feat(libexpr): mark start of infinite recursion in error traces

[roberth]

Motivation

Trace useful. Found a need for this again when debugging the tracing cache (STF)

When Nix reports a trace for cyclic expression, it will now let you discern the cycle from the code that leads up to the cycle.

Context

• Improved after taking from Stack trace improvements #8623

Assisted-By: Claude Code (Claude Opus 4.7)
Assisted-By: Claude Code (Claude Opus 5)

---

Add 👍 to pull requests you find important.

The Nix maintainer team uses a GitHub project board to schedule and track reviews.

[xokdvium]

Suggested change

	template<typename... Args>
	explicit InfiniteRecursionError(EvalState & state, const Value * v, const Args &... args)
	: CloneableError(state, args...)
	, v(v)
	template<typename... Args>
	explicit InfiniteRecursionError(EvalState & state, const Value * v, Args &&... args)
	: CloneableError(state, std::forward<Args>(args)...)
	, v(v)

But it probably doesn't matter much.

[roberth]

Doesn't do anything atm. Would take 4 more changes:

#	Link	Signature	File
1	EvalState::error<T>	const Args &... args	eval.hh:488
2	EvalErrorBuilder<T> ctor	const Args &... args	eval-error.hh:129
3	InfiniteRecursionError ctor	const Args &... args	eval-error.hh:75
4	EvalBaseError ctor	const Args &... formatArgs	eval-error.hh:39
5	BaseError ctor	Args &&... args → HintFmt(...)	error.hh:142 ✅ already forwards

Also errors are supposedly not in the hot path (tryEval aside)

[xokdvium]

That's still good to do locally, even if we are not forwarding everything up the call stack.

[xokdvium]

I can improve this in a follow-up.

[roberth]

I already asked Claude to do it #16309, if it's any help :)