Fix: Prevent infinite polling when continue_execution called multiple times

[Piotr Tylenda (pitylend)]

Problem

When calling continue_execution while the debugger is already running (not paused at a breakpoint), the waitForStateChange method would poll indefinitely waiting for a state change that would never come. This caused the MCP tool to hang forever.

Reproduction Steps

1. Start debug session with an attach configuration
2. Set a breakpoint
3. Trigger code path to hit breakpoint
4. Call continue_execution - works fine
5. Call continue_execution again - HANGS FOREVER

Root Cause

The early exit condition in waitForStateChange was:

if(beforeState.hasLocationInfo()&&!currentState.hasLocationInfo()&&afterState.sessionActive){returnfalse;// Keep polling}

This only handled the case where we HAD location info before continuing. When calling continue a second time:

• beforeState already had NO location info (process was running)
• The condition beforeState.hasLocationInfo() was false
• So we never entered the early exit logic and polled forever

Solution

Added tracking of consecutive attempts where the process is running without location info:

if(!currentState.hasLocationInfo()&&currentState.sessionActive){runningWithoutLocationAttempts++;if(runningWithoutLocationAttempts>=maxRunningAttempts){returncurrentState;// Stop polling after 3 attempts}}elseif(currentState.hasLocationInfo()){runningWithoutLocationAttempts=0;// Reset when we hit a breakpoint}

This handles both scenarios:

1. Paused → Continue → Running (normal case, returns after a few attempts)
2. Running → Continue → Still Running (second continue, now returns after 3 attempts instead of hanging forever)

Testing Performed

Full debugging session test:

Step	Action	Before Fix	After Fix
1	Start debug session with attach config	✅	✅
2	Set breakpoint on health endpoint	✅	✅
3	Call health endpoint (background)	✅	✅
4	Debugger pauses at breakpoint	✅	✅
5	get_variables_values	✅	✅
6	First continue_execution	✅ Returns promptly	✅ Returns promptly
7	Second continue_execution	❌ HANGS FOREVER	✅ Returns promptly
8	Third continue_execution	N/A (stuck)	✅ Returns promptly
9	stop_debugging	N/A	✅ Session ends cleanly

Backward Compatibility

Fully backward compatible. The fix only affects the edge case where:

• The debugger is already running (not paused)
• continue_execution is called again

Normal debugging workflows (step, continue from breakpoint, etc.) are unaffected.