[bercedev]

Fixes #61.

A #[doc = ...] whose value isn't a string literal, such as #[doc = include_str!("...")], used to make the derive panic with "not implemented". Since #69, which I wrote to skip #[doc(alias = "...")] and #[doc(hidden)], it's skipped silently instead. That's arguably worse here: a struct then gets no Display impl at all, and the error only shows up wherever the type is formatted, as "Foo doesn't implement std::fmt::Display".

display() now returns a compile error, spanned on the value, when a #[doc = ...] value isn't a string literal:

error: displaydoc can only read doc comments that are string literals; give the format with #[displaydoc("...")] instead
 --> tests/ui/non_string_doc_expr.rs:4:9
  |
4 | #[doc = include_str!("non_string_doc_expr.txt")]
  |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

A type with #[displaydoc("...")] is unaffected, since that attribute is read first. #69 isn't released yet, so compared with 0.2.7 this replaces the panic with a proper error.

Tests:

• New UI test tests/ui/non_string_doc_expr.rs with its .stderr, added to compile_tests.rs. On master the file compiles, because the derive skips the attribute and emits nothing, so the compile_fail case fails there.
• cargo test on stable 1.99.1 passes (the UI tests are ignored off nightly). The UI tests pass on nightly-2026-10-22.
• cargo fmt --check and cargo clippy --all-targets are clean.
• CHANGELOG.md has an entry under Unreleased.

[Manishearth]

Suggested change

	"displaydoc can only read doc comments that are string literals; give the format with #[displaydoc(\"...\")] instead",
	"displaydoc can only read doc comments that are string literals; supply the Display text with #[displaydoc(\"...\")] instead",

[bercedev]

applied it, and updated tests/ui/non_string_doc_expr.stderr to match. the ui tests pass on nightly