I am seeing a Windows-only failure in a Bazel build that looks like rustc is not accepting an externed crate even though the referenced artifact exists.
The reduced repro is a direct rustc.exe invocation from the Bazel action context that compiles a file containing only use serde as _; and fn main() {} from a response file.
That command is given the same target triple, sysroot, extern path for serde, and dependency search paths that were dumped from the failing Bazel action.
The serde rlib exists on disk, the dependency directories exist on disk, and the same general Bazel flow works for the repository on Linux and macOS.
On Windows, the direct smoke compile still fails with E0463 saying it cannot find serde, which makes this look deeper than a missing BUILD dependency or missing generated artifact.
The Harper-side tracker is harpertoken/harper#347 and the investigation PR with the diagnostic workflow is harpertoken/harper#344.
If this points to known Windows behavior in rules_rust or to a specific expectation around how extern rlibs should be surfaced to rustc in this context, guidance would be useful.
I am seeing a Windows-only failure in a Bazel build that looks like rustc is not accepting an externed crate even though the referenced artifact exists.
The reduced repro is a direct rustc.exe invocation from the Bazel action context that compiles a file containing only use serde as _; and fn main() {} from a response file.
That command is given the same target triple, sysroot, extern path for serde, and dependency search paths that were dumped from the failing Bazel action.
The serde rlib exists on disk, the dependency directories exist on disk, and the same general Bazel flow works for the repository on Linux and macOS.
On Windows, the direct smoke compile still fails with E0463 saying it cannot find serde, which makes this look deeper than a missing BUILD dependency or missing generated artifact.
The Harper-side tracker is harpertoken/harper#347 and the investigation PR with the diagnostic workflow is harpertoken/harper#344.
If this points to known Windows behavior in rules_rust or to a specific expectation around how extern rlibs should be surfaced to rustc in this context, guidance would be useful.