peakschris
commented
5 months ago The zstd failure is very similar to this: https://gist.github.com/sbabcoc/28f3b53ae2765444100ee6a183a939b2
Open peakschris opened 5 months ago
peakschris
commented
5 months ago The zstd failure is very similar to this: https://gist.github.com/sbabcoc/28f3b53ae2765444100ee6a183a939b2
This is a very early cut of windows support.
The build is failing on cargo build script zstd-sys Three test cases are passing.
I'm finding rules_rust very flaky on windows, particularly around its generation of very long filenames that do not work on windows or with msvc. The rules_rust maintainers have mitigated this to some extent in their builds by using D:\b as output_root, but the issues are hitting this repo too.
It may be that if the binaries can be cross compiled on linux, then rules_rust does not need to be a dependency on the windows platform and things might go more smoothly.
The current blocker is zstd crate that is failing to build on windows.
@alexeagle This is as far as I can go right now with this; it needs a CI or rust expert to get past the current hurdle. If that can be solved I can have another look.
One question: Can we avoid generating zip files with this ruleset for py_binary? I was hoping that this ruleset already does that, it seems not (the zip files and hermetic toolchain together are extremely slow to create and unzip on windows). Are there any plans to directly invoke python with system interpreter?