"git submodule update" in build.rs with std::process::Command has no effect

Question

I'm trying to port C++ software to Rust and one of the parts is fetching a git submodule with a C++ library. I'd like to integrate that action into my custom build.rs script instead of forcing a user to do that manually before running cargo build. Here is my build.rs:

fn main() {
    std::process::Command::new("git")
        .args([
            "submodule",
            "update",
            "--init",
            "--depth 1",
            "--recommend-shallow",
    ])
    .output()
    .expect("Failed to fetch git submodules!");

    // here C++ code compilation with cc::Build happens
}

Unfortunately: the command didn't execute -> the submodule wasn't fetched -> C++ compiler started complaining about missing headers -> cargo threw an error. Otherwise than that build.rs is totally fine, because it worked perfectly after running git submodule update manually before cargo build. Suprisingly enough changing the command to echo cargo:warning=test, catching the output of the std::process::Command and outputting it with io::stdout().write_all(&output.stdout).unwrap(); resulted in cargo correctly reporting a warning. touching a test file and later rming it after the compilation worked as well. Why git doesn't work, but these commands do?

@tifrel I won't paste stdout here since it's just compiler's "dude, where's my header?" rambling. Stderr is a bit more interesting, though, because it's git submodule's usage info, which is printed every time you enter a bad command. It made me think and I've just split the --depth 1 argument into 2 separate ones. Sure enough stderr is now empty, but the build script still errors out with litany of errors in stdout. — Ziemowit Zabawa
– Ziemowit Zabawa, Commented Apr 3, 2022 at 20:46

tifrel · Accepted Answer · 2022-04-03 22:11:51Z

2

Given your own answer, I'd also like to suggest taking a look at the return type of std::process::Command::output. This is an std::io::Result, and while I cannot tell for sure, I would expect this to fail e.g. if the specified command is not accessible, but return an Ok variant on non-zero exit codes.

You can (and probably should) inspect the exit code from the obtained Output struct whenever you run std::process::Command, such that you can pinpoint the step at which your build script fails.

answered Apr 3, 2022 at 22:11

tifrel

4711 gold badge8 silver badges22 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

bguiz · Accepted Answer · 2022-04-04 01:28:47Z

1

I accidentally dropped the submodule while trying to solve this issue, which I detected by running git submodule update manually and achieving no effect. cargo build works perfectly automagically after the fix described in my comment. Please pretend this didn't happen and you haven't seen this question.

edited Apr 4, 2022 at 1:28

bguiz

28.5k49 gold badges165 silver badges256 bronze badges

answered Apr 3, 2022 at 21:31

Ziemowit Zabawa

538 bronze badges

Collectives™ on Stack Overflow

"git submodule update" in build.rs with std::process::Command has no effect

2 Answers 2

Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related