process: rename and deprecate process.reallyExit(), optionally expose result.

[eljefedelrodeodeljefe]

• Version: all
• Platform: all
• Subsystem: process

In the wake of #6456 I want to suggest to rename process.reallyExit() as it's name is poorly chosen. It was introduced in f6a7fe2 and imo doesn't fit the intention of the commit - the repo was young back then.

An adequate name would be os.exit() since the exit call is implemented by the OSes differently. Optionally we could think about exposing this via documentation, since it also eases the API inconsistencies related - it would basically say, if you really want to exit, no matter what, call this.

For deprecation purposes we could alias this one in lib/process.js.

For API examples see other modern languages.

It's also doubtful whether the use of exit(3) is correct in that context and _Exit(3) (see here) would have been more appropriate. Discussion would be more academic though, imo.

[its2mc]

Hey.. what is the differnce between this and proces.exit()

[eljefedelrodeodeljefe]

To simplify: the current implementation of process.exit() in JS isn't really exiting, but rather checks for exit codes emits an event and invokes process.reallyExit() (see here), which in turn is calling a cpp function wrapping exit(). cpp Setupnodeinstance() is in turn handling the atexit() of this call. It's messy....

[jasnell]

process.reallyExit() is the undocumented native function exposed from inside node.cc that causes the immediate exit of the Node process. It's called by process.exit() after the exit event is triggered. It's really not intended for direct use by anyone and shouldn't really be on the process object without a _ prefix as it's not something that users likely should be calling themselves.

[eljefedelrodeodeljefe]

Yes. My point is that is very poor style and could be renamed for internal purposes also. It's one of the oldest part of the repo and frankly I think Ryan has implemented this w/o anticipating node's popularity. As described above, it would be worthwhile to expose it also - for clarity, especially in light of #6456

[eljefedelrodeodeljefe]

...poor style as in: as is this the whole style around exiting.

[its2mc]

Ah.. okay.. what are the options for improving it as a function? I suppose the reallyExit() function executes the stack cleanup and all that..

[its2mc]

well to clarify cleanup in general.. not sure whats cleaned up anyway.. :D

[eljefedelrodeodeljefe]

Unfortunately it does not. That is subject to another problem though. As it stands (and as far as my cpp knowledge goes) node anyhow is exiting really abruptly and the OS takes care of cleaning up. The OS is doing a good job at it in general.

[its2mc]

okay.. doesnt that create incosistencies among different OSs though?

[its2mc]

Also.. regarding es6 and proxies, can functions such as reallyExit.. and in general process. functions be interceptable.. basically can I make my own custom node global. :D

[its2mc]

Thanks for all your responses^^

[eljefedelrodeodeljefe]

Yes. Very badly, since exit(3) is highly OS-y. The effects shouldn't be too noticeable, yet what we have in the standard case "I want to exit my application normally or with a specified error" is considered bad practice for not unwinding stacks properly. My bet is that IoT devices have a hard time. It is of course possible to do and/or expose exit(3), but rather in "I am panicking" mode.

Sure you can define the behaviour yourself. In that case it's really easy also.

[its2mc]

Thanks..