... I don't think it's about de-allocating memory twice.

[atomaki]

If I'm not using any Irrlicht facilities, application runs OK (without segfaults)

Code: Select all

 
#include <stdexcept>
#include <exception>
#include "irrlicht.h"
 
int main() {
    throw std::runtime_error("err");
}
 

Code: Select all

 
$ g++ -g -O0 -Wall -Wextra -std=c++11  -o main main2.cc \
         -lIrrlicht -lGL -lXxf86vm -lXext -lX11 -lXcursor 
 

 

Code: Select all

 
$ ./main 
terminate called after throwing an instance of 'std::runtime_error'
  what():  err
Aborted (core dumped)
 

[CuteAlien]

Very confusing. Are you using an Irrlicht from a distributor? In that case maybe download it fresh from our website. And then start by compiling the examples (just call ./buildAllExamples.sh in the examples folder). If those run, then try adding the runtime_error stuff to one of them.

[atomaki]

Irrlicht downloaded from our website:
http://irrlicht.sourceforge.net/?page_id=10

version (1.8.3)
I've compiled and run examples without any errors.

[CuteAlien]

Running out of ideas. Maybe check with ldd if it really links to the libs you expect it to link to (ldd yourappname). You can also copy/paste the results in here in case anything looks suspicious to one of us.

[atomaki]

Code: Select all

 
$ ldd main
    linux-gate.so.1 =>  (0xb7727000)
    libGL.so.1 => /usr/lib/nvidia-304/libGL.so.1 (0xb7628000)
    libXxf86vm.so.1 => /usr/lib/i386-linux-gnu/libXxf86vm.so.1 (0xb7622000)
    libX11.so.6 => /usr/lib/i386-linux-gnu/libX11.so.6 (0xb74ed000)
    libstdc++.so.6 => /usr/lib/i386-linux-gnu/libstdc++.so.6 (0xb7405000)
    libm.so.6 => /lib/i386-linux-gnu/libm.so.6 (0xb73bf000)
    libgcc_s.so.1 => /lib/i386-linux-gnu/libgcc_s.so.1 (0xb73a2000)
    libc.so.6 => /lib/i386-linux-gnu/libc.so.6 (0xb71f4000)
    libnvidia-tls.so.304.131 => /usr/lib/nvidia-304/tls/libnvidia-tls.so.304.131 (0xb71ef000)
    libnvidia-glcore.so.304.131 => /usr/lib/nvidia-304/libnvidia-glcore.so.304.131 (0xb54e9000)
    libXext.so.6 => /usr/lib/i386-linux-gnu/libXext.so.6 (0xb54d6000)
    libdl.so.2 => /lib/i386-linux-gnu/libdl.so.2 (0xb54d1000)
    libxcb.so.1 => /usr/lib/i386-linux-gnu/libxcb.so.1 (0xb54af000)
    /lib/ld-linux.so.2 (0xb7728000)
    libXau.so.6 => /usr/lib/i386-linux-gnu/libXau.so.6 (0xb54aa000)
    libXdmcp.so.6 => /usr/lib/i386-linux-gnu/libXdmcp.so.6 (0xb54a3000)
 

[CuteAlien]

Is this a 32-bit system? Maybe I should try reproducing on that (got to admit I haven't tested on 32-bit in a rather long time...).

[hendu]

For giggles, link your standalone program that works with -lGL and see if it starts to fail.