Using error injection

Here's a very manual way of running the error injection in ACCEPT applications.

1. You will need to switch to the errorinjection branch on both the accept and accept-apps repositories. Build ACCEPT as normal. Use activate.sh or otherwise ensure that you have ACCEPT's bin directory on your $PATH.

2. Error injection requires a liberror.cpp to define the error model. We'll use the accept-apps/perfect/sar/kernels/bp application as an example, since it has one checked into the repo already.

3. Go to the application directory and type make build_orig. This will produce an accept_config.txt (but you don't need to meddle with that).

4. (You can optionally write make run_orig to see the output without error injection.)

5. Write an inject_config.txt file to set up the parameters. For example:

    $ cat inject_config.txt 
    default 1
    cmult 2
   

   List the function name (mangled, if it's C++) and the parameter. You can also choose a default for other instructions.

6. Run inject_config.py. This will transfer your parameters from inject_config.txt to ACCEPT's accept_config.txt.

7. Type make run_opt. This will compile, instrument, and execute the error-injected version of the application.