read_video_from_file() causes seg fault with Python 3.9

[prabhat00155]

🐛 Describe the bug

https://app.circleci.com/pipelines/github/pytorch/vision/10411/workflows/5783f083-26e2-4209-b9c6-976de289d637/jobs/791355

Program received signal SIGSEGV, Segmentation fault.
0x00007f87b3031d22 in ?? () from /root/project/env/lib/libswscale.so.5
Missing separate debuginfos, use: debuginfo-install glibc-2.17-324.el7_9.x86_64
(gdb) bt
#0  0x00007f87b3031d22 in ?? () from /root/project/env/lib/libswscale.so.5
#1  0x00007f87b3030be4 in ?? () from /root/project/env/lib/libswscale.so.5
#2  0x00007f87b3006af4 in sws_scale () from /root/project/env/lib/libswscale.so.5
#3  0x00007f87b4ab1932 in ffmpeg::(anonymous namespace)::transformImage(SwsContext*, unsigned char const* const*, int*, ffmpeg::VideoFormat, ffmpeg::VideoFormat, unsigned char*, unsigned char**, int*) [clone .isra.9] ()
   from /root/project/torchvision/video_reader.so
#4  0x00007f87b4ab1bd9 in ffmpeg::VideoSampler::sample(unsigned char const* const*, int*, ffmpeg::ByteStorage*) () from /root/project/torchvision/video_reader.so
#5  0x00007f87b4ab2ac7 in ffmpeg::VideoStream::copyFrameBytes(ffmpeg::ByteStorage*, bool) () from /root/project/torchvision/video_reader.so
#6  0x00007f87b4aaea9a in ffmpeg::Stream::decodePacket(AVPacket const*, ffmpeg::DecoderOutputMessage*, bool, bool*) () from /root/project/torchvision/video_reader.so
#7  0x00007f87b4aaa183 in ffmpeg::Decoder::processPacket(ffmpeg::Stream*, AVPacket*, bool*, bool*) () from /root/project/torchvision/video_reader.so
#8  0x00007f87b4aaa80a in ffmpeg::Decoder::getFrame(unsigned long) () from /root/project/torchvision/video_reader.so
#9  0x00007f87b4aafe81 in ffmpeg::SyncDecoder::decode(ffmpeg::DecoderOutputMessage*, unsigned long) () from /root/project/torchvision/video_reader.so
#10 0x00007f87b4adbb55 in vision::video_reader::(anonymous namespace)::readVideo(bool, at::Tensor const&, std::string, double, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long) ()
   from /root/project/torchvision/video_reader.so
#11 0x00007f87b4add99a in vision::video_reader::read_video_from_file(std::string, double, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long) () from /root/project/torchvision/video_reader.so
#12 0x00007f87b4ae0d8f in std::decay<c10::guts::infer_function_traits<c10::impl::detail::WrapFunctionIntoRuntimeFunctor_<c10::List<at::Tensor> (*)(std::string, double, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long), c10::List<at::Tensor>, c10::guts::typelist::typelist<std::string, double, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long> > >::type::return_type>::type c10::impl::call_functor_with_args_from_stack_<c10::impl::detail::WrapFunctionIntoRuntimeFunctor_<c10::List<at::Tensor> (*)(std::string, double, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long), c10::List<at::Tensor>, c10::guts::typelist::typelist<std::string, double, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long> >, false, 0ul, 1ul, 2ul, 3ul, 4ul, 5ul, 6ul, 7ul, 8ul, 9ul, 10ul, 11ul, 12ul, 13ul, 14ul, 15ul, 16ul, 17ul, 18ul, std::string, double, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long>(c10::OperatorKernel*, c10::DispatchKeySet, std::vector<c10::IValue, std::allocator<c10::IValue> >*, std::integer_sequence<unsigned long, 0ul, 1ul, 2ul, 3ul, 4ul, 5ul, 6ul, 7ul, 8ul, 9ul, 10ul, 11ul, 12ul, 13ul, 14ul, 15ul, 16ul, 17ul, 18ul>, c10::guts::typelist::typelist<std::string, double, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long>*) () from /root/project/torchvision/video_reader.so
#13 0x00007f87b4ae0e92 in c10::impl::make_boxed_from_unboxed_functor<c10::impl::detail::WrapFunctionIntoRuntimeFunctor_<c10::List<at::Tensor> (*)(std::string, double, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long), c10::List<at::Tensor>, c10::guts::typelist::typelist<std::string, double, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long, long> >, false>::call(c10::OperatorKernel*, c10::OperatorHandle const&, c10::DispatchKeySet, std::vector<c10::IValue, std::allocator<c10::IValue> >*) () from /root/project/torchvision/video_reader.so
