fix(dotnet): stop duplicating failures on failing test runs (#2501)

[Tailoo]

Summary

Fixes #2501 — on failing dotnet test runs the orchestrator prepended the full raw stdout ahead of the filtered summary. The Failed Tests: section already reproduces each failure (name + message + clipped stack) parsed from TRX/console, so every failure was printed twice — inflating output +65% vs raw, scaling linearly with failure count.

Fix

Gate the raw-stdout prepend behind a new test_needs_raw_fallback:

• Skip the prepend when the structured Failed Tests: section already carries detail → dedup.
• Keep it when the filter is blind: no failures parsed (build failure / crash where nothing reaches failed_tests), or a parsed failure has empty detail (e.g. a self-closing <UnitTestResult> with no <ErrorInfo>). Nothing is lost.

build / restore / passthrough keep prior behavior (needs_raw_fallback = true).

Why this shape

The bug lived in run_dotnet_with_binlog — the one dotnet layer with no test coverage (it spawns real dotnet). Consistent with the existing dotnet test pattern, the decision is extracted into two pure, unit-tested fns rather than adding a net-new orchestration/e2e harness (no dotnet filter has one):

• test_needs_raw_fallback(&summary) -> bool
• compose_failure_output(success, needs_raw_fallback, stdout, stderr, filtered) -> String

Regression analysis (traced through code)

Scenario	failed_tests	Behavior	Result
Normal failing test (detail parsed)	populated, detail present	drop raw prepend	deduped, detail preserved
Failure with empty details	populated, detail empty	keep raw fallback	raw message not lost
Compile/build failure, no tests ran	empty	keep raw fallback	build errors preserved
Crash / no TRX	empty	keep raw fallback	raw preserved
counts_unavailable	empty	keep raw fallback	raw preserved

Warnings: / Errors: sections come from the build summary, independent of the prepend → still shown once.

Tests

7 new unit tests (all green), covering the dedup decision and every fallback branch. cargo fmt + cargo clippy --all-targets clean, cargo test --all passes.

🤖 Generated with Claude Code

[CLAassistant]

All committers have signed the CLA.

[aeppling]

Hey @Tailoo , thanks for addressing this issue.

