Stream: t-compiler/const-eval

Topic: Running miri tests on windows

Nick12 (Oct 05 2019 at 11:24, on Zulip):

Is it correct if, to run the miri tests, I did the following:
rustup component add miri
cargo miri setup
set MIRI_SYSROOT=path_given_by_miri_setup
cargo test (in the miri repo)

I was under the suspicion I could only test on *nix after reading since it suggests using the binary (Which I'm assuming can't be used on windows sans maybe msys/cygwin/mingw stuff). Would it make sense to modify that?

Christian Poveda (Oct 05 2019 at 14:25, on Zulip):

I think there is no problem with testing miri on windows

Christian Poveda (Oct 05 2019 at 14:25, on Zulip):

but what do you want to test? your locally built miri?

Nick12 (Oct 07 2019 at 04:34, on Zulip):


Christian Poveda (Oct 07 2019 at 04:36, on Zulip):

I think the miri script doesn't work correctly in windows

RalfJ (Oct 09 2019 at 13:35, on Zulip):

@Nick12 the instructions only mention *nix because Linux is the only platform on which I ever tried any of this. I don't have a Windows on which I could do development, nor do I have much time or motivation for tending to the countless quirks of that platform. so we mostly rely on Rust's excellent cross-platform nature and on CI to catch big "oopsies", but we try to be careful in our documentation.
If you find steps that work for Miri on Windows, a PR or patches to make our scripts work on more platforms would be more than welcome!

RalfJ (Oct 09 2019 at 13:36, on Zulip):

basically we support Windows insofar as it doesn't require much work and/or platform-specific knowledge, and the reason is a lack of a Windows expert to care enough to do better than that.

Nick12 (Oct 09 2019 at 15:33, on Zulip):

Makes sense. I might do that in the weekend

RalfJ (Oct 09 2019 at 15:54, on Zulip):

also see, someone else seems to be looking into similar issues

Last update: Nov 15 2019 at 21:15UTC