I have a program that throws an uncaught exception somewhere. All I get is a report of an exception being thrown, and no information as to where it was thrown. It seems illogical for a program compiled to contain debug symbols not to notify me of where in my code an exception was generated.
Is there any way to tell where my exceptions are coming from short of setting 'catch throw' in gdb and calling a backtrace for every single thrown exception?
If an exception is uncaught, the special library function std::terminate() is automatically called. Terminate is actually a pointer to a function and default value is the Standard C library function std::abort(). If no cleanups occur for an uncaught exception†, it may actually be helpful in debugging this problem as no destructors are called.
†It is implementation-defined whether or not the stack is unwound before std::terminate() is called.
A call to abort() is often useful in generating a core dump that can be analyzed to determine the cause of the exception. Make sure that you enable core dumps via ulimit -c unlimited (Linux).
You can install your own terminate() function by using std::set_terminate(). You should be able to set a breakpoint on your terminate function in gdb. You may be able to generate a stack backtrace from your terminate() function and this backtrace may help in identifying the location of the exception.
There is a brief discussion on uncaught exceptions in Bruce Eckel's Thinking in C++, 2nd Ed that may be helpful as well.
Since terminate() calls abort() by default (which will cause a SIGABRT signal by default), you may be able to set a SIGABRT handler and then print a stack backtrace from within the signal handler. This backtrace may help in identifying the location of the exception.
Note: I say may because C++ supports non-local error handling through the use of language constructs to separate error handling and reporting code from ordinary code. The catch block can be, and often is, located in a different function/method than the point of throwing. It has also been pointed out to me in the comments (thanks Dan) that it is implementation-defined whether or not the stack is unwound before terminate() is called.
Update: I threw together a Linux test program called that generates a backtrace in a terminate() function set via set_terminate() and another in a signal handler for SIGABRT. Both backtraces correctly show the location of the unhandled exception.
Update 2: Thanks to a blog post on Catching uncaught exceptions within terminate, I learned a few new tricks; including the re-throwing of the uncaught exception within the terminate handler. It is important to note that the empty throw statement within the custom terminate handler works with GCC and is not a portable solution.
Code:
#ifndef _GNU_SOURCE
#define _GNU_SOURCE
#endif
#ifndef __USE_GNU
#define __USE_GNU
#endif
#include <execinfo.h>
#include <signal.h>
#include <string.h>
#include <iostream>
#include <cstdlib>
#include <stdexcept>
void my_terminate(void);
namespace {
// invoke set_terminate as part of global constant initialization
static const bool SET_TERMINATE = std::set_terminate(my_terminate);
}
// This structure mirrors the one found in /usr/include/asm/ucontext.h
typedef struct _sig_ucontext {
unsigned long uc_flags;
struct ucontext *uc_link;
stack_t uc_stack;
struct sigcontext uc_mcontext;
sigset_t uc_sigmask;
} sig_ucontext_t;
void crit_err_hdlr(int sig_num, siginfo_t * info, void * ucontext) {
sig_ucontext_t * uc = (sig_ucontext_t *)ucontext;
// Get the address at the time the signal was raised from the EIP (x86)
void * caller_address = (void *) uc->uc_mcontext.eip;
std::cerr << "signal " << sig_num
<< " (" << strsignal(sig_num) << "), address is "
<< info->si_addr << " from "
<< caller_address << std::endl;
void * array[50];
int size = backtrace(array, 50);
std::cerr << __FUNCTION__ << " backtrace returned "
<< size << " frames\n\n";
// overwrite sigaction with caller's address
array[1] = caller_address;
char ** messages = backtrace_symbols(array, size);
// skip first stack frame (points here)
for (int i = 1; i < size && messages != NULL; ++i) {
std::cerr << "[bt]: (" << i << ") " << messages[i] << std::endl;
}
std::cerr << std::endl;
free(messages);
exit(EXIT_FAILURE);
}
void my_terminate() {
static bool tried_throw = false;
try {
// try once to re-throw currently active exception
if (!tried_throw++) throw;
}
catch (const std::exception &e) {
std::cerr << __FUNCTION__ << " caught unhandled exception. what(): "
<< e.what() << std::endl;
}
catch (...) {
std::cerr << __FUNCTION__ << " caught unknown/unhandled exception."