#14 0x00007f881b9ef98b in std::_Function_handler<void (std::vector<c10::IValue, std::allocator<c10::IValue> >&), torch::jit::(anonymous namespace)::createOperatorFromC10(c10::OperatorHandle const&)::{lambda(std::vector<c10::IValue, std::allocator<c10::IValue> >&)#1}>::_M_invoke(std::_Any_data const&, std::vector<c10::IValue, std::allocator<c10::IValue> >&) () from /root/project/env/lib/python3.9/site-packages/torch/lib/libtorch_cpu.so
#15 0x00007f88239dbb53 in torch::jit::invokeOperatorFromPython(std::vector<std::shared_ptr<torch::jit::Operator>, std::allocator<std::shared_ptr<torch::jit::Operator> > > const&, pybind11::args, pybind11::kwargs const&) ()
   from /root/project/env/lib/python3.9/site-packages/torch/lib/libtorch_python.so
#16 0x00007f88239af864 in torch::jit::initJITBindings(_object*)::{lambda(std::string const&)#132}::operator()(std::string const&) const::{lambda(pybind11::args, {lambda(std::string const&)#132}::kwargs)#1}::operator()(pybind11, pybind11::args) const ()
   from /root/project/env/lib/python3.9/site-packages/torch/lib/libtorch_python.so
#17 0x00007f88239b002f in void pybind11::cpp_function::initialize<torch::jit::initJITBindings(_object*)::{lambda(std::string const&)#132}::operator()(std::string const&) const::{lambda(pybind11::args, pybind11::kwargs)#1}, pybind11::object, {lambda(std::string const&)#132}, pybind11::args, pybind11::name, pybind11::doc>(torch::jit::initJITBindings(_object*)::{lambda(std::string const&)#132}::operator()(std::string const&) const::{lambda(pybind11::args, pybind11::kwargs)#1}&&, pybind11::object (*)({lambda(std::string const&)#132}, pybind11::args), pybind11::name const&, pybind11::doc const&)::{lambda(pybind11::detail::function_call&)#3}::_FUN(pybind11::detail) () from /root/project/env/lib/python3.9/site-packages/torch/lib/libtorch_python.so
#18 0x00007f882312d63b in pybind11::cpp_function::dispatcher(_object*, _object*, _object*) () from /root/project/env/lib/python3.9/site-packages/torch/lib/libtorch_python.so
#19 0x0000555bedf0d964 in cfunction_call () at /tmp/build/80754af9/python-split_1629314979626/work/Objects/methodobject.c:539
#20 0x0000555bedf07c0f in _PyObject_MakeTpCall () at /tmp/build/80754af9/python-split_1629314979626/work/Objects/call.c:191
#21 0x0000555bedf97789 in _PyObject_VectorcallTstate (kwnames=0x0, nargsf=<optimized out>, args=0x555bf2271418, callable=<optimized out>, tstate=<optimized out>) at /tmp/build/80754af9/python-split_1629314979626/work/Include/cpython/abstract.h:116
#22 PyObject_Vectorcall () at /tmp/build/80754af9/python-split_1629314979626/work/Include/cpython/abstract.h:127
#23 call_function (kwnames=0x0, oparg=<optimized out>, pp_stack=<synthetic pointer>, tstate=0x555bef577810) at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:5072
#24 _PyEval_EvalFrameDefault () at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:3487
#25 0x0000555bedf4d552 in _PyEval_EvalFrame () at /tmp/build/80754af9/python-split_1629314979626/work/Include/internal/pycore_ceval.h:40
#26 _PyEval_EvalCode () at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:4327
#27 0x0000555bedf4e0b7 in _PyFunction_Vectorcall () at /tmp/build/80754af9/python-split_1629314979626/work/Objects/call.c:396
#28 0x0000555bedec6336 in _PyObject_VectorcallTstate (kwnames=0x7f87b4b1dfa0, nargsf=<optimized out>, args=<optimized out>, callable=0x7f87b4b2d550, tstate=<optimized out>)
    at /tmp/build/80754af9/python-split_1629314979626/work/Include/cpython/abstract.h:118
#29 PyObject_Vectorcall () at /tmp/build/80754af9/python-split_1629314979626/work/Include/cpython/abstract.h:127
#30 call_function (kwnames=0x7f87b4b1dfa0, oparg=<optimized out>, pp_stack=<synthetic pointer>, tstate=<optimized out>) at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:5072
#31 _PyEval_EvalFrameDefault (tstate=<optimized out>, f=0x555bf226e6f0, throwflag=<optimized out>) at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:3535
#32 0x0000555bedf4d74b in _PyEval_EvalFrame () at /tmp/build/80754af9/python-split_1629314979626/work/Include/internal/pycore_ceval.h:40
#33 _PyEval_EvalCode () at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:4327
