Test ordering of glue_drop and glue_takew in self-re-assignment.

[nrc]

closes#3290

[brson]

We're closing a lot of 1.0 bugs this week! 💃

[alexcrichton]

Looks like this just needs a pub fn main to appease the check-fast target (which will hopefully be gone soon!)

[nrc]

You mean that fn main should be pub fn main? Is that true in general for tests? How do you recognise from those logs that check-fast is the issue? And does this mean I should run make check and make check-fast before PRing? (So many questions...)

[alexcrichton]

You mean that fn main should be pub fn main

yes

Is that true in general for tests

Just for the run-pass suite. The reason is that we have a check-fast suite on windows which basically concatenates all the tests together, and for resolve to work each main function must be pub to be usable. See #12304 for removing check-fast. (this suite is only run on bsd for a slow aws bot and windows for slow spawn times).

How do you recognise from those logs that check-fast is the issue

The bsd bot is complaining about error: function main is private with some weird t_XXX::main() expression, which is what the test concatenator is generating.

And does this mean I should run make check and make check-fast before PRing

Perhaps, I always forget pub fn main, but for super exhaustiveness check-fast will catch this bug.

[nrc]

Thanks for the info!

New version with pub in.