scapy/CONTRIBUTING.md

How to contribute

Contributors are essential to Scapy (as they are to most open source projects). Here is some advice to help you help the project!

Project objectives

We try to keep Scapy as powerful as possible, to support as many protocols and platforms as possible, to keep and make the code (and the commit history) as clean as possible.

Since Scapy can be slow and memory consuming, we try to limit CPU and memory usage, particularly in parts of the code often called.

What to contribute

You want to spend time working on Scapy but have no (or little) idea what to do? You can look for open issues labeled "contributions wanted", or look at the contributions roadmap

If you have any ideas of useful contributions that you cannot (or do not want to) do yourself, open an issue and include "contributions wanted" in the title.

Once you have chosen a contribution, open an issue to let other people know you're working on it (or assign the existing issue to yourself) and track your progress. You might want to ask whether you're working in an appropriate direction, to avoid the frustration of seeing your contribution rejected after a lot of work.

Reporting issues

Bugs

If you have installed Scapy through a package manager (from your Linux or BSD system, from PyPI, etc.), please get and install the current development code, and check that the bug still exists before submitting an issue.

If you're not sure whether a behavior is a bug or not, submit an issue and ask, don't be shy!

Enhancements / feature requests

If you want a feature in Scapy, but cannot implement it yourself or want some hints on how to do that, open an issue and include "enhancement" in the title.

Explain if possible the API you would like to have (e.g., give examples of function calls, packet creations, etc.).

Submitting pull requests

Coding style & conventions

• The code should be PEP-8 compliant; you can check your code with pep8 and the command tox -e flake8

• Pylint can help you write good Python code (even if respecting Pylint rules is sometimes either too hard or even undesirable; human brain needed!).

• Google Python Style Guide is a nice read!

• Avoid creating unnecessary list objects, particularly if they can be huge (e.g., when possible, use for line in fdesc instead of for line in fdesc.readlines(); more generally prefer generators over lists).

Tests

Please consider adding tests for your new features or that trigger the bug you are fixing. This will prevent a regression from being unnoticed. Do not use the variable _ in your tests, as it could break them.

If you find yourself in a situation where your tests locally succeed but fail if executed on the CI, try to enable the debugging option for the dissector by setting conf.debug_dissector = 1.

New protocols

New protocols can go either in scapy/layers or to scapy/contrib. Protocols in scapy/layers should be usually found on common networks, while protocols in scapy/contrib should be uncommon or specific.

To be precise, scapy/layers protocols should not be importing scapy/contrib protocols, whereas scapy/contrib protocols may import both scapy/contrib and scapy/layers protocols.

The detailed requirements are explained in Design patterns on Scapy's doc.

Features

Protocol-related features should be implemented within the same module as the protocol layers(s) (e.g., traceroute() is implemented in scapy/layers/inet.py).

Other features may be implemented in a module (scapy/modules) or a contribution (scapy/contrib).

Core

If you contribute to Scapy's core (e.g., scapy/base_classes.py, scapy/packet.py, etc.), please be very careful with performances and memory footprint, as it is easy to write Python code that wastes memory or CPU cycles.

As an example, Packet().__init__() is called each time a layer is parsed from a string (during a network capture or a PCAP file read). Adding inefficient code here will have a disastrous effect on Scapy's performances.

Logging

Scapy has an internal logging system based on logging.

In the past, Scapy was generally too verbose on packet dissection, leading many new users to disable all logs, which makes it harder for them to find real issues afterwards. You should comply with these guidelines to make sure logging in Scapy remains helpful.

• If you want the log message to only be displayed when using Scapy through the interactive console, use scapy.error.log_interactive. You are free to use any log level.
• Otherwise, always use scapy.error.log_runtime.
  • On packet dissection, of packet layers you should remain AT OR BELOW the logging.INFO level, unless the issue is critical or tied to security. For instance: "DNS Decompression loop detected !" is allowed as WARNING, but "Could not dissect packet" or "Invalid value detected" are not.
  • On packet build or any command or function that is called by the user or the root program, you are free and welcomed to use the WARNING or ERROR levels, to signal that a packet was wrongly built for instance.
• If you are working on Scapy's core, you may use: scapy.error.log_loading only while Scapy is loading, to display import errors for instance.

Python 2 and 3 compatibility

The project aims to provide code that works both on Python 2 and Python 3. Therefore, some rules need to be applied to achieve compatibility:

• byte-string must be defined as b"\x00\x01\x02"
• exceptions must comply with the new Python 3 format: except SomeError as e:
• lambdas must be written using a single argument when using tuples: use lambda x, y: x + f(y) instead of lambda (x, y): x + f(y).
• use int instead of long
• use list comprehension instead of map() and filter()
• __bool__ = __nonzero__ must be used when declaring __nonzero__ methods
• __next__ = next must be used when declaring next methods in iterators
• StopIteration must NOT be used in generators (but it can still be used in iterators)
• io.BytesIO must be used instead of StringIO when using bytes
• __cmp__ must not be used.

Code review

Maintainers tend to be picky, and you might feel frustrated that your code (which is perfectly working in your use case) is not merged faster.

Please don't be offended, and keep in mind that maintainers are concerned about code maintainability and readability, commit history (we use the history a lot, for example to find regressions or understand why certain decisions have been made), performances, integration in Scapy, API consistency (so that someone who knows how to use Scapy will know how to use your code), etc.

Thanks for reading, happy hacking!