Open sourcefrog opened 7 months ago
It looks like it's still happening, from a similar command
/home/mbp/.rustup/toolchains/beta-x86_64-unknown-linux-gnu/bin/cargo nextest run --package cargo-nextest --package integration-tests --package nextest-filtering --package nextest-metadata --package nextest-runner
However strangely when I run that outside cargo-mutant it seems to pass...
I'm going to unassign and deprioritize this for now because while it might be pointing to a bug in cargo-mutants or a failure to handle some reasonable behavior, it also seems really entangled with some quirks in Nextest.
Using the in-progress support for running in-place (#157), I can reproduce this running in-place, so it's not a problem with copying.
The log shows that cargo-mutants runs
and that does fail in a similar way in this tree:
Very interestingly, invoking it with
cargo +beta
and otherwise the same args does pass, and also without the+beta
. It also passes with the explicit path to e.g.~/.cargo/bin/cargo
.But, it fails if I give it the path to a toolchain-specific binary, like
~/.rustup/toolchains/beta-x86_64-unknown-linux-gnu/bin/cargo
.I guess rustup installs a wrapper binary in
~/.cargo/bin/cargo
(??) and perhaps it's setting an environment variable or otherwise doing something that nextest depends upon.The nextest errors have a badly printed
Vec<u8>
like this:But after decoding that I can see the errors are really like this
I can reproduce the same failure if I run the fixture binary directly, outside of the tests: