Adding New Op with Converter Caused Core Dump

[dreamibor]

Platform:
Ubuntu 18.04 & MacOS 13.0 Ventura

So basically, I want to add a new op to MNN and I got segmentation fault, which doesn't quite make sense to me and it has bothered me for a while.

What I did:

1. added 'SliceOnnx' to enum OpType;
2. added 'SliceOnnx' to union OpParameter;
3. added
   table SliceOnnx { starts:[int]; ends:[int]; axes:[int]; steps:[int]; }
   to CaffeOps.fbs
4. remove OnnxSlice.cpp under tools/converter/source/optimizer/onnxextra
5. added SliceOnnx.cpp under tools/converter/source/onnx with:

//
//  SliceOnnx.cpp
//  MNNConverter
//
//  Created by MNN on 2019/07/17.
//  Copyright © 2018, Alibaba Group Holding Limited
//

#include "onnxOpConverter.hpp"

DECLARE_OP_CONVERTER(SliceOnnx);

MNN::OpType SliceOnnx::opType() {
    return MNN::OpType_SliceOnnx;
}

MNN::OpParameter SliceOnnx::type() {
    return MNN::OpParameter_SliceOnnx;
}

void SliceOnnx::run(MNN::OpT *dstOp, const onnx::NodeProto *onnxNode,
                      OnnxScope* scope) {
    return;
}

REGISTER_CONVERTER(SliceOnnx, Slice);

The compilation works fine but when I call the MNNConvert, it gives me the following error messge with LLDB:

(lldb) settings set -- target.run-args  "-f" "ONNX" "--modelFile" "minimum_slice/minimum_slice.onnx" "--MNNModel" "minimum_slice/minimum_slice.mnn" "--bizCode" "MNN"
(lldb) r
Process 50470 launched: '/Users/xxx/Develop/MNN/build/MNNConvert' (arm64)
hw.cpufamily: 458787763 , size = 4
The device support dot:1, support fp16:1, support i8mm: 0
Start to Convert Other Model Format To MNN Model...
[22:51:01] /Users/xxx/Develop/MNN/tools/converter/source/onnx/onnxConverter.cpp:40: ONNX Model ir version: 8
Start to Optimize the MNN Net...
Process 50470 stopped
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x0)
    frame #0: 0x0000000100192ff8 libMNN_Express.dylib`MNN::CreateSliceOnnx(flatbuffers::FlatBufferBuilder&, MNN::SliceOnnxT const*, std::__1::function<unsigned long long (void*)> const*) + 24
libMNN_Express.dylib`MNN::CreateSliceOnnx:
->  0x100192ff8 <+24>: ldp    x8, x9, [x1]
    0x100192ffc <+28>: subs   x10, x9, x8
    0x100193000 <+32>: b.eq   0x10019308c               ; <+172>
    0x100193004 <+36>: asr    x2, x10, #2
Target 0: (MNNConvert) stopped.
(lldb) bt
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x0)
  * frame #0: 0x0000000100192ff8 libMNN_Express.dylib`MNN::CreateSliceOnnx(flatbuffers::FlatBufferBuilder&, MNN::SliceOnnxT const*, std::__1::function<unsigned long long (void*)> const*) + 24
    frame #1: 0x000000010018f150 libMNN_Express.dylib`MNN::CreateOp(flatbuffers::FlatBufferBuilder&, MNN::OpT const*, std::__1::function<unsigned long long (void*)> const*) + 92
    frame #2: 0x0000000100188b7c libMNN_Express.dylib`MNN::Express::Expr::create(MNN::OpT const*, std::__1::vector<MNN::Express::VARP, std::__1::allocator<MNN::Express::VARP> >, int) + 544
    frame #3: 0x0000000100ae7b98 libMNNConvertDeps.dylib`MNN::Express::Program::createUnit(std::__1::map<int, MNN::Express::VARP, std::__1::less<int>, std::__1::allocator<std::__1::pair<int const, MNN::Express::VARP> > >&, std::__1::vector<int, std::__1::allocator<int> >&, std::__1::vector<std::__1::unique_ptr<MNN::OpT, std::__1::default_delete<MNN::OpT> >, std::__1::allocator<std::__1::unique_ptr<MNN::OpT, std::__1::default_delete<MNN::OpT> > > > const&, MNN::OpT*, std::__1::vector<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >, std::__1::allocator<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > > > const&, std::__1::set<MNN::OpT*, std::__1::less<MNN::OpT*>, std::__1::allocator<MNN::OpT*> >&, std::__1::set<int, std::__1::less<int>, std::__1::allocator<int> >&) + 2252
    frame #4: 0x0000000100ae8af0 libMNNConvertDeps.dylib`MNN::Express::Program::create(std::__1::vector<std::__1::unique_ptr<MNN::OpT, std::__1::default_delete<MNN::OpT> >, std::__1::allocator<std::__1::unique_ptr<MNN::OpT, std::__1::default_delete<MNN::OpT> > > > const&, std::__1::vector<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >, std::__1::allocator<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > > > const&, std::__1::vector<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >, std::__1::allocator<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > > > const&, bool, bool) + 156
    frame #5: 0x0000000100ae0248 libMNNConvertDeps.dylib`MNN::Express::RunExtraPass(std::__1::unique_ptr<MNN::NetT, std::__1::default_delete<MNN::NetT> >&, std::__1::unordered_map<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >, MNN::Express::VARP, std::__1::hash<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > >, std::__1::equal_to<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > >, std::__1::allocator<std::__1::pair<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const, MNN::Express::VARP> > > const&) + 60
    frame #6: 0x0000000100ae0dd8 libMNNConvertDeps.dylib`MNN::Express::optimizeNetImpl(std::__1::unique_ptr<MNN::NetT, std::__1::default_delete<MNN::NetT> >&, std::__1::unordered_map<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >, MNN::Express::VARP, std::__1::hash<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > >, std::__1::equal_to<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > >, std::__1::allocator<std::__1::pair<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const, MNN::Express::VARP> > > const&) + 1096
    frame #7: 0x0000000100ae35bc libMNNConvertDeps.dylib`optimizeNet(std::__1::unique_ptr<MNN::NetT, std::__1::default_delete<MNN::NetT> >&, bool, modelConfig&) + 660
    frame #8: 0x000000010099a5a0 libMNNConvertDeps.dylib`MNN::Cli::convertModel(modelConfig&) + 3000
    frame #9: 0x0000000100003e6c MNNConvert`main + 196
    frame #10: 0x000000018906fe50 dyld`start + 2544

This is so weird since the error message shows the error happens at creating the new OP, and the function CreateSliceOnnx() is generated under schema/current/CaffeOp_generated.h

Any ideas on this? Thanks in advance.