refactor: split integration tests from CLI (part 1) (#22308)

This PR separates integration tests from CLI tests into a new project
named `cli_tests`. This is a prerequisite for an integration test runner
that can work with either the CLI binary in the current project, or one
that is built ahead of time.

## Background

Rust does not have the concept of artifact dependencies yet
(https://github.com/rust-lang/cargo/issues/9096). Because of this, the
only way we can ensure a binary is built before running associated tests
is by hanging tests off the crate with the binary itself.

Unfortunately this means that to run those tests, you _must_ build the
binary and in the case of the deno executable that might be a 10 minute
wait in release mode.

## Implementation

To allow for tests to run with and without the requirement that the
binary is up-to-date, we split the integration tests into a project of
their own. As these tests would not require the binary to build itself
before being run as-is, we add a stub integration `[[test]]` target in
the `cli` project that invokes these tests using `cargo test`.

The stub test runner we add has `harness = false` so that we can get
access to a `main` function. This `main` function's sole job is to
`execvp` the command `cargo test -p deno_cli`, effectively "calling"
another cargo target.

This ensures that the deno executable is always correctly rebuilt before
running the stub test runner from `cli`, and gets us closer to be able
to run the entire integration test suite on arbitrary deno executables
(and therefore split the build into multiple phases).

The new `cli_tests` project lives within `cli` to avoid a large PR. In
later PRs, the test data will be split from the `cli` project. As there
are a few thousand files, it'll be better to do this as a completely
separate PR to avoid noise.

2 files changed, 72 insertions, 0 deletions

diff --git a/test_util/src/lib.rs b/test_util/src/lib.rs
index 9bf68cbc2..dda62a591 100644
--- a/test_util/src/lib.rs
+++ b/test_util/src/lib.rs
@@ -32,6 +32,7 @@ pub mod lsp;
 mod npm;
 pub mod pty;
 pub mod servers;
+pub mod spawn;
 
 pub use builders::DenoChild;
 pub use builders::TestCommandBuilder;
diff --git a/test_util/src/spawn.rs b/test_util/src/spawn.rs
new file mode 100644
index 000000000..bfd83e9b2
--- /dev/null
+++ b/test_util/src/spawn.rs
@@ -0,0 +1,71 @@
+// Copyright 2018-2024 the Deno authors. All rights reserved. MIT license.
+use anyhow::Error;
+use std::convert::Infallible;
+
+/// For unix targets, we just replace our current process with the desired cargo process.
+#[cfg(unix)]
+pub fn exec_replace_inner(
+  cmd: &str,
+  args: &[&str],
+) -> Result<Infallible, Error> {
+  use std::ffi::CStr;
+  use std::ffi::CString;
+
+  let args = args
+    .iter()
+    .map(|arg| CString::new(*arg).unwrap())
+    .collect::<Vec<_>>();
+  let args: Vec<&CStr> =
+    args.iter().map(|arg| arg.as_ref()).collect::<Vec<_>>();
+
+  let err = nix::unistd::execvp(&CString::new(cmd).unwrap(), &args)
+    .expect_err("Impossible");
+  Err(err.into())
+}
+
+#[cfg(windows)]
+pub fn exec_replace_inner(
+  cmd: &str,
+  args: &[&str],
+) -> Result<Infallible, Error> {
+  use std::os::windows::io::AsRawHandle;
+  use std::process::Command;
+  use win32job::ExtendedLimitInfo;
+  use win32job::Job;
+
+  // Use a job to ensure the child process's lifetime does not exceed the current process's lifetime.
+  // This ensures that if the current process is terminated (e.g., via ctrl+c or task manager),
+  // the child process is automatically reaped.
+
+  // For more information about this technique, see Raymond Chen's blog post:
+  // https://devblogs.microsoft.com/oldnewthing/20131209-00/?p=2433
+  // Note: While our implementation is not perfect, it serves its purpose for test code.
+
+  // In the future, we may directly obtain the main thread's handle from Rust code and use it
+  // to create a suspended process that we can then resume:
+  // https://github.com/rust-lang/rust/issues/96723
+
+  // Creates a child process and assigns it to our current job.
+  // A more reliable approach would be to create the child suspended and then assign it to the job.
+  // For now, we create the child, create the job, and then assign both us and the child to the job.
+  let mut child = Command::new(cmd).args(&args[1..]).spawn()?;
+
+  let mut info = ExtendedLimitInfo::default();
+  info.limit_kill_on_job_close();
+  let job = Job::create_with_limit_info(&info)?;
+  job.assign_current_process()?;
+  let handle = child.as_raw_handle();
+  job.assign_process(handle as _)?;
+
+  let exit = child.wait()?;
+  std::process::exit(exit.code().unwrap_or(1));
+}
+
+/// Runs a command, replacing the current process on Unix. On Windows, this function blocks and
+/// exits.
+///
+/// In either case, the only way this function returns is if it fails to launch the child
+/// process.
+pub fn exec_replace(command: &str, args: &[&str]) -> Result<Infallible, Error> {
+  exec_replace_inner(command, args)
+}