How do you run an analysis?

[typesanitizer]

Pasting my Reddit comment here:

let's say that given a crate's source code, I want to create a data structure containing type signatures of all functions with proper naming, trait/lifetime constraints, post proc macro-expansion. I do not care about codegen (so being able to avoid it would a big plus).

What would be the best way of doing so? (Best = if there is some change in the compiler/cargo, I get the change soon).

I got a suggestion that I should use this crate. Looking at the source code here/in rls, I can't really tell what is going on :(. For example, if I have

    let host = AnalysisHost::new(Target::Debug);
    let p = Path::new("/home/theindigamer/Code/rls-analysis");
    // Not entirely sure what the second path should be
    host.reload(p, Path::new("."));
    println!("Roots: {:?}", host.def_roots());

I'd naively expect that I'll get the ids of rls-analysis crate and its dependencies (transitively), and then I can call for_each_child_def to recursively traverse all the modules in a crate. Instead I get an empty vector, so it seems like nothing has been analyzed yet. Am I vaguely on the right track or way off the mark?

The other option I'm exploring is using syn to parse stuff (which is working ok) and implement the module traversal/Cargo.toml parsing myself.

I understand that writing a long explanation might be too much work and you may not have the time; in that case, a pointer to roughly the right place(s) in the code would still be appreciated.

[nrc]

The this works is that you need to provide save-analysis data to this crate. The easiest way to do that is to run a nightly compiler with -Zsave-analysis. This crate then needs pointing at the directory with the save-analysis data in it for all crates. As an example of doing all this, see running rustc and updating rls-analysis from date

[typesanitizer]

I think you accidentally linked to the same thing twice. But yeah, I can see the save analysis data generated in JSON after running it! However, a couple of lines in the linked code don't make sense to me. With cargo 1.31.0-nightly (ad6e5c003 2018-09-28), I actually see the data being generated under target/debug/deps/save-analysis whereas these lines

https://github.com/nrc/cargo-src/blob/34b3b4cae63ed2de53672135cac8a4ffa5b26a71/src/build.rs#L19

https://github.com/nrc/cargo-src/blob/34b3b4cae63ed2de53672135cac8a4ffa5b26a71/src/build.rs#L71-L75

make me think it should be saved under target/rls/debug/deps/save-analysis. Anyways, if I ignore that for a moment, the path_prefix (1st) and base_dir (2nd) arguments weren't clear to me so I looked at this:

https://github.com/nrc/cargo-src/blob/3c1a3327a2a92a61275f711cfa9b27fd43712f31/src/file_controller/mod.rs#L187-L190

and it seems path_prefix should be "." and base_dir should be "/home/.../crate-dir" -- is that right? I've tried various different combinations (including appending `"/target") but none of them work.

(Not sure why the code snippets aren't being rendered 🙁 )

[nrc]

Sorry the second link should have been: https://github.com/nrc/cargo-src/blob/master/src/file_controller/mod.rs#L187-L192

You can generate the data wherever you like. The RLS creates a separate directory for building (target/rls) and so looks in there. That is how CargoAnalysisLoader is setup, but you can pass your own loader if you want to use a different directory

path_prefix and base_dir are somewhat confusing. base_dir is used for converting relative paths to absolute ones, I think it points either to the crate directory or the src directory. path_prefix should point to the crate directory