rustdoc --version --verbose no longer displays commit hashes + date since beta 1.67

[matthiaskrgr]

stable is fine:
~/.rustup/toolchains/stable-x86_64-unknown-linux-gnu/bin/rustdoc --version --verbose

rustdoc 1.66.1 (90743e729 2023-01-10)
binary: rustdoc
commit-hash: 90743e7298aca107ddaa0c202a4d3604e29bfeb6
commit-date: 2023-01-10
host: x86_64-unknown-linux-gnu
release: 1.66.1
LLVM version: 15.0.2

but beta and nightly are no longer showing commit-hash and commit-date information:
~/.rustup/toolchains/beta-x86_64-unknown-linux-gnu/bin/rustdoc --version --verbose

rustdoc 1.67.0-beta.8 (7a9ae0ce6 2023-01-14)
binary: rustdoc
commit-hash: unknown
commit-date: unknown
host: x86_64-unknown-linux-gnu
release: 1.67.0-beta.8
LLVM version: 15.0.6

~/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/bin/rustdoc --version --verbose

rustdoc 1.68.0-nightly (333ee6c46 2023-01-18)
binary: rustdoc
commit-hash: unknown
commit-date: unknown
host: x86_64-unknown-linux-gnu
release: 1.68.0-nightly
LLVM version: 15.0.6

[matthiaskrgr]

it's weird because the info is still displayed correctly in the very first oneliner-summary, but if there was some tool that reads the data by looking at commit-hash or commit-date we still should display the right thing.

[ehuss]

This appears to have been introduced by #104184. The issue is that the macro uses CFG_VER_DATE/CFG_VER_HASH which is only set when compiling rustc.

One option is to try to share some code with rustc_cargo_env. Another option is to go the other way and get rustc to share some code with prepare_tool_cargo.

[jyn514]

Mentoring instructions: set

rust/src/bootstrap/compile.rs

Lines 792 to 794 in 4e163af

	if let Some(ref ver_date) = builder.rust_info().commit_date() {
	cargo.env("CFG_VER_DATE", ver_date);
	}

in

rust/src/bootstrap/tool.rs

Line 235 in 7c3f0d6

pub fn prepare_tool_cargo(

(Maybe we should try to unify those functions so they don't duplicate so much code and problems like this don't happen in the future?)

[jyn514]

I think it should be possible to write a run-make test ensuring this works properly.

[zephaniahong]

@rustbot claim

[anth0nyleung]

Hi @zephaniahong! im starting my journey of contributing to Rust, i found this issue a good starter for myself, do you think you would be open to let me take this chance to get familiar with the code base?

[zephaniahong]

@anth0nyleung Hey! Go for it!🤗

[zephaniahong]

Alternatively, you could check out #108111. I THINK that might be more straightforward. But I'll let you decide(:

[anth0nyleung]

@rustbot claim

[anth0nyleung]

I tried doing this change in tool.rs

diff --git a/src/bootstrap/tool.rs b/src/bootstrap/tool.rs
index ca5f500f93b..6788151f806 100644
--- a/src/bootstrap/tool.rs
+++ b/src/bootstrap/tool.rs
@@ -286,6 +286,9 @@ pub fn prepare_tool_cargo(
     if let Some(date) = info.commit_date() {
         cargo.env("CFG_COMMIT_DATE", date);
     }
+    if let Some(ver_date) = info.commit_date() {
+        cargo.env("CFG_VER_DATE", ver_date);
+    }
     if !features.is_empty() {
         cargo.arg("--features").arg(&features.join(", "));
     }

Then i did a ./x.py build and tried running this command rustdoc --version --verbose on the stage2 rustodc

$> ./build/aarch64-apple-darwin/stage2/bin/rustdoc --version --verbose
rustdoc 1.69.0-dev
binary: rustdoc
commit-hash: unknown
commit-date: unknown
host: aarch64-apple-darwin
release: 1.69.0-dev
LLVM version: 15.0.7

But it appears that the commit-hash and commit-date is still unknown. Im wondering if the issue is that builder.config.ignore_git is passed as true in GitInfo::new()

let info = GitInfo::new(builder.config.ignore_git, &dir);

This is my first time making changes and testing in this project, please feel free to correct me if im testing / building in the wrong way.

[jyn514]

@anth0nyleung I'd suggest changing

rust/compiler/rustc_driver_impl/src/lib.rs

Lines 702 to 705 in ae5473c

	unw(option_env!("CFG_VERSION")),
	unw(option_env!("CFG_VER_HASH")),
	unw(option_env!("CFG_VER_DATE")),
	unw(option_env!("CFG_RELEASE")),

to use env! instead of option_env so it's easier to tell when these variables are unset without having to build a full stage 2 compiler.

[Mark-Simulacrum]

Retagging as stable/stable regression, this already shipped in 1.67 as far as I can tell.

[duckymirror]

@anth0nyleung Are you still (interested in) working on this issue? I'd like to claim it otherwise.

[anth0nyleung]

@duckymirror please go ahead and claim it!

[duckymirror]

@rustbot claim