<< std::endl;
}
void * array[50];
int size = backtrace(array, 50);
std::cerr << __FUNCTION__ << " backtrace returned "
<< size << " frames\n\n";
char ** messages = backtrace_symbols(array, size);
for (int i = 0; i < size && messages != NULL; ++i) {
std::cerr << "[bt]: (" << i << ") " << messages[i] << std::endl;
}
std::cerr << std::endl;
free(messages);
abort();
}
int throw_exception() {
// throw an unhandled runtime error
throw std::runtime_error("RUNTIME ERROR!");
return 0;
}
int foo2() {
throw_exception();
return 0;
}
int foo1() {
foo2();
return 0;
}
int main(int argc, char ** argv) {
struct sigaction sigact;
sigact.sa_sigaction = crit_err_hdlr;
sigact.sa_flags = SA_RESTART | SA_SIGINFO;
if (sigaction(SIGABRT, &sigact, (struct sigaction *)NULL) != 0) {
std::cerr << "error setting handler for signal " << SIGABRT
<< " (" << strsignal(SIGABRT) << ")\n";
exit(EXIT_FAILURE);
}
foo1();
exit(EXIT_SUCCESS);
}
Output:
my_terminate caught unhanded exception. what(): RUNTIME ERROR! my_terminate backtrace returned 10 frames [bt]: (0) ./test(my_terminate__Fv+0x1a) [0x8048e52] [bt]: (1) /usr/lib/libstdc++-libc6.2-2.so.3 [0x40045baa] [bt]: (2) /usr/lib/libstdc++-libc6.2-2.so.3 [0x400468e5] [bt]: (3) /usr/lib/libstdc++-libc6.2-2.so.3(__rethrow+0xaf) [0x40046bdf] [bt]: (4) ./test(throw_exception__Fv+0x68) [0x8049008] [bt]: (5) ./test(foo2__Fv+0xb) [0x8049043] [bt]: (6) ./test(foo1__Fv+0xb) [0x8049057] [bt]: (7) ./test(main+0xc1) [0x8049121] [bt]: (8) ./test(__libc_start_main+0x95) [0x42017589] [bt]: (9) ./test(__eh_alloc+0x3d) [0x8048b21] signal 6 (Aborted), address is 0x1239 from 0x42029331 crit_err_hdlr backtrace returned 13 frames [bt]: (1) ./test(kill+0x11) [0x42029331] [bt]: (2) ./test(abort+0x16e) [0x4202a8c2] [bt]: (3) ./test [0x8048f9f] [bt]: (4) /usr/lib/libstdc++-libc6.2-2.so.3 [0x40045baa] [bt]: (5) /usr/lib/libstdc++-libc6.2-2.so.3 [0x400468e5] [bt]: (6) /usr/lib/libstdc++-libc6.2-2.so.3(__rethrow+0xaf) [0x40046bdf] [bt]: (7) ./test(throw_exception__Fv+0x68) [0x8049008] [bt]: (8) ./test(foo2__Fv+0xb) [0x8049043] [bt]: (9) ./test(foo1__Fv+0xb) [0x8049057] [bt]: (10) ./test(main+0xc1) [0x8049121] [bt]: (11) ./test(__libc_start_main+0x95) [0x42017589] [bt]: (12) ./test(__eh_alloc+0x3d) [0x8048b21]
As you say, we can use 'catch throw' in gdb and call 'backtrace' for every single thrown exception. While that's usually too tedious to do manually, gdb allows automation of the process. That allows seeing the backtrace of all exceptions that are thrown, including the last uncaught one:
gdb>
set pagination off
catch throw
commands
backtrace
continue
end
run
Without further manual intervention, this generates lots of backtraces, including one for the last uncaught exception:
Catchpoint 1 (exception thrown), 0x00a30 in __cxa_throw () from libstdc++.so.6
#0 0x0da30 in __cxa_throw () from /usr/.../libstdc++.so.6
#1 0x021f2 in std::__throw_bad_weak_ptr () at .../shared_ptr_base.h:76
[...]
terminate called after throwing an instance of 'std::bad_weak_ptr'
what(): bad_weak_ptr
Program received signal SIGABRT, Aborted.
Here's a great blog post wrapping this up: http://741mhz.com/throw-stacktrace [on archive.org]
You can create a macro like:
#define THROW(exceptionClass, message) throw exceptionClass(__FILE__, __LINE__, (message) )
...and it will give you the location where the exception is thrown (admittedly not the stack trace). It's necessary for you to derive your exceptions from some base class that takes the above constructor.
