injected code results in errors that emacs hates

[brson]

For example:

<core-macros>:3:10: 3:61 note: in expansion of #error

Emacs wants to load <core-macros> and interrupts your workflow to prompt for it. Injected code should come from files on disk (like intrinsic.rs) and we should use the correct file path when inserting them into the AST, so that the error messages work.

Alternately, we can come up with another way to indicate injected code so that the diagnostic emitter can generate error messages that emacs doesn't complain about. This could be a better solution, since getting dumped into core_macros.rs may not be so useful.

[graydon]

intrinsic.rs is moving to core now that we have #[lang] items; core-macros will also go away when we figure out a mechanism for pulling in syntax extensions and/or macro definitions in other crates.

In general I think "pointing to source files on disk" isn't going to work so well, since it implies the end-user has a source-file. I'd prefer not to be shipping .rs files as part of the end-user distribution / make install file set.

[catamorphism]

Bumping to 0.7; this seems like low priority compared to the million and one traits bugs.

[catamorphism]

Not critical for 0.7. Nominating for milestone 5, production-ready.

[graydon]

should say something that does not make emacs mad, but there is no file to point to.

[graydon]

accepted for production-ready milestone

[emberian]

still relevant

[bblum]

these macro errors produce a "note: expansion site" message, right? perhaps the emacs plugin could be configured to recognise a string such as "" and look for the expansion site message below.

[pnkfelix]

part of #8793

[pnkfelix]

assigning P-low.

[steveklabnik]

Triage: I don't use emacs, but these errors are the same, so I'm guessing this is still the same.

[steveklabnik]

Moved to rust-lang/rust-mode#4