node:test logs verbose feature

[mtsluna]

What is the problem this feature will solve?

This feature solve the problem to turn on and turn off the logs (info/warn/error/log/etc) in the testing runner (node 18)

What is the feature you are proposing to solve the problem?

Add a simple flag in the node commands that allow to turn on the logs in combination of the next command:
node --test --verbose with the testing runner in node 18.

What alternatives have you considered?

No response

[MoLow]

Hi @mtsluna what type of logs do you refer to? do you have a code example to describe how such a log might look, or maybe a reference to a test runner (i.e mocha, jest,tap) that supports this?

[ErickWendel]

I think this could be kind of an interesting feature. Although, I'd suggest adding a --silent to hide logs as it wouldn't have to be a breaking change. What do you think?

cc @benjamingr

[MoLow]

@ErickWendel what logs would --silent mute?

[ErickWendel]

@ErickWendel what logs would --silent mute?

when we run node --tests we see the node tap output right? the --silent would be used on CI and actually omit all logs so if everything is going right it shouldn't show anything and exit with code 0

[mtsluna]

Hello @MoLow and @ErickWendel when you run jest test and you put the flag --silent=false, you can see all the logs in all levels in the test console. In this case in the node:test api you cant enable the logs, you only can view the results of the execution.

[benjamingr]

@nodejs/test_runner @cjihrig

[cjihrig]

I don't hate the idea of a quiet mode, but I do have two thoughts:

1. If we figure out reporters, we could theoretically have a "quiet" reporter instead of this other mode of operation.
2. Since the test runner is part of Node core, we need to balance adding functionality with not adding a ton of new CLI flags to core, and I'm not sure this is worth adding a new flag to core.

[MoLow]

the --silent would be used on CI and actually omit all logs so if everything is going right it shouldn't show anything and exit with code 0

the tap output is still useful in the CI in case of a failure - users are usually interested at the specific failing test details

if we figure out reporters, we could theoretically have a "quiet" reporter instead of this other mode of operation.

+1 on that, also reporting to a file instead of stdout will be much easier and might solve this usecase

Since the test runner is part of Node core, we need to balance adding functionality with not adding a ton of new CLI flags to core, and I'm not sure this is worth adding a new flag to core.

this is probably out of scope - but we might want to introduce some kind of a .testrc file - for allowing configuration of the test runner (what reporters to use, concurrency etc), can be cleaner than adding lots of flags

[khaosdoctor]

this is probably out of scope - but we might want to introduce some kind of a .testrc file - for allowing configuration of the test runner (what reporters to use, concurrency etc), can be cleaner than adding lots of flags

+1 on that, it would greatly help sharing configuration outside of the project, although I see #44241 has already been implemented

---

Another thought on that would be to check the NODE_ENV for whether you're running the app on development (as those are quite common) if you're running in test, production, or anything other than development then console.logs would be suppressed, otherwise, they'd be shown.

[cjihrig]

I am strongly -1 to using NODE_ENV for anything in the test runner.

[MoLow]

I am waiting to see how #43973 evolves to figure out how to globally configure node,
until then #44241 is a perfect way to achieve quite mode

using NODE_ENV sounds like a hack to me

[khaosdoctor]

I am strongly -1 to using NODE_ENV for anything in the test runner.

Any particular reason? Just curious on the drawbacks that I might not have thought 😅

[cjihrig]

A few reasons:

• Despite popularity, NODE_ENV is an anti-pattern, especially when the default value is not production. I've seen multiple apps running in production without setting NODE_ENV properly, so their performance was not as good as it could/should have been.
• This is a test runner, so I'd expect NODE_ENV to be 'test' or 'development' all the time.

[khaosdoctor]

I see! Thanks for the explanation.

[himself65]

I don't see any usage of NODE_ENV in the nodejs source code. Only in npm and third-party libraries.
So -1 on this, because I think nodejs runtime shouldn't rely on any environment variables, this will make users hard to debug

[himself65]

Not only console.log, I'm using util.debugLog, but it also cannot display to the console or somewhere else

[Qix-]

@himself65 to be fair, debugLog already relies on NODE_DEBUG (akin to debug).

[cjihrig]

I'm going to close this. Support for reporters just landed in a1b27b2, so you can customize the output how you'd like.