Invalid cargo metadata output and Lazy instance has previously been poisoned in Docker build

[Palmik]

Hi folks,

Since a few days ago I have been unable to build my project in Docker. The first error, on the first sqlx::query! invocation, is:

Invalid `cargo metadata` output: Error("EOF while parsing a value", line: 1, column: 0)

Followed by

Lazy instance has previously been poisoned

In full:

 error: proc macro panicked
  --> /home/rust/src/xxx:32:22
   |
32 |           let result = sqlx::query!(
   |  ______________________^
33 | |             r#"
34 | |             xxx
41 | |         )
   | |_________^
   |
   = help: message: Invalid `cargo metadata` output: Error("EOF while parsing a value", line: 1, column: 0)
   = note: this error originates in a macro (in Nightly builds, run with -Z macro-backtrace for more info)

error: proc macro panicked
  --> /home/rust/src/yyy:52:22
   |
52 |           let result = sqlx::query!(
   |  ______________________^
53 | |             r#"
54 | |             yyy
61 | |         )
   | |_________^
   |
   = help: message: Lazy instance has previously been poisoned
   = note: this error originates in a macro (in Nightly builds, run with -Z macro-backtrace for more info)

What I have tried:

• Different docker images (rust:1.53.0, ekidd/rust-musl-builder:1.51.0)
• Pinning the version of sqlx to 5.5.2 and 5.5.5
• Removing Cargo.lock and cargo clean

I have no idea why the error would appear only while building in Docker. I copy over my Cargo.lock file. When I build outside of Docker, I use the same rust version, I use export SQLX_OFFLINE=1, the same build command and everything works.

Relevant discussion:

• https://stackoverflow.com/questions/67999435/rust-sqlx-lazy-instance-has-previously-been-poisoned
• cargo fix cannot be used with sqlx offline mode #1229

Any tips are welcome.

[msdrigg]

Just had this happen. I solved it by fixing my .env file. My rust log like was missing the last quotation mark, so I fixed it and the problem resolved itself

-RUST_LOG = "info,tower_http=trace,grab_backend=debug
+RUST_LOG = "info,tower_http=trace,grab_backend=debug"

[c-git]

Just bumped into this but my message is the same as the OP. So I'm not missing a quotation mark. I haven't made any changes to my .env file recently.

Tried a cargo clean and that didn't help.

[c-git]

TLDR cargo metadata would fail in the workspace of one of my dependencies and once that was fixed I could then compile.

Finally found out what was wrong in my case. I was trying to build one of my crates in workspace A. It depends on a crate in workspace B. A different create from my dependency in workspace B had an error in it's Cargo.toml. This causing cargo metadata to fail in the workspace B cargo metadata would pass in the workspace A. However sqlx works in the context of the workspace where the crate is defined so even if I could run cargo metadata workspace A sqlx would fail because the same command in the workspace B would fail.

PS. Error was removing log from the workspace Cargo.toml not realizing on of the other crates still needed it.