[Attn: gcc maintainers] Re: unhandled C++ exceptions not propagating
Ken Brown
kbrown@cornell.edu
Mon Aug 17 17:36:24 GMT 2020
On 8/16/2020 6:38 PM, Ken Brown via Cygwin wrote:
> On 8/16/2020 10:56 AM, Ken Brown via Cygwin wrote:
>> On 8/16/2020 9:52 AM, Takashi Yano via Cygwin wrote:
>>> On Sun, 16 Aug 2020 09:21:24 -0400
>>> Ken Brown via Cygwin <cygwin@cygwin.com> wrote:
>>>> On 8/15/2020 11:53 PM, Takashi Yano via Cygwin wrote:
>>>>> On Sat, 15 Aug 2020 20:38:01 -0300
>>>>> David McFarland via Cygwin <cygwin@cygwin.com> wrote:
>>>>>>
>>>>>> I was just debugging a c++ app (b2 build system from boost), and noticed
>>>>>> that it would appear to exit unexpectedly without an error. This turned
>>>>>> out to be when an unhandled C++ exception was thrown.
>>>>>>
>>>>>> On a fresh install of cygwin with gcc-g++, this program will throw an
>>>>>> exception from the std::string constructor:
>>>>>>
>>>>>> c++ -x c++ - <<END
>>>>>> #include <string>
>>>>>> int main() { std::string str(nullptr); return 0; }
>>>>>> END
>>>>>>
>>>>>> When it's executed from the shell it returns zero, but execution stops
>>>>>> at the exception.
>>>>>>
>>>>>> $ ./a; echo $?
>>>>>> 0
>>>>>>
>>>>>> When executed under gdb, the exception is caught, and the process exits
>>>>>> non-zero when continued:
>>>>>>
>>>>>> (gdb) r
>>>>>> Starting program: /home/corngood/a
>>>>>> [New Thread 4300.0x1390]
>>>>>> [New Thread 4300.0x1d24]
>>>>>> [New Thread 4300.0x1d48]
>>>>>> [New Thread 4300.0x80c]
>>>>>> gdb: unknown target exception 0x20474343 at 0x7ff8d2cfa799
>>>>>> Thread 1 "a" received signal ?, Unknown signal.
>>>>>> 0x00007ff8d2cfa799 in RaiseException () from
>>>>>> /cygdrive/c/WINDOWS/System32/KERNELBASE.dll
>>>>>> (gdb) c
>>>>>> Continuing.
>>>>>> [Thread 4300.0x1d24 exited with code 541541187]
>>>>>> [Thread 4300.0x1870 exited with code 541541187]
>>>>>> [Thread 4300.0x1d48 exited with code 541541187]
>>>>>> [Thread 4300.0x1390 exited with code 541541187]
>>>>>> [Inferior 1 (process 4300) exited with code 04021641503]
>>>>>> (gdb)
>>>>>>
>>>>>> When executed under strace, it exits with an error as expected:
>>>>>>
>>>>>> $ strace -o /dev/null a; echo $?
>>>>>> 67
>>>>>>
>>>>>> That's as far as I've investigated so far.
>>>>>
>>>>> Is this the same issue with
>>>>> https://cygwin.com/pipermail/cygwin/2019-October/242795.html ?
>>>>>
>>>>> As far as I tested, this does not occur in 32-bit cygwin.
>>>>
>>>> This doesn't seem to be a new issue. I found the following report from 6 years
>>>> ago, on Cygwin 1.7.30 with GCC 4.9.0:
>>>>
>>>> https://stackoverflow.com/questions/24402412/program-executed-on-cygwin-does-not-report-a-thrown-exception
>>>>
>>>
>>> Thanks for the information. But I still wonder why 32-bit and 64-bit
>>> cygwin behave differently.
