debug-info test failure on windows

[klutzy]

failures:
    [debug-info] debug-info/by-value-self-argument-in-trait-impl.rs
    [debug-info] debug-info/var-captured-in-nested-closure.rs

by-value-self-argument-in-trait-impl test fails to print the first self.

(gdb) print self
$1 = <optimized out>

var-captured-in-nested-closure test fails to print closure_local.

(gdb) print closure_local
No symbol "closure_local" in current context

cc #10452

[jdm]

We should check whether we're disabling all optimization when building the debuginfo tests.

[klutzy]

I manually built the first one: rustc.exe -Z extra-debug-info --opt-level=0 <test>.rs but gdb still prints $1 = <optimized out>. Does it indicate that something is missing?

[jdm]

I... don't really know what that means.

[alexcrichton]

We pump llvm through an extra g++ pass on windows, right? Is it possible that there's some optimizations accidentally getting tacked on at that stage?

[michaelwoerister]

by-value self is not really supported yet (see "Known issues" http://michaelwoerister.github.io/2013/09/27/what-you-call-the-present.html). I'm surprised that the first test case passes on other platforms.

[michaelwoerister]

I opened a separate issue about by-value-self: #10549

[alexcrichton]

Now that #10549 is closed, it sounds like this just needs to be verified to be working on today's compiler

[klutzy]

confirmed that both tests work with nightly. yay!

[alexcrichton]

Closing, the tests are checked in, and don't have ignore directives for windows. Once we start running debuginfo tests on windows, these should hopefully "just work".

[alexcrichton]

Oops, looks like debug-info/var-captured-in-nested-closure.rs is still xfail'd for windows,