Wrong demangling under miri

[karolzwolak]

The demangling is wrong when run under miri. This might be a bug in miri.

A lot of demangling tests fail when run under miri.
To reproduce:

• install miri rustup +nightly component add miri (see using miri)
• run some tests under miri: cargo +nightly miri test --no-fail-fast demangle (filtering tests because they are slow under miri)

[khuey]

Given that this crate has almost zero use of unsafe code I would assume the bug is in miri until demonstrated otherwise.

[karolzwolak]

Yeah exactly. And the behavior changes under miri — it's not just that the miri detects UB or anything like that. Still this should probably should stay open here until it's solved on the miri side in case anyone stumbles onto it too.

It'd be nice though if you could confirm that you can reproduce this and this isn't some quirk on my system.

[khuey]

Yes, I can confirm that I can reproduce it. From looking at two failing tests real quickly it appears that under miri the no_params option is (incorrectly) set somehow.