Support Sync trait for rpm::Error

[marxin]

I would like to use a return type equal to anyhow::Result<RPMFile> for a multi-threaded code and I face the following issue:

        let header = rpm::PackageMetadata::parse(&mut buf_reader).context("RPM parsing failed")?;

error[E0599]: the method `context` exists for enum `Result<PackageMetadata, Error>`, but its trait bounds were not satisfied
   --> src/main.rs:140:67
    |
140 |         let header = rpm::PackageMetadata::parse(&mut buf_reader).context("RPM parsing failed")?;
    |                                                                   ^^^^^^^ method cannot be called on `Result<PackageMetadata, Error>` due to unsatisfied trait bounds
    |
   ::: /home/marxin/.rustup/toolchains/stable-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/result.rs:502:1
    |
502 | pub enum Result<T, E> {
    | --------------------- doesn't satisfy `_: Context<PackageMetadata, Error>`
    |
   ::: /home/marxin/Programming/rpm-rs/src/errors.rs:9:1
    |
9   | pub enum Error {
    | --------------
    | |
    | doesn't satisfy `rpm::Error: Sync`
    | doesn't satisfy `rpm::Error: anyhow::context::ext::StdError`
    | doesn't satisfy `rpm::Error: std::marker::Send`
    |
    = note: the following trait bounds were not satisfied:
            `rpm::Error: anyhow::context::ext::StdError`
            which is required by `Result<PackageMetadata, rpm::Error>: anyhow::Context<PackageMetadata, rpm::Error>`
            `rpm::Error: std::marker::Send`
            which is required by `Result<PackageMetadata, rpm::Error>: anyhow::Context<PackageMetadata, rpm::Error>`
            `rpm::Error: Sync`
            which is required by `Result<PackageMetadata, rpm::Error>: anyhow::Context<PackageMetadata, rpm::Error>`

[marxin]

@dralley Is there a straightforward way we can take to implement this?

[dralley]

Dunno, you could try annotating the error type with Send and Sync and seeing which variants the compiler complains about.

[marxin]

One can annotate neither Sync nor Send, however, I noticed the problem goes away if I remove usage of all the Box<dyn std::error::Error> members in the error enum variants. Can't we just replace all the inner errors with String fields that will be constructed with inner.to_string() ?
To be honest, I'm just learning Rust, but that's definitely something that makes the library more usable with anyhow and multi-threaded execution.

[marxin]

Which might not be possible due to:

error[E0599]: the method `as_dyn_error` exists for reference `&String`, but its trait bounds were not satisfied
   --> src/errors.rs:69:9

which is caused by fact that the GPG errors use dyn Error.

[dralley]

Apologies, I won't be able to look at this much until the weekend (really busy at the moment) but I'm definitely interested in making it easier to use.

[cmeister2]

You can do it like this if you like: e61be9b

[marxin]

You can do it like this if you like: e61be9b

Great, it works for me! Thank you and can you please make a pull request with the changes?

[cmeister2]

I'll hold off on it for now because I haven't checked for definite that the wrapped errors are *Sync* and *Send*. If they're not then there could be some fun breakages. I just wanted to propose a potential way forward.

…

On Thu, 9 Nov 2023 at 17:15, Martin Liška ***@***.***> wrote: You can do it like this if you like: e61be9b <e61be9b> Great, it works for me! Thank you and can you please make a pull request with the changes? — Reply to this email directly, view it on GitHub <#191 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAPA36I6W37LWEL7HDSPYL3YDUFTRAVCNFSM6AAAAAA6VA5UROVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMBUGIZTONJYGY> . You are receiving this because you commented.Message ID: ***@***.***>

[dralley]

@cmeister2 @marxin Are you OK with the solution in #198 ?

[cmeister2]

@cmeister2 @marxin Are you OK with the solution in #198 ?

Yes, it looks very elegant.

[marxin]

@cmeister2 @marxin Are you OK with the solution in #198 ?

Yep, I like it much.