TracePoint.allow_reentry

Generally, while a TracePoint callback is running, other registered callbacks are not called to avoid confusion from reentrance. This method allows reentrance within a given block. Use this method carefully to avoid infinite callback invocation.

If called when reentrance is already allowed, it raises a RuntimeError.

Example:

# Without reentry
# ---------------

line_handler = TracePoint.new(:line) do |tp|
  next if tp.path != __FILE__ # Only works in this file
  puts "Line handler"
  binding.eval("class C; end")
end.enable

class_handler = TracePoint.new(:class) do |tp|
  puts "Class handler"
end.enable

class B
end

# This script will print "Class handler" only once: when inside the :line
# handler, all other handlers are ignored.

# With reentry
# ------------

line_handler = TracePoint.new(:line) do |tp|
  next if tp.path != __FILE__ # Only works in this file
  next if (__LINE__..__LINE__+3).cover?(tp.lineno) # Prevent infinite calls
  puts "Line handler"
  TracePoint.allow_reentry { binding.eval("class C; end") }
end.enable

class_handler = TracePoint.new(:class) do |tp|
  puts "Class handler"
end.enable

class B
end

# This will print "Class handler" twice: inside the allow_reentry block in the :line
# handler, other handlers are enabled.

Note that the example shows the principal effect of the method, but its practical usage is for debugging libraries that sometimes require other libraries’ hooks to not be affected by the debugger being inside trace point handling. Precautions should be taken against infinite recursion in this case (note that we needed to filter out calls by itself from the :line handler, otherwise it would call itself infinitely).