tests: mock saveLhr and assert no unit test source changes

[patrickhulce]

Summary
We wanted to mock the saveLhr function, not just spy it. Also adds the git diff test to our github unit test suite :)

Related Issues/PRs
fixes https://github.com/GoogleChrome/lighthouse/pull/11509/files#r499844389 and our travis builds

[googlebot]

☹️ Sorry, but only Googlers may change the label cla: yes.

[connorjclark]

Wanna add to the smokes step too?

[patrickhulce]

yeah I'll add everywhere I just noticed it happens on update:sample-json too

[patrickhulce]

the tests should now fail before I fix the update:sample-json path, making sure we're catching this now :)

[connorjclark]

LGTM

[patrickhulce]

soooooo @connorjclark after looking at all the places affected by the writing lhr.json change I'm not sure I like it as a thing.

This PR changes the extension to .report.json for now, but why did we need the LHR -A writing? The original PR doesn't have a description and -A creates a JSON already when you just add --output=json so I'm not super clear why we need it also written to artifacts. Writing audit output to a folder that's supposed to be the artifacts doesn't feel like the optimal solution (and rears its head here in several places according to the git clean failures in numerous tests that are difficult to mock), so maybe there is another way we can work to solve whatever the root cause is that inspired #11509?

[connorjclark]

soooooo @connorjclark after looking at all the places affected by the writing lhr.json change I'm not sure I like it as a thing.

I thought the mock you added would prevent this?

[patrickhulce]

I thought the mock you added would prevent this?

It prevents it in those specific tests but apparently we exercise audit mode in several other places that weren't touched in #11509.

Do you have any thoughts on the rest of my comment?

[connorjclark]

What if we save in the "artifact folder" if: -A is set and --output includes json? That should avoid the test issues.

Currently it's too cumbersome to hold on to a LHR for a run, compared to how easy it is for artifacts. especially combined with your new script to save the latest-run folder.

Would also be ok with a new folder in .tmp: historic LHRs. maybe even just always do it for every run if LH_DEBUG or DEBUG is true or w/e....

[patrickhulce]

What if we save in the "artifact folder" if: -A is set and --output includes json? That should avoid the test issues.

For future readers, @connorjclark and I discussed offline and settled on this solution with the addition of -G being used.

[brendankenny]

For future readers, @connorjclark and I discussed offline and settled on this solution with the addition of -G being used.

amusing to look back at the original -GAR proposal thread