Currently we filter files that contain command line arguments out.
- I do not think its that common, out of 390 files only 47 had them
- Costs outweight the benefits right now but would be cool in the future
They have been filtered out if they matched one of:
´´´´
r"ACCEPT\s+.\s+FROM\s+COMMAND-LINE",
r"ACCEPT\s+.\s+FROM\s+ARGUMENT-NUMBER",
r"ACCEPT\s+.*\s+FROM\s+ARGUMENT-VALUE",
´´´
The problem is that sometimes a command line argument might be the name of a file. e.g. input.txt
but could also be a float 12.21 number.
- if it is a number of other normal input we just need to catch it, store it in metadata.json and then pass it when we execute the script
- If is is a file we also need to generate sample data for that file and then construct it when we execute the programms
How do detect what it is?
It is possible for sure, probably by asking the llm what that is and so on.
Currently we filter files that contain command line arguments out.
They have been filtered out if they matched one of:
´´´´
r"ACCEPT\s+.\s+FROM\s+COMMAND-LINE",
r"ACCEPT\s+.\s+FROM\s+ARGUMENT-NUMBER",
r"ACCEPT\s+.*\s+FROM\s+ARGUMENT-VALUE",
´´´
The problem is that sometimes a command line argument might be the name of a file. e.g. input.txt
but could also be a float 12.21 number.
How do detect what it is?
It is possible for sure, probably by asking the llm what that is and so on.