Support panics across stack growth boundaries

[oli-obk]

No description provided.

[alexcrichton]

Ok looks like the segfault here is showing up on MSVC

[alexcrichton]

FWIW it's likely the case that https://github.com/alexcrichton/stacker/pull/4 needs to be dusted off to solve the MSVC issues

[oli-obk]

Would you be ok with disabling windows support for now and using the fallback?

[alexcrichton]

It's fine to not support windows in this crate, but it's not fine to do that in the compiler unfortunately

[oli-obk]

it's not fine to do that in the compiler unfortunately

The compiler currently does not use this at all, using stacker while only supporting linux/mac would fix the regression issues we have open. There are none on windows as of now.

[alexcrichton]

The stack size isn't really magically larger on Windows though, so the lack of reports on Windows doesn't mean it doesn't exist there, just that it hasn't been run into yet. It's the same bug on all platforms, and we can't fix it on 2/3 tier 1 platforms.

[oli-obk]

True, but I don't see #8 getting ready for being merged before the beta freeze.

Isn't it better to fix a bug on 2/3 of the platforms with a plan for fixing the last 1/3rd later, than to let the regression run into beta and stable?

[alexcrichton]

I personally feel (others may wish to weigh in) that we can't fix 2/3 tier 1 platforms. This fix must be uniform across all tier 1 platforms.

[alexcrichton]

I think this is supported and tested with https://github.com/alexcrichton/stacker/pull/8, so closing!