-
Notifications
You must be signed in to change notification settings - Fork 1.7k
Closed
Labels
C-bugCategory: Clippy is not doing the correct thingCategory: Clippy is not doing the correct thing
Description
Appveyor failures:
-
windows-gnu
target doesn't support compiler plugins currently: GNU toolchain on Windows can't build compiler plugins since nightly 2017-12-15. rust#47029- Suggest disabling
windows-gnu
while the above issue is open.
- Suggest disabling
- 32-bit tests are broken in general. Travis never runs these so the failures don't matter there.
- I believe the underlying cause is missing the expected
.stderr
output for thetransmute_32bit
tests. Unfortunately the script for generating these is a shell-script for unix systems. Anyone have a working 32-bit cross compilation env that can run shell scripts?
- I believe the underlying cause is missing the expected
- dogfood test fails on permissions. Can't build
cargo-clippy
whilecargo-clippy
is running. Windows locks the executable and prevents modifying it while it's being executed.- One option would be to copy the
cargo-clippy.exe
under a different name before running the test itself. - Another option would be to use
cargo run --release
to build a release version while clippy itself would attempt to build a debug version. However this would also require building release version ofclippy-driver
which isn't a direct dependency ofcargo-clippy
.
- One option would be to copy the
Other failures
Plus a pile of random test failures tied to CRLF/LF differences. These don't seem to matter for Appveyor, but do fail on my local Windows machine. Ensuring the stderr
/stdout
reference files use LF line endings is a start. There's also a couple of places in the lints where this difference affects the output.
needless_continue.rs
skips whitespace using only\n
as match. On Windows\r
might exist in the file as well.doc.rs
fails in calculating the word positions within/*...*/
comment blocks. I believe this is caused by Rust stripping\r
s from the#[doc(..)
attributes so they aren't included in the line lengths, but they are still part of the span information maybe. Tried having a look at this, but never found the root cause. Adding a random -1 length/offset per line somewhere there "fixed" the issue on Windows so it seems to be off-by-one error per line.method.rs
I believe had a failure as well, but didn't dig into it when I realized these weren't Appveyor failures.
phansch, rassouly, goodmanjonathan, unv-annihilator, ryazanov and 1 more
Metadata
Metadata
Assignees
Labels
C-bugCategory: Clippy is not doing the correct thingCategory: Clippy is not doing the correct thing