>>
>> I'm no C++ expert, but my understanding after a brief internet search is that
>> an uncaught exception is supposed to cause std::terminate to be called. So I
>> installed gcc-debuginfo and ran the OP's test case under gdb with a breakpoint
>> at 'terminate'. On 32-bit Cygwin I got the expected behavior:
>>
>> Thread 1 "exception_test" hit Breakpoint 1, std::terminate ()
>> at /usr/src/debug/gcc-9.3.0-2/libstdc++-v3/libsupc++/eh_terminate.cc:96
>> 96 __atomic_load (&__terminate_handler, &func, __ATOMIC_ACQUIRE);
>> (gdb) bt
>> #0 std::terminate ()
>> at /usr/src/debug/gcc-9.3.0-2/libstdc++-v3/libsupc++/eh_terminate.cc:96
>> #1 0x57c45909 in __cxxabiv1::__cxa_throw (obj=obj@entry=0x800396e8,
>> tinfo=tinfo@entry=0x57c646a8 <typeinfo for std::logic_error>,
>> dest=0x57be6c60 <std::logic_error::~logic_error()>)
>> at /usr/src/debug/gcc-9.3.0-2/libstdc++-v3/libsupc++/eh_throw.cc:95
>> #2 0x57c49223 in std::__throw_logic_error (
>> __s=__s@entry=0x57c60160 <cplus_demangle_builtin_types+10272>
>> "basic_string::
>> _S_construct null not valid")
>> at /usr/src/debug/gcc-9.3.0-2/libstdc++-v3/src/c++11/functexcept.cc:66
>> [...]
>>
>> On 64-bit Cygwin, however, I got the behavior reported by the OP:
>>
>> gdb: unknown target exception 0x20474343 at 0x7ff8cccca719
>>
>> Thread 1 "exception_test" received signal ?, Unknown signal.
>> 0x00007ff8cccca719 in RaiseException () from /c/WINDOWS/System32/KERNELBASE.dll
>> (gdb) c
>> Continuing.
>> [Thread 4508.0x3e20 exited with code 541541187]
>> [Thread 4508.0x5bc exited with code 541541187]
>> [Thread 4508.0x5ee8 exited with code 541541187]
>> [Thread 4508.0x1cd0 exited with code 541541187]
>> [Inferior 1 (process 4508) exited with code 04021641503]
>>
>> Note that the breakpoint at 'terminate' is not reached. I tried putting a
>> breakpoint at __cxa_throw and stepping through from there, but I don't really
>> understand what's going on. Someone who knows the innards of exception
>> handling will have to look at this.
>
> One further comment, and then I'll shut up
I lied.
> and leave it to people who know what
> they're talking about: In the 64-bit case, the function _Unwind_RaiseException
> in libgcc/unwind-seh.c gets called, and it calls the Win32 function
> RaiseException as follows:
>
> RaiseException (STATUS_GCC_THROW, 0, 1, (ULONG_PTR *)&exc);
>
> Then the following comment appears:
>
> /* The exception handler installed in crt0 will continue any GCC
> exception that reaches there (and isn't marked non-continuable).
> Returning allows the C++ runtime to call std::terminate. */
>
> Apparently this doesn't actually happen.
I'm beginning to think this is a bug in the Cygwin build of gcc. The file
unwind-seh.c is compiled on 64-bit Cygwin because __SEH__ is defined. But it
makes assumptions, such as those in the comment quoted above, that are not valid
on Cygwin.
In fact, if you look at the main exception handler on Cygwin
(exceptions.cc:626), it is clear that the exception code STATUS_GCC_THROW does
not lead to continuation. It falls under the default case:
default:
/* If we don't recognize the exception, we have to assume that
we are doing structured exception handling, and we let
something else handle it. */
return ExceptionContinueSearch;
[STATUS_GCC_THROW is 0x20474343, which shows up in the gdb output above as an
unknown target exception.]
So it seems to me that either Cygwin's exception handler has to learn to handle
the exception codes defined in unwind-seh.c, or else gcc should be built with
__SEH__ undefined.
JonY, Achim, Corinna, am I missing something?
Ken
More information about the Cygwin
mailing list