CRASH (f86115ad python) core/unix/module.c:101 assert
Created by: varun729
Describe the bug Running DynamoRIO with python application that uses Tensorflow results in a crash. DynamoRIO can successfully run with a trivial helloworld python program, and even with a python application that uses PyTorch.
The goal is to use dynamorio with drcachesim to collect memory trace, but it crashes. This is similar to the issue #2929 and the suggestion there was to try debugging without any DynamoRIO client. So, providing details here from the debug run.
To Reproduce Steps to reproduce the behavior:
- Run the python 3.7 with tensorflow 1.15 installed in Anaconda environment.
- Precise command line for running the application.
drrun -debug -- python -c "import tensorflow as tf"
- Exact output or incorrect behavior. debug_drio_python_tensorflow.txt
Please also answer these questions:
- What happens when you run without any client? Crashes
- What happens when you run with debug build ("-debug" flag to drrun/drconfig/drinject)? Throws assert at core/unix/module.c:101
Expected behavior Program should run to completion when run without any client.
Versions
- What version of DynamoRIO are you using? f86115ad
- Does the latest build from https://github.com/DynamoRIO/dynamorio/releases solve the problem? using the latest build
- What operating system version are you running on? Ubuntu 18.04, Linux 5.4.0-72-generic
- Is your application 32-bit or 64-bit? 64-bit