编写涉及Signal Trap与Process.kill的测试用例仅在CI环境失败，如何解决？

问题背景 #

我为一个进程关闭监听模块编写测试用例，本地环境运行正常，但在GitHub CI中执行测试时失败。推测原因是测试中使用的Process.kill指令直接终止了CI作业，日志中可见'terminated'相关记录。使用版本：

• ruby 2.7.8
• rails 7.0.8.4

实现代码 #

module ProcessShutdownListener
  def initialize(...)
    setup_graceful_shutdown
    super
  end

  def shutdown_requested?
    @_shutdown_requested
  end

  private

  def setup_graceful_shutdown
    @_shutdown_requested = false
    Signal.trap("TERM") do
      graceful_shutdown
    end

    Signal.trap("INT") do
      graceful_shutdown
    end
  end

  def raise_error_on_process_shutdown?
    false
  end

  def graceful_shutdown
    @_shutdown_requested = true

    if raise_error_on_process_shutdown?
      raise ShutdownRequested
    end
  end
end

测试用例代码 #

class BasicExample
  include ProcessShutdownListener

  def raise_error_on_process_shutdown?
    false # default
  end
end

class AutomaticRaiseErrorExample
  include ProcessShutdownListener

  def raise_error_on_process_shutdown?
    true
  end
end

describe ProcessShutdownListener do

  describe "responding to process shutdown signals" do
    context "when #raise_error_on_process_shutdown? is false" do
      subject! { BasicExample.new }

      describe "#shutdown_requested?" do
        context "when process receives the TERM Signal" do
          it "returns true" do
            Process.kill("TERM", 0)
            expect(subject.shutdown_requested?).to be true
          end
        end

        context "when process receives the INT Signal" do
          it "returns true" do
            Process.kill("INT", 0)
            expect(subject.shutdown_requested?).to be true
          end
        end
      end
    end

    context "when #raise_error_on_process_shutdown? is true" do
      subject! { AutomaticRaiseErrorExample.new }

      context "when process receives the TERM Signal" do
        it "raises an error" do
          expect { Process.kill("TERM", Process.pid) }.to raise_error(ShutdownRequested)
        end
      end
    end
  end
end

解决方案 #

GitHub CI环境中，测试进程收到TERM/INT信号后会直接终止整个作业，这是导致测试失败的核心原因。以下是几种可行的测试方案：

1. 用子进程隔离信号影响 #

通过fork子进程执行信号测试逻辑，避免信号影响到父进程（CI测试进程）：

context "when process receives the TERM Signal" do
  it "returns true" do
    read, write = IO.pipe

    pid = fork do
      read.close
      subject = BasicExample.new
      Process.kill("TERM", Process.pid)
      write.puts(subject.shutdown_requested?.to_s)
      exit
    end

    write.close
    result = read.gets.chomp == "true"
    Process.wait(pid)
    expect(result).to be true
  end
end

2. 直接模拟信号触发逻辑 #

不发送真实信号，而是直接调用信号处理的逻辑块：

context "when process receives the TERM Signal" do
  it "returns true" do
    # 获取TERM信号的处理Proc
    term_trap = Signal.trap("TERM", nil)
    Signal.trap("TERM", term_trap) # 恢复原信号处理
    term_trap.call # 直接触发处理逻辑

    expect(subject.shutdown_requested?).to be true
  end
end

3. 直接调用核心处理方法 #

对于需要验证异常抛出的场景，可直接调用graceful_shutdown方法，绕开发送信号的步骤：

context "when process receives the TERM Signal" do
  it "raises an error" do
    expect { subject.graceful_shutdown }.to raise_error(ShutdownRequested)
  end
end

内容的提问来源于stack exchange，提问作者queroga_vqz