Defer glob failures

[mitsuhiko]

This is an experimental feature to address #283

[JohnnyMorganz]

Thanks for the quick turnaround!
This does now continue through all snapshot failures and print out a count in the end, but it is very noisy, and it also regenerates all the snapshots (my expectation is that cargo test shouldn't regenerate snapshots whilst cargo insta test does, but that might be wrong)

Once it goes through all the failures it then prints out all the snapshot differences, and since there are 66 of them its a lot:

(Also, minor typo failuress at the end there)

Is it possible to make it instead print something equivalent to rusttest output (maybe behind a flag for quiet output?)

tests__standard@ambiguous-syntax.lua  ... ok
tests__full_moon_test_suite@anonymous-functions-2  ... FAILED

[mitsuhiko]

The default of INSTA_UPDATE is auto mode where if it's not in CI it will write new files. Try again with INSTA_UPDATE=no then it should not generate any changed snapshots like it would in CI.

[JohnnyMorganz]

The default of INSTA_UPDATE is auto mode where if it's not in CI it will write new files. Try again with INSTA_UPDATE=no then it should not generate any changed snapshots like it would in CI.

Ah yep, that stopped it, thanks.
The main thing is just the noisy "snapshot differences" section still, a quieter overview would be awesome.

Also, just a side note, since its not failing on first snapshot error, cargo test can take a little bit longer (2s vs 7s when keep going). It doesn't really matter to me, but it may be inconvenient for others (its heavily dependent on the code being tested).

[mitsuhiko]

It might be interesting to couple it with the failfast option of rusttest. That way it aligns with what users might already use. Will give that a whirl.

[mitsuhiko]

Turns out that the fail fast flag is only passed to the cargo wrapper so I won't be able to access this. Adding another flag to envvars is unlikely to be discoverable. I will try around with this a bit.