This env param does work even without RunEnvironmentInfo (similar to args).

@cameron-martin Are you sure that's true? As far as I know the args parameter is treated specially for all rules, but env is only for native rules.

Looks like I'm too eager to clean up my issues list. I'll reopen this as you're probably right about this.

Hi! I would also like this! For now I'm working around this with a patch, using this rather small change:

https://github.com/bazelbuild/bazel-skylib/compare/main...redsun82:bazel-skylib:env-native-binary?expand=1

I would have contributed it as a PR, but I'm a bit unsure about:

• setting up tests
• bazel version compatibility (is RunEnvironmentInfo supported across all versions bazel-skylib must support? If not, how can that code be made compatible across versions?)

@redsun82 You could use bazel_features, which provides you a way to test for certain global symbols: https://github.com/bazel-contrib/bazel_features/tree/main?tab=readme-ov-file#accessing-globals
RunEnvironmentInfo is already supported: https://github.com/bazel-contrib/bazel_features/blob/443861571a389ddc16d17690ab8e46ee87b4ea57/private/globals.bzl#L5

ah, nice one @fmeum

For the moment I've opened this draft PR, I'll update it sometime soon with the RunEnvironmentInfo check and have a look at the existing native_binary tests.

I've updated the PR, which turned out a bit trickier than I thought. For the moment I decided to replicate part of what bazel_features does to avoid complicating the legacy workflow-based integration instructions, which is not ideal. If you deem that we should rather go for bazel_features to simplify the code but update the instructions to integrate bazel_skylib into a WORKFLOW file accordingly with the new dependency, I can update the PR further.

I've opened another PR based on the previous one but using bazel_features (and modifying the release instructions for WORKSPACE setup). There was a bit of gymnastics involved in getting the doc update right, but I guess the two approaches can be compared now.

I've got a first approval on #484 a couple of weeks ago (thanks @jgertm!). I've now updated the PR, and added location expansion to env variables (I actually have a use case for passing LD_PRELOAD with a library in data).