2.8.1: ExecJS.compile consistently exits 1 in one environment and 0 elsewhere. about execjs

[jkeam]

Looks like this might be the issue. Read on 🎉

This transpiled JS doesn't work (https://gist.githubusercontent.com/jrafanie/96475c6f34a55269e1add042bded92b0/raw/919de4331b9befa5ff3a1a22479b2a792d6b9a90/source_map_path_2.8.1.js)

And this one does (https://gist.githubusercontent.com/jkeam/141bb30cc0fdb5c42e21fd15c41b908d/raw/84ac502ad1bd3af7eb2bd42fbbc0b07bce78dda2/test_write.js)

The only difference between the two is that in the working version, we commented out this line:

delete this.setTimeout;

What it looks like based on our debugging session is that DynaTrace is trying to call the setTimeout method, but in the broken version that method does not exist and an error is thrown and the script returns a non-zero exit status code.

While not exactly the same issue, @radikaled found this issue #64 which has similar odd behavior of DynaTrace.

[Fryguy]

Note that we believed this was because the failing environment was FIPS enabled, but we've tried on other FIPS enabled envs and we can't duplicate it.

[Fryguy]

As @jkeam said one thing we noticed was DynaTrace was installed on the openshift cluster - we are trying to evaluate whether that is actually what causes the issue, but it is looking likely suspect. We see the following stack trace which includes dynatrace, and separately we saw setTimeout removal involved.

sh-4.4$ node --trace-exit --trace-uncaught test281
["ok"](node:225) WARNING: Exited the environment with code 1
    at exit (internal/process/per_thread.js:173:13)
    at n (/opt/dynatrace/oneagent/agent/bin/1.219.162.20210726-100948/any/nodejs/nodejsagent.js:15740:49)
    at emit (events.js:314:20)
    at internal/process/execution.js:165:25

and

sh-4.4$ node --abort-on-uncaught-exception test281
["ok"]Uncaught ReferenceError: setTimeout is not defined

FROM
Illegal instruction (core dumped)

---

I see in the FAQ, that it talks about setTimeout, but what I can't understand from the README is why. That is not clear from the README, leading me to question why these methods are removed.

[jrafanie]

It looks like one way to get around this issue is to use execjs 2.8.1 with uglifier 4.2.0+ since it defers the context call asset compilation so it doesn't happen on Uglifier.new which happens when specifying the assets.js_compressor in the rails application configuration. With uglifier 4.2.0 and execjs 2.8.1, we can pre-compile the assets in an environment without dynatrace, then at runtime we can avoid shelling out to nodejs, which was conflicting with dynatrace.

I'll update the issue if this doesn't work.