You can mark main tight places in your code as noexcept to locate an exception, then use libunwind (just add -lunwind to linker parameters) (tested with clang++ 3.6):
demagle.hpp:
#pragma once
char const *
get_demangled_name(char const * const symbol) noexcept;
demangle.cpp:
#include "demangle.hpp"
#include <memory>
#include <cstdlib>
#include <cxxabi.h>
namespace
{
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wglobal-constructors"
#pragma clang diagnostic ignored "-Wexit-time-destructors"
std::unique_ptr< char, decltype(std::free) & > demangled_name{nullptr, std::free};
#pragma clang diagnostic pop
}
char const *
get_demangled_name(char const * const symbol) noexcept
{
if (!symbol) {
return "<null>";
}
int status = -4;
demangled_name.reset(abi::__cxa_demangle(symbol, demangled_name.release(), nullptr, &status));
return ((status == 0) ? demangled_name.get() : symbol);
}
backtrace.hpp:
#pragma once
#include <ostream>
void
backtrace(std::ostream & _out) noexcept;
backtrace.cpp:
#include "backtrace.hpp"
#include <iostream>
#include <iomanip>
#include <limits>
#include <ostream>
#include <cstdint>
#define UNW_LOCAL_ONLY
#include <libunwind.h>
namespace
{
void
print_reg(std::ostream & _out, unw_word_t reg) noexcept
{
constexpr std::size_t address_width = std::numeric_limits< std::uintptr_t >::digits / 4;
_out << "0x" << std::setfill('0') << std::setw(address_width) << reg;
}
char symbol[1024];
}
void
backtrace(std::ostream & _out) noexcept
{
unw_cursor_t cursor;
unw_context_t context;
unw_getcontext(&context);
unw_init_local(&cursor, &context);
_out << std::hex << std::uppercase;
while (0 < unw_step(&cursor)) {
unw_word_t ip = 0;
unw_get_reg(&cursor, UNW_REG_IP, &ip);
if (ip == 0) {
break;
}
unw_word_t sp = 0;
unw_get_reg(&cursor, UNW_REG_SP, &sp);
print_reg(_out, ip);
_out << ": (SP:";
print_reg(_out, sp);
_out << ") ";
unw_word_t offset = 0;
if (unw_get_proc_name(&cursor, symbol, sizeof(symbol), &offset) == 0) {
_out << "(" << get_demangled_name(symbol) << " + 0x" << offset << ")\n\n";
} else {
_out << "-- error: unable to obtain symbol name for this frame\n\n";
}
}
_out << std::flush;
}
backtrace_on_terminate.hpp:
#include "demangle.hpp"
#include "backtrace.hpp"
#include <iostream>
#include <type_traits>
#include <exception>
#include <memory>
#include <typeinfo>
#include <cstdlib>
#include <cxxabi.h>
namespace
{
[[noreturn]]
void
backtrace_on_terminate() noexcept;
static_assert(std::is_same< std::terminate_handler, decltype(&backtrace_on_terminate) >{});
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wglobal-constructors"
#pragma clang diagnostic ignored "-Wexit-time-destructors"
std::unique_ptr< std::remove_pointer_t< std::terminate_handler >, decltype(std::set_terminate) & > terminate_handler{std::set_terminate(backtrace_on_terminate), std::set_terminate};
#pragma clang diagnostic pop
[[noreturn]]
void
backtrace_on_terminate() noexcept
{
std::set_terminate(terminate_handler.release()); // to avoid infinite looping if any
backtrace(std::clog);
if (std::exception_ptr ep = std::current_exception()) {
try {
std::rethrow_exception(ep);
} catch (std::exception const & e) {
std::clog << "backtrace: unhandled exception std::exception:what(): " << e.what() << std::endl;
} catch (...) {
if (std::type_info * et = abi::__cxa_current_exception_type()) {
std::clog << "backtrace: unhandled exception type: " << get_demangled_name(et->name()) << std::endl;
} else {
std::clog << "backtrace: unhandled unknown exception" << std::endl;
}
}
}
std::_Exit(EXIT_FAILURE); // change to desired return code
}
}
There is good article concerning the issue.
You did not pass information about what OS / Compiler you use.
In Visual Studio C++ Exceptions can be instrumented.
See "Visual C++ Exception-Handling Instrumentation" on ddj.com
My article "Postmortem Debugging", also on ddj.com includes code to use Win32 structured exception handling (used by the instrumentation) for logging etc.
I've got code to do this in Windows/Visual Studio, let me know if you want an outline. Don't know how to do it for dwarf2 code though, a quick google suggests that there's a function _Unwind_Backtrace in libgcc that probably is part of what you need.
Check this thread, perhaps it helps:
Catching all unhandled C++ exceptions?
I made good experiences with that software:
http://www.codeproject.com/KB/applications/blackbox.aspx
It can print out a stack trace to a file for any unhandled exception.
