Tests failing because of a possible analysis regression?

[Xanewok]

Starting from nightly-2017-07-07 (x86_64-unknown-linux-gnu), these are the new failing ones, apart from the ones mentioned in #400:

test::test_goto_def
test::test_highlight
test::test_hover
test::test_rename

It seems that it's handling this statement: let world = "world"; incorrectly, e.g.:

(mistaking &str for &&str)

thread 'test::test_hover' panicked at 'Could not find `[{"language":"rust","value":"&str"}]` in `{"jsonrpc":"2.0","id":11,"result":{"contents":[{"language":"rust","value":"&&str"}]

(finds this world, but not the original one)

thread 'test::test_goto_def' panicked at 'Could not find `"start":{"line":20,"character":8}` in `{"jsonrpc":"2.0","id":11,"result":[{"uri":"file:///h
ome/xanewok/repos/rls/test_data/goto_def/src/main.rs","range":{"start":{"line":21,"character":27},"end":{"line":21,"character":32}}}]}`', /checkout/s
rc/libcore/option.rs:823:4

[nrc]

cc rust-dev-tools/rls-analysis#75

[nrc]

I believe the problem is with either macro spans or the println macro - the definition is pointing at the redefinition in the macro, rather than the original definition. This might be a regression in spans or macros or it might be an issue in our processing of the data (e.g, we're not ignoring the macro definitions).

[Xanewok]

Tests are fixed and got re-enabled in 134a685.