Copyright (c) 2010-2012 Ben Bass <firstname.lastname@example.org>
All rights reserved.
rpcpdb is a wrapper around the Python pdb debugger which makes it more suitable for use in RPC contexts.
It is designed to fulfil the need to debug a function on an already-running server which uses threads or processes to dispatch each remote procedure call, without having to change the source code to the server in any way. There is no disruption to other clients and rpc calls while the selected call is being debugged.
In particular, a mixin class is provided which adds the debug_func and undebug_func methods to your RPC server. These allow breakpoints to be controlled by another RPC client.
For an example, run the xmlrpc_server.py server process, then run one or more xmlrpc_client.py processes which will continually perform RPC requests against it. xmlrpc_debug.py can then be run to inject a debug breakpoint in a method which the clients are continually calling; the next client to call that function will be remotely debuggable, while other clients carry on oblivious.
An alternative example added in v0.2 is the 'test_server.py' which does not use RPC but runs functions in background threads. It gives a good overview of usage in a single module without extra dependencies.
Currently the debug interface is provided via a UNIX socket, but this will be extended in future.
RPC framework support is intended to cover XMLRPC, RPyC and Pyro in the initial stages.
- Update examples, tidy up, document.
- Support other interfaces than termsock / UNIX socket.
- v0.2 - Add trigger criteria which check incoming arguments so the user can ensure the right client triggers the breakpoint
See the file "LICENSE" for information terms & conditions for usage, and a DISCLAIMER OF ALL WARRANTIES.
- add .hgignore file
- add updb creation without specifying a socket name. One will be created 'at random' and the name available through the debug_socket read-only property.
- Added ignore_count functionality for skipping a matched breakpoint n times
- No functional changes; updates to Trove classifiers only
- Python 3 support (3.2+ only)
- parameter matching to trigger debugger (conditional debugging)
- fix issue where select call in termsock was continually finding writable FDs, causing 100% CPU usage
- first alpha release.