Run-state follow-ups: startup-event fix, stoppable Watcher, C API run state, AgentApp tests#11
Merged
Conversation
…er sends register_event(event_type::startup) detached a thread capturing raw 'this' that slept 500 ms before publishing: any agent destroyed within that window was a use-after-free (AgentApp::connect() takes this path whenever events are enabled). The thread is now a member, waits on a condition variable with the per-agent stop request as predicate, and is woken and joined by shutdown()/disconnect() while the sockets are still open. Since that thread publishes concurrently with the owner thread and ZMQ sockets are not thread-safe, sends on the publisher socket are now serialized with a mutex (a pre-existing race for any agent publishing within 500 ms of connecting).
Watcher::watch() looped on while(true) with no way to end it: the broker leaked its settings watcher in a detached thread and tests had to intentionally leak Watchers. watch() now runs while neither stop() nor a process-wide stop (Mads::Runtime::stop_process(), e.g. SIGINT) has been requested, and all platform waits are bounded (~1 s) so a stop is noticed promptly. The broker now owns its watcher thread and joins it during shutdown in both daemon and interactive modes.
New C functions completing the Mads::Runtime story for C and Python agents: agent_stop() stops one agent's loops, agent_running() reports whether an agent should keep going (typical loop: while(agent_running(a)) receive), mads_stop_process()/mads_process_running() handle the process-wide level. Backed by a new public Agent::running() accessor (the condition loop() itself checks). The Python ctypes wrapper gains Agent.stop(), the Agent.running property, and module-level stop_process()/process_running().
New unit suite for AgentAppT: CLI option plumbing (settings/name/id/ queue-size/crypto), the standard early-exit options (--help, --version, --save-settings against an in-process fake broker and the local-settings error path), apply_receive_timeout/apply_queue_size, set_agent_name, restart_if_requested's no-restart path, and the events-enabled connect/disconnect flow over loopback. The usage example now shows how to stop the loop via the agent's Runtime and points at RUNTIME.md. src/agent_app.hpp leaves the gcovr/codecov ignore lists: 78% covered, with the remainder being std::exit/execvp paths and the UDP discovery branch. Project coverage: 83.6% (215 tests).
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## v2 #11 +/- ##
=======================================
+ Coverage 52.5% 53.1% +0.6%
=======================================
Files 10 11 +1
Lines 1830 2081 +251
Branches 847 943 +96
=======================================
+ Hits 961 1107 +146
- Misses 324 342 +18
- Partials 545 632 +87 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Follow-ups to the
Mads::Runtimerefactor merged in #10, completing the run-state story for v2.4.0. Five commits:a43a564fix: startup-event use-after-free.register_event(event_type::startup)published from a detached thread holding a rawthisthrough a 500 ms delay; any agent destroyed within that window (e.g. a short-lived events-enabledAgentApp) was a use-after-free. The thread is now owned by the agent, wakes early on shutdown/disconnect via a condition variable, and is joined while the sockets are still open. Publisher sends are also serialized with a mutex, closing a pre-existing data race between the delayed event thread and the owner thread.3f86a3bfeat: stoppableWatcher.Watcher::watch()can now be ended withstop()or by a process-wide stop request; all platform waits are bounded (~1 s). The broker joins its settings-watcher thread on shutdown in both daemon and interactive modes instead of leaking it.b585fbcfeat: run state in the C API and Python. Newagent_stop(),agent_running(),mads_stop_process(),mads_process_running()(backed by a new publicAgent::running()accessor — the exact conditionloop()checks), plusAgent.stop(), theAgent.runningproperty, and module-levelstop_process()/process_running()in the ctypes wrapper. C and Python agents can now drive and end receive loops cleanly instead of relying on signals.340ba90test: AgentApp coverage. 11 new test cases forAgentAppT(CLI plumbing,--help/--version/--save-settingsagainst an in-process fake broker, crypto option propagation, events-enabled connect/disconnect over loopback); the usage example now shows stopping the loop via the agent's Runtime and points at RUNTIME.md;src/agent_app.hppleaves the coverage ignore lists at 78% covered.ad2677fdocs: CHANGES.md entries under the unreleased v2.4.0 and new RUNTIME.md sections for C/Python agents and the Watcher.Verification
agent_app.hpp).ad2677fis green end to end, including the Codecov upload.MADS_BUILD_APPS=ON, MongoDB, tools) compiles cleanly — 272 targets including the modified broker and plugin loaders.🤖 Generated with Claude Code
https://claude.ai/code/session_01QM4yHBNnLAp1fyoeSLc4C1
Generated by Claude Code