rustc_env_file can now render workspace status stamp format strings

[UebelAndre]

This change allows for rust_* targets to specify the stamp attribute to determine if or how format strings ({VAR}) will be rendered in files passed to the rustc_env_files attribute.

load("@rules_rust//rust:defs.bzl", "rust_binary")
load("@bazel_skylib//rules:write_file.bzl", "write_file")

write_file(
    name = "env",
    out = "stamp.env",
    content = ["BUILD_TIMESTAMP={BUILD_TIMESTAMP}"],
)

rust_binary(
    name = "bin",
    srcs = ["main.rs"],
    rustc_env_files = [":env"],
    stamp = 1, # forced stamping
)

[UebelAndre]

@hlopko Friendly ping, would love to get these changes in 😅

[illicitonion]

Someone with more C++ than me should probably review the C++, but otherwise looks great to me! :)

[hlopko]

Only minor things, but looks good to me. I apologize for the delay.

[hlopko]

I don't have any issue with this implementation, but here https://docs.bazel.build/versions/main/skylark/lib/transition.html#transition I see that //command_line_option:cpu works for transitions for accessing --cpu, so maybe //command_line_option:stamp works as well and can be used as any other build settings. May be worth trying out.

[UebelAndre]

This implementation was inspired by rules_pkg which has been in place for some time. I'd rather stick with the same approach since it seems to have withstood the test of time (so far).

[hlopko]

Ack. But by trying to use //command_line_option:stamp you may discover that it's already possible to access value of --stamp flag from Starlark today, negating the need for the workaround.

[hlopko]

(But if I had to bet, I'd bet on //command_line_option thing not working as a normal build setting, I'd expect the label only exists in the context of configuration transitions)

[UebelAndre]

These two options may not be what you were thinking but demonstrating that they don't work.

diff --git a/rust/private/stamp.bzl b/rust/private/stamp.bzl
index 1a7cab6..3292225 100644
--- a/rust/private/stamp.bzl
+++ b/rust/private/stamp.bzl
@@ -48,7 +48,7 @@ def stamp_build_setting(name, visibility = ["//visibility:public"]):
     _stamp_build_setting(
         name = name,
         value = select({
-            ":stamp_detect": True,
+            "//command_line_option:stamp": True,
             "//conditions:default": False,
         }),
         visibility = visibility,

ERROR: Analysis of target '//rust/private:stamp' failed; build aborted: no such package 'command_line_option': BUILD file not found in any of the following directories. Add a BUILD file to a directory to mark it as a package.

diff --git a/rust/private/stamp.bzl b/rust/private/stamp.bzl
index 1a7cab6..7f0f963 100644
--- a/rust/private/stamp.bzl
+++ b/rust/private/stamp.bzl
@@ -41,7 +41,7 @@ _stamp_build_setting = rule(
 def stamp_build_setting(name, visibility = ["//visibility:public"]):
     native.config_setting(
         name = "stamp_detect",
-        values = {"stamp": "1"},
+        values = {"//command_line_option:stamp": "1"},
         visibility = visibility,
     )

ERROR: /rules_rust/rust/private/BUILD.bazel:15:20: in values attribute of config_setting rule //rust/private:stamp_detect: error while parsing configuration settings: unknown option: '//command_line_option:stamp'. Since this rule was created by the macro 'stamp_build_setting', the error might have been caused by the macro implementation

I've used //command_line_option with transitions (exactly what you linked to earlier) but I feel the build setting is simpler than transitions and would opt to keep this as is.

[UebelAndre]

This is something I'm happy to make another PR for if I can understand the appropriate change. Since this is an exciting change for some, I'm gonna merge this since ultimately this is an implementation detail that does not impact the API.

[UebelAndre]

@hlopko Thanks for the review! Some quick follow up questions if you don't mind 😅