process.argv[0] contains full node.exe path on Windows, docs say it must be just "node"

[Dmitry-Me]

• Version: 4.4.6.0
• Platform: Windows 7 Enterprise SP1
• Subsystem: no idea

Documentation https://nodejs.org/api/process.html#process_process_argv says process.argv[0] must be node but I run this code on Windows

console.log('argv[0]: ' + process.argv[0]);

and it outputs this

argv[0]: C:\Program Files (x86)\nodejs\node.exe

which kind of isn't the same as node. It was node on some earlier versions - perhaps a couple years ago.

Here's how I run my code: I started cmd.exe, cdd into C:\Program Files (x86)\nodejs, so it's now the current path and I run

node fullPathToJsFile

[Fishrock123]

./node -p 'process.argv[0]'
/Users/Jeremiah/Documents/node/out/Release/node

The docs are arguably not entirely correct. It will be the node binary, wherever that is. (Including the path.)

[Fishrock123]

More specifically, how you invoked the node binary. node ... will have node but ./ will append the path to the local node.

[Dmitry-Me]

I started cmd, cdd into C:\Program Files (x86)\nodejs, so it's now the current path. I run

 node fullPathToJsFile

[addaleax]

There’s this line which will always set process.argv[0] to process.execPath, so usually it won’t matter how node was invoked.

[tarungarg546]

I am up for taking this up.
What needs to done exactly modifying docs?

[cjihrig]

The process.argv docs should be updated to say that the first element will be process.execPath. It would probably be good to link there too.

[Dmitry-Me]

@cjihrig Why did the behavior change since perhaps two years ago? I remember clearly that a while ago it was node, not full path.

[addaleax]

#1194 was the PR that changed this for Windows.

[cjihrig]

I honestly don't recall.

[tarungarg546]

#7449

[Dmitry-Me]

@addaleax @cjihrig Okay, may it be that changes from #1194 should be further adapted so that old behavior is preserved? Backward compatibility, you know...

[cjihrig]

At this point, the change has been in a wild long enough that backward compatibility would be hard to achieve.

[Dmitry-Me]

Reviewed by four contributors. That's impressive.

[addaleax]

@Dmitry-Me I’m not a hundred percent sure what you’re trying to say. Documenting the current behaviour is definitely the right thing to do, independently of whether a change to the actual logic should be made or not.

The reviews concerned the correctness of the docs change and I wouldn’t take them to make any statement on what the behaviour should be.

I won’t keep you from reopening this issue and making convincing points as to why the behaviour should be changed back, but keep a few things in mind:

• The original change in src: don't error at startup when cwd doesn't exist #1194 brought parity between Windows and the rest of the world; something that is quite valuable for a cross-platform runtime like Node.
• That change happened over a year ago, and people may have come to rely on the current behaviour.
• Checking process.argv[0] against a specific value to test for a Node-like environment is not very reliable, or even a very common way to check it.

[Dmitry-Me]

@addaleax I'm suprised to see so many contributors signing off a commit. It usually takes up to two contributors (elsewhere). I won't make any conclusions.

[addaleax]

@Dmitry-Me I think it’s just that there is a tendency for small changes that are easy to review to gain more reviewers, that’s all. :)