ctx.expand_location() on py_binary expands to multiple files

[jacky8hyf]

Description of the bug:

When one runs ctx.expand_location() on the expression $(locations <a py_binary>) for a custom rule, multiple files are returned.

However, this is not the case for genrule's.

Which category does this issue belong to?

No response

What's the simplest, easiest way to reproduce this bug? Please provide a minimal example if possible.

BUILD

load(":myrule.bzl", "myrule")

py_binary(
    name = "a",
    srcs = ["a.py"],
)

genrule(
    name = "gen",
    tools = [":a"],
    outs = ["gen.txt"],
    cmd = "echo $(locations :a); : > $@",
)

myrule(
    name = "mytarget",
    deps = [
        ":a",
    ],
    value = "$(locations :a)",
)

myrule.bzl

def _impl(ctx):
    print(ctx.expand_location(ctx.attr.value, ctx.attr.deps))

myrule = rule(
    implementation = _impl,
    attrs = {
        "deps": attr.label_list(),
        "value": attr.string(),
    },
)

And create an empty WORKSPACE and a.py file.

With the above setup:

• If you run bazel build :gen, it will print bazel-out/k8-opt-exec-2B5CBBC6/bin/a, indicating that $(locations :a) correctly resolved to the generated binary.
• If you run bazel build :mytarget, it will print ./a.py bazel-out/k8-fastbuild/bin/a, indicating that $(locations :a) resolves to multiple files.

I tried turning tools into deps for my :gen target. Then, bazel build :gen also prints two paths.

I tried adding cfg = "exec" to deps's attr.label_list, but the result is the same.

This makes me unable to resolve to the correct binary to execute without some hacky code (e.g. CL)

How do I call ctx.expand_location so it produces the same output like genrule.tools does?

Which operating system are you running Bazel on?

Linux

What is the output of bazel info release?

release 6.4.0

If bazel info release returns development version or (@non-git), tell us how you built Bazel.

No response

What's the output of git remote get-url origin; git rev-parse master; git rev-parse HEAD ?

No response

Is this a regression? If yes, please try to identify the Bazel commit where the bug was introduced.

I don't think so.

Have you found anything relevant by searching the web?

https://stackoverflow.com/questions/57497270/bazel-how-to-run-py-binary-to-produce-input-files-before-testing

Any other information, logs, or outputs that you want to share?

No response

[pcjanzen]

related: #16381

[jacky8hyf]

Does this mean that https://github.com/bazelbuild/bazel/pull/16381/files/d8176177c95ff90c471db2c46085dfbc93fca4c1, the issue should be fixed?

[xich]

It looks like the fix was rolled back: ab71a10

Is there any resolution to this? Even if the the locations variants gave the ".py" and the executable back in a stable order, that would help. Right now I'm filtering the list.