fix FIPS regression in 4.x

[sneak]

Checklist

Affected core subsystem(s)

tiny change to lib/_tls_wrap.js

Description of change

This fixes a regression introduced in 4.2.4 where under certain circumstances (I believe mocha test framework, but didn't isolate) the code in getDefaultSessionIdContext() would throw an exception because process.config.variables was undefined and thus doesn't have attributes to access.

Why the project's libs in use were mucking about in process.config like that we'll never know, but this defensive patch makes it work again (like in 4.2.3).

#3755 (comment)

[jasnell]

this can simply be:

if (process.config && process.config.variables && process.config.variables.openssl_fips)

Using the in syntax is not common in core.

[sneak]

thanks for the feedback! will update now.

FWIW I was guarding against the case where they are truthy but perhaps not an object (lacking the required attribute).

[jasnell]

Small nit but generally this is ok. My apologies, however, in that I wasn't clear on my first note. Fixes to v4.x need to be opened as pull requests to the v4.x-staging branch rather than directly against v4.x. We only land commits into v4.x when we're actually cutting a release. Doing so allows us to stage out what actually lands in any particular v4 release. Can I ask you to please make the change I suggest above, but close this PR and open a new one against v4.x-staging?

[sneak]

Per request, created new PRs against v4.x-staging and v5.x (couldn't find a v5.x-staging).

[jasnell]

Thank you @sneak ... I appreciate it. Just an FYI, We use the staging branches for LTS branches only (v4.x-staging, v0.12-staging and v0.10-staging). Since v5 is not an LTS stream, there is no staging branch for it.

[sneak]

Makes perfect sense. Glad to help - may my hours lost yesterday not bite anyone else using mocha/supertest/whatever broke it. (I know the people upstream of the app I am using are going to hit this same bug when they upgrade their node version.)