pdb wrapper with remote access via tcp socket
rpdb - remote debugger based on pdb
rpdb is a wrapper around pdb that re-routes stdin and stdout to a socket handler. By default it opens the debugger on port 4444:
import rpdb; rpdb.set_trace()
But you can change that by simply instantiating Rpdb manually:
import rpdb debugger = rpdb.Rpdb(port=12345) debugger.set_trace()
It is known to work on Jython 2.5 to 2.7, Python 2.5 to 3.1. It was written originally for Jython since this is pretty much the only way to debug it when running it on Tomcat.
set_trace(), your script will “hang” and the only way to get it
to continue is to access rpdb using telnet, netcat, etc..:
nc 127.0.0.1 4444
Installation in CPython (standard Python)
pip install rpdb
For a quick, ad hoc alternative, you can copy the entire rpdb subdirectory (the directory directly containing the __init__.py file) to somewhere on your $PYTHONPATH.
Installation in a Tomcat webapp
Just copy the rpdb directory (the one with the __init__.py file) in your WEB-INF/lib/Lib folder along with the standard Jython library (required).
Trigger rpdb with signal
set_trace() can be triggered at any time by using the TRAP signal handler.
This allows you to debug a running process independantly of a specific failure
import rpdb rpdb.handle_trap() # As with set_trace, you can optionally specify addr and port rpdb.handle_trap("0.0.0.0", 54321)
handle_trap will overwrite the existing handler for SIGTRAP if one has
already been defined in your application.
- The socket is not always closed properly so you will need to ^C in netcat and Esc+q in telnet to exit after a continue or quit.
- There is a bug in Jython 2.5/pdb that causes rpdb to stop on ghost breakpoints after you continue (‘c’), this is fixed in 2.7b1.
- Give access to attributes of stdin and stdout (by @fuxpavel).
- Add rpdb.post_mortem(), similar to pdb.post_mortem() (by @CamDavidsonPilon).
- Write addr/port to stderr instead of stdout (thanks to @onlynone).
- Allow for dynamic host port (thanks to @onlynone).
- Make q/quit do proper cleanup (@kenmanheimer)
- Benignly disregard repeated rpdb.set_trace() to same port as currently active session (@kenmanheimer)
- Extend backwards compatibility down to Python 2.5 (@kenmanheimer)
- Expose the addr, port arguments to the set_trace method (thanks to @niedbalski).
- Remove a try/finally that seemed to shift the trace location (thanks to k4ml@github).
- Catch IOError raised by print in initialization, it may not work in some environments (e.g. mod_wsgi). (Menno Smits)