Any child processes launched by that child process will also be wrapped in a similar fashion.
This is a bit of a brutal hack, designed primarily to support code
coverage reporting in cases where tests or the system under test are
loaded via child processes rather than via
It can also be handy if you want to run your own mock executable instead of some other thing when child procs call into it.
var wrap =// wrap(wrapperArgs, environs)var unwrap =// later to undo the wrapping, you can call the returned function
In this example, the
/path/to/my/main.js file will be used as the
"main" module, whenever any Node or io.js child process is started,
whether via a call to
exec, whether node is invoked
directly as the command or as the result of a shebang
/path/to/my/main.js, you can do whatever instrumentation or
environment manipulation you like. When you're done, and ready to run
the "real" main.js file (ie, the one that was spawned in the first
place), you can do this:
// /path/to/my/main.js// process.argv === 'foo=bar'// and process.env.FOO === '1'// my wrapping manipulationsprocess// now run the instrumented and covered or whatever codes
Spawn-wrap responds to two environment variables, both of which are preserved through child processes.
SPAWN_WRAP_DEBUG=1 in the environment will make this module dump a
lot of information to stderr.
SPAWN_WRAP_SHIM_ROOT can be set to a path on the filesystem where
the shim files are written in a
.node-spawn-wrap-<id> folder. By
default this is done in
$HOME, but in some environments you may wish
to point it at some other root. (For example, if
$HOME is mounted
as read-only in a virtual machine or container.)
CONTRACTS and CAVEATS
The initial wrap call uses synchronous I/O. Probably you should not be using this script in any production environments anyway.
Also, this will slow down child process execution by a lot, since we're adding a few layers of indirection.
The contract which this library aims to uphold is:
- Wrapped processes behave identical to their unwrapped counterparts for all intents and purposes. That means that the wrapper script propagates all signals and exit codes.
- If you send a signal to the wrapper, the child gets the signal.
- If the child exits with a numeric status code, then the wrapper exits with that code.
- If the child dies with a signal, then the wrapper dies with the same signal.
- If you execute any Node child process, in any of the various ways that such a thing can be done, it will be wrapped.
- Children of wrapped processes are also wrapped.
(Much of this made possible by foreground-child.)
There are a few ways situations in which this contract cannot be adhered to, despite best efforts:
- In order to handle cases where
nodeis invoked in a shell script, the
PATHenvironment variable is modified such that the the shim will be run before the "real" node. However, since Windows does not allow executing shebang scripts like regular programs, a
node.cmdfile is required.
- Signal propagation through
dashdoesn't always work. So, if you use
child_process.exec()on systems where
dash, then the process may exit with a status code > 128 rather than indicating that it received a signal.
cmd.exeis even stranger with how it propagates and interprets unix signals. If you want your programs to be portable, then probably you wanna not rely on signals too much.
- It is possible to escape the wrapping, if you spawn a bash
script, and that script modifies the
PATH, and then calls a specific