os/signal: TestCtrlBreak failures

[gopherbot]

#!watchflakes
default <- pkg == "os/signal" && test == "TestCtrlBreak"

Issue created automatically to collect these failures.

Example (log):

=== RUN   TestCtrlBreak
    signal_windows_test.go:96: Program exited with error: exit status 1
        2026/06/16 13:42:19 Timeout waiting for Ctrl+Break
--- FAIL: TestCtrlBreak (5.07s)

— watchflakes

[gopherbot]

Found new dashboard test flakes for:

#!watchflakes
default <- pkg == "os/signal" && test == "TestCtrlBreak"

2026-06-16 20:11 go1.26-windows-386 release-branch.go1.26@fe91cea4 os/signal.TestCtrlBreak (log)

=== RUN   TestCtrlBreak
    signal_windows_test.go:96: Program exited with error: exit status 1
        2026/06/16 13:42:19 Timeout waiting for Ctrl+Break
--- FAIL: TestCtrlBreak (5.07s)

— watchflakes

[dmitshur]

CC @golang/runtime, @golang/windows.

[dmitshur]

From triage meeting, @prattmic points out this test has a time.Sleep(1 * time.Second) so this might be a flake due to the subprocess not starting in that constant time.

[dagood]

Looks like the typical fix (e.g. os/signal.TestLibraryCtrlHandler) is to have the test program write "ready" to stdout and use that as the indicator, rather than rely on timing. I'll submit a CL.

[gopherbot]

Change https://go.dev/cl/792020 mentions this issue: os/signal: avoid flaky TestCtrlBreak by waiting for child stdout