Skip to content

Worktrunk is a CLI for Git worktree management, designed for parallel AI agent workflows

fix(test): run --source error test non-interactively to prevent a PTY wedge (#3161)

Maximilian Roos3fb96deb71f9
diff --git a/tests/integration_tests/shell_wrapper.rs b/tests/integration_tests/shell_wrapper.rs
index f2f075c93..e52e491ad 100644
--- a/tests/integration_tests/shell_wrapper.rs
+++ b/tests/integration_tests/shell_wrapper.rs
@@ -339,6 +339,29 @@ fn exec_in_pty_interactive(
working_dir: &std::path::Path,
env_vars: &[(&str, &str)],
inputs: &[&str],
+) -> (String, i32) {
+ exec_in_pty_shell(shell, script, working_dir, env_vars, inputs, true)
+}
+
+/// Like [`exec_in_pty_interactive`] but lets the caller choose whether `bash` /
+/// `zsh` run interactively (`-i`).
+///
+/// An interactive shell does job-control initialization at startup: if it
+/// decides it isn't the PTY's foreground process group it sends its own process
+/// group `SIGTTIN` and stops — producing no output and never exiting. Under
+/// parallel load that occasionally wedged `test_source_flag_forwards_errors`
+/// until the 180s harness timeout. A non-interactive shell skips job control
+/// entirely, so the wedge can't happen; pass `interactive = false` for tests
+/// that don't assert on job-control output. Tests that *do* (e.g.
+/// `test_zsh_no_job_control`) pass `interactive = true`.
+#[cfg(test)]
+fn exec_in_pty_shell(
+ shell: &str,
+ script: &str,
+ working_dir: &std::path::Path,
+ env_vars: &[(&str, &str)],
+ inputs: &[&str],
+ interactive: bool,
) -> (String, i32) {
use portable_pty::CommandBuilder;
use std::io::Write;
@@ -402,14 +425,17 @@ fn exec_in_pty_interactive(
cmd.env("USER", "testuser");
cmd.env("SHELL", shell_binary);
- // Run in interactive mode to simulate real user environment.
- // This ensures tests catch job control message leaks like "[1] 12345" and "[1]+ Done".
- // Interactive shells have job control enabled by default.
+ // Interactive shells (`-i`) enable job control, so tests can catch
+ // job-control message leaks like "[1] 12345" and "[1]+ Done". Callers that
+ // don't assert on job control pass `interactive = false` to skip `-i` and
+ // avoid the startup job-control wedge (see `exec_in_pty_shell`).
match shell {
"zsh" => {
// Isolate from user rc files
cmd.env("ZDOTDIR", "/dev/null");
- cmd.arg("-i");
+ if interactive {
+ cmd.arg("-i");
+ }
cmd.arg("--no-rcs");
cmd.arg("-o");
cmd.arg("NO_GLOBAL_RCS");
@@ -419,7 +445,9 @@ fn exec_in_pty_interactive(
cmd.arg(script);
}
"bash" => {
- cmd.arg("-i");
+ if interactive {
+ cmd.arg("-i");
+ }
cmd.arg("-c");
cmd.arg(script);
}
@@ -1622,8 +1650,17 @@ approved-commands = ["echo 'fish background task'"]
("WORKTRUNK_TEST_EPOCH", "1735776000"),
];
- let (combined, exit_code) =
- exec_in_pty_interactive(shell, &final_script, &worktrunk_source, &env_vars, &[]);
+ // Run non-interactively: this test only checks that the `--source`
+ // branch forwards wt's error, never job-control output. Skipping `-i`
+ // avoids the interactive-shell startup wedge that flaked this test.
+ let (combined, exit_code) = exec_in_pty_shell(
+ shell,
+ &final_script,
+ &worktrunk_source,
+ &env_vars,
+ &[],
+ false,
+ );
let output = ShellOutput {
combined,
exit_code,