Tests

[soulseekah]

Lots of permissions failures, about 10 tests are failing.

[soulseekah]

Bisect shows 1803ef7 as one of the culprits.

[soulseekah]

6 fails to go.

[soulseekah]

[soulseekah]

[soulseekah]

[soulseekah]

GravityView_Edit_Entry_Test::test_field_visibility_and_editability_caps's success depends on the number of forms that have been created... weirdest thing I've ever seen.

I modify the create_many call in GVCommon_Test::test_get_forms from 5 to 3 and it works. 1 and 2 works as well, anything over 5 stops working.

...It's like it fails unless there are more than 100 forms present.

I literally do a --filter GravityView_Edit_Entry_Test::test_field_visibility_and_editability_caps --repeat 20, it works for the first 4 times, then it stops working for 5 times, and then works for 11 times!

I'm stumped.

Here are 250 test runs of the same test:

Fails first 4 then fails another 50 after some good ones.

It actually fails if the form ID starts with a 1... 1, 17, 110, 1093, etc. 🗡️

[soulseekah]

input_1 is contained in for="input_17_2"... doh

2 to go 🏁

[soulseekah]

0?

[soulseekah]

[soulseekah]

2 😄

[soulseekah]

1 🍇

[soulseekah]

0. But not on all versions:

ugh

[soulseekah]

A regression in WordPress nightly:

https://core.trac.wordpress.org/ticket/28636

[zackkatz]

@soulseekah I thought the nightlies wouldn't show as errors in the build?

allow_failures:
      - php: nightly

[soulseekah]

See: https://core.trac.wordpress.org/ticket/28636#comment:34

[soulseekah]

See: https://core.trac.wordpress.org/ticket/28636#comment:34

nightly should be fine.

[soulseekah]

Errors: 1, Failures: 1... continuing the battle.

The error is from #877 d04c10a which is a good thing, strangely (and hopefully). It should be failing but for other reasons.

The failure is being worked on.

[soulseekah]

Bah, I don't know... it's soooooo bad:

But we're almost there: https://travis-ci.org/gravityview/GravityView/builds/604386137

[soulseekah]

There seems to be a bug in latest. GravityView_Merge_Tags_Test::test_replace_date_updated human time diff stuff. I would love to look into it if no update is released in a week or two, after handling more pressing matters. Okay?

[zackkatz]

@soulseekah Yep, sounds good.

[soulseekah]

It does look like the human test is time-sensitive in some way. Running it 100 times gets about 2% fails:

Maybe if the interval is sometimes too short (less than 1 second) then it doesn't output 0 seconds, but outputs the date. I'll check later.

[soulseekah]

Seems to have helped locally: