Just as a heads up for @davidtwco : My first task is going to be to put in links to the locations where each file actually ended up
(you can see how I've been encoding it so far; my technique has been a pretty simple minded `find src/test/ui -name '*whatever-the-file-was.rs' invocation.)
but also, I think at least some of these cases are instances where the test is/was weak, in the sense that a lot of these tests assume lexical lifetimes for no reason: e.g. one should add a use of the borrow, rather than let it have an artificially short life time.
I'll add either add notes about such cases, or put up a PR fixing them.
(I think you should be able to find similar updates to tests if you look at the PR where we originally landed compare-mode ... I think ...)
Sounds good. If you've not got time to fix them but leave some notes then I can put a PR in.
Had a look over your PR and it seemed fine to me @pnkfelix. Might be worth getting another pair of eyes on it though.
@davidtwco I was curious : what technique/methodology did you use to validate the .nll.stderr files when you portes compile-fail to ui/?
A quick glance at each of them to see if it looked right. Not particularly rigorous but there were a lot of them.
Because even with this relatively small set of tests, i found my eyes quickly glazed over ...
I guess we should just plan a third round of comparison between .stderr and .nll.stderr. Maybe I’ll try to make a tool this time for equating the ones with only cosmetic deltas ...