Add test execution time to wptreport.json format

[jgraham]

This was suggested in # #9972.

Basically the execution time is the timestamp of the test_end event minus the timestamp of the test_start event. So to fix this look in tools/wptrunner/wptrunner/formatters.py, add a time field to the data and populate it with the execution time per test.

[ghost]

Can I work on it?

[jgraham]

Yes.

[foolip]

This would be great!

[marimeireles]

Hey, @ManiKishan
Are you still working on this?
Thanks.

[ghost]

No,sorry I would not be able to work on it now.
Sorry.

[marimeireles]

I could try to fix this, if you don't mind.

[jgraham]

Go for it.

[foolip]

@marimeireles do you still intend to work on this?

[marimeireles]

Hey,
Not in the moment, sorry!

[foolip]

Oh yay, @gsnedders for president! @jugglinmike @Hexcles, have you seen this show up in runs yet? How much does it change the wptreport size?

[jugglinmike]

Yup, the data is being collected!

$ ssh bocoup-wptd-build-master 'sudo -u wptdmaster cat /mnt/buildmaster-data/chunk-results/f556cda8f3ab2c14851a63ecd3def3c03de8182d/edge-17-windows-10-sauce/10_of_100.json | python -m json.tool | head -n 10'
{
    "results": [
        {
            "duration": 3385,
            "message": null,
            "status": "PASS",
            "subtests": [],
            "test": "/css/css-writing-modes/background-size-document-root-vrl-004.html"
        },
        {

I don't have a great way to check the size of the final reports, though, since we remove them immediately following successful upload.

[Hexcles]

I will have the numbers soon (fixing a bug in receiver now).

[jgraham]

wptreport is not optimized for size, if we are concerned about non-compressed size there's definitely changes we could make. What's the motivation there?

[foolip]

There's a request size limit in the wpt.fyi receiver, and we're not miles away from it. Hopefully this doesn't push it over the limit.

[Hexcles]

@foolip we only transport the compressed form

Adding a field like this wouldn't blow the compressed size up too much due to the nature of dictionary-based compression algorithms.

[foolip]

Well, with enough precision the numbers will be increasingly random, and not compress as well.