#34 0x0000555bedf4e0b7 in _PyFunction_Vectorcall () at /tmp/build/80754af9/python-split_1629314979626/work/Objects/call.c:396
#35 0x0000555bedec6ff4 in _PyObject_VectorcallTstate (kwnames=0x0, nargsf=<optimized out>, args=0x555bf226e698, callable=0x7f87b4b2d940, tstate=<optimized out>) at /tmp/build/80754af9/python-split_1629314979626/work/Include/cpython/abstract.h:118
#36 PyObject_Vectorcall () at /tmp/build/80754af9/python-split_1629314979626/work/Include/cpython/abstract.h:127
#37 call_function (kwnames=0x0, oparg=<optimized out>, pp_stack=<synthetic pointer>, tstate=0x555bef577810) at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:5072
#38 _PyEval_EvalFrameDefault (tstate=<optimized out>, f=0x555bf226e4a0, throwflag=<optimized out>) at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:3487
#39 0x0000555bedf4d552 in _PyEval_EvalFrame () at /tmp/build/80754af9/python-split_1629314979626/work/Include/internal/pycore_ceval.h:40
#40 _PyEval_EvalCode () at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:4327
#41 0x0000555bedf4e0b7 in _PyFunction_Vectorcall () at /tmp/build/80754af9/python-split_1629314979626/work/Objects/call.c:396
#42 0x0000555bedec6336 in _PyObject_VectorcallTstate (kwnames=0x7f883a6eb4f0, nargsf=<optimized out>, args=<optimized out>, callable=0x7f87b4b2de50, tstate=<optimized out>)
    at /tmp/build/80754af9/python-split_1629314979626/work/Include/cpython/abstract.h:118
#43 PyObject_Vectorcall () at /tmp/build/80754af9/python-split_1629314979626/work/Include/cpython/abstract.h:127
#44 call_function (kwnames=0x7f883a6eb4f0, oparg=<optimized out>, pp_stack=<synthetic pointer>, tstate=<optimized out>) at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:5072
#45 _PyEval_EvalFrameDefault (tstate=<optimized out>, f=0x555bef5cfd80, throwflag=<optimized out>) at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:3535
#46 0x0000555bedf4d552 in _PyEval_EvalFrame () at /tmp/build/80754af9/python-split_1629314979626/work/Include/internal/pycore_ceval.h:40
#47 _PyEval_EvalCode () at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:4327
#48 0x0000555bedffcfcc in _PyEval_EvalCodeWithName (qualname=0x0, name=0x0, closure=0x0, kwdefs=0x0, defcount=0, defs=0x0, kwstep=2, kwcount=0, kwargs=<optimized out>, kwnames=<optimized out>, argcount=<optimized out>, args=<optimized out>, 
    locals=<optimized out>, globals=<optimized out>, _co=<optimized out>) at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:4359
#49 PyEval_EvalCodeEx () at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:4375
#50 0x0000555bedf4e57b in PyEval_EvalCode (co=co@entry=0x7f883a5b8190, globals=globals@entry=0x7f883a684b00, locals=locals@entry=0x7f883a684b00) at /tmp/build/80754af9/python-split_1629314979626/work/Python/ceval.c:826
#51 0x0000555bedffd07b in run_eval_code_obj () at /tmp/build/80754af9/python-split_1629314979626/work/Python/pythonrun.c:1219
#52 0x0000555bee0334d5 in run_mod () at /tmp/build/80754af9/python-split_1629314979626/work/Python/pythonrun.c:1240
#53 0x0000555beded9f34 in pyrun_file (fp=0x555bef5ada10, filename=0x7f883a64d6b0, start=257, globals=0x7f883a684b00, locals=0x7f883a684b00, closeit=1, flags=0x7ffd6d8540f8) at /tmp/build/80754af9/python-split_1629314979626/work/Python/pythonrun.c:1138
#54 0x0000555bee0381cf in pyrun_simple_file (flags=0x7ffd6d8540f8, closeit=1, filename=0x7f883a64d6b0, fp=0x555bef5ada10) at /tmp/build/80754af9/python-split_1629314979626/work/Python/pythonrun.c:449
#55 PyRun_SimpleFileExFlags () at /tmp/build/80754af9/python-split_1629314979626/work/Python/pythonrun.c:482
#56 0x0000555bee0382eb in PyRun_AnyFileExFlags () at /tmp/build/80754af9/python-split_1629314979626/work/Python/pythonrun.c:91
#57 0x0000555bee03891c in pymain_run_file (cf=0x7ffd6d8540f8, config=0x555bef576300) at /tmp/build/80754af9/python-split_1629314979626/work/Modules/main.c:379

[pmeier]

Since we only excluded 3.9 explicitly (which was the latest supported Python version at the time)

vision/setup.py

Line 353 in 13755c9

if sys.platform != "linux" or (sys.version_info.major == 3 and sys.version_info.minor == 9):

3.10 will be build against ffmpeg. So far there are no issues: #7295. So I guess, we can lift at least the version restriction?