Panic building a crate without a cargo.toml

[jyn514]

How could crates not have a cargo.toml and still be published to crates.io ???

Aug 09 02:50:11 docsrs cratesfyi[27117]: 2020/08/09 02:50:11 [INFO] rustwide::crates::cratesio: extracting crate reactiontime 0.1.0 into /home/cratesfyi/workspace/builds/reactiontime-0.1.0/source
Aug 09 02:50:11 docsrs cratesfyi[27117]: 2020/08/09 02:50:11 [INFO] rustwide::prepare: validating manifest of crates.io crate reactiontime 0.1.0 on toolchain nightly
Aug 09 02:50:11 docsrs cratesfyi[27117]: 2020/08/09 02:50:11 [ERROR] cratesfyi::build_queue: Failed to build package reactiontime-0.1.0 from queue: missing Cargo.toml
Aug 09 02:50:11 docsrs cratesfyi[27117]: Backtrace:    0: failure::backtrace::internal::InternalBacktrace::new
Aug 09 02:50:11 docsrs cratesfyi[27117]:    1: <failure::backtrace::Backtrace as core::default::Default>::default
Aug 09 02:50:11 docsrs cratesfyi[27117]:    2: rustwide::prepare::Prepare::prepare
Aug 09 02:50:11 docsrs cratesfyi[27117]:    3: rustwide::build::BuildBuilder::run
Aug 09 02:50:11 docsrs cratesfyi[27117]:    4: cratesfyi::docbuilder::rustwide_builder::RustwideBuilder::build_package
Aug 09 02:50:11 docsrs cratesfyi[27117]:    5: cratesfyi::build_queue::BuildQueue::process_next_crate
Aug 09 02:50:11 docsrs cratesfyi[27117]:    6: <std::panic::AssertUnwindSafe<F> as core::ops::function::FnOnce<()>>::call_once
Aug 09 02:50:11 docsrs cratesfyi[27117]:    7: cratesfyi::utils::queue_builder::queue_builder
Aug 09 02:50:11 docsrs cratesfyi[27117]:    8: std::sys_common::backtrace::__rust_begin_short_backtrace
Aug 09 02:50:11 docsrs cratesfyi[27117]:    9: core::ops::function::FnOnce::call_once{{vtable.shim}}
Aug 09 02:50:11 docsrs cratesfyi[27117]:   10: <alloc::boxed::Box<F> as core::ops::function::FnOnce<A>>::call_once
Aug 09 02:50:11 docsrs cratesfyi[27117]:              at /rustc/c7087fe00d2ba919df1d813c040a5d47e43b0fe7/src/liballoc/boxed.rs:1008
Aug 09 02:50:11 docsrs cratesfyi[27117]:       <alloc::boxed::Box<F> as core::ops::function::FnOnce<A>>::call_once
Aug 09 02:50:11 docsrs cratesfyi[27117]:              at /rustc/c7087fe00d2ba919df1d813c040a5d47e43b0fe7/src/liballoc/boxed.rs:1008
Aug 09 02:50:11 docsrs cratesfyi[27117]:       std::sys::unix::thread::Thread::new::thread_start
Aug 09 02:50:11 docsrs cratesfyi[27117]:              at /rustc/c7087fe00d2ba919df1d813c040a5d47e43b0fe7/src/libstd/sys/unix/thread.rs:87
Aug 09 02:50:11 docsrs cratesfyi[27117]:   11: start_thread
Aug 09 02:50:11 docsrs cratesfyi[27117]:   12: __clone

[Nemo157]

It has cargo.toml instead, only way I can think of this happening is it was published by something other than cargo, and crates.io doesn't actually validate the content of the crate archive (or that validation is lacking compared to what cargo does).

[Nemo157]

Ohhhhh, case-insensitive filesystem maybe? Probably needs a cargo bug ticket.

[pietroalbini]

I think we can change rustwide to rename cargo.toml to Cargo.toml if the latter is missing.

[Nemo157]

Cargo doesn't support it either, trying to build with a dependency on reactiontime on a case-sensitive filesystem gives

error: failed to read `/home/nemo157/.cargo/registry/src/github.tiyicn.workers.dev-1ecc6299db9ec823/reactiontime-0.1.0/Cargo.toml`

[jyn514]

I think the fix here should be to give a build failure if Cargo.toml is missing. In a sense this is really the same as #797, but we also panic on the error.

[jyn514]

Err .. it turns out this isn't actually a panic, just error logging. So there's no panics going on here. The only issue is that the build failure isn't recorded (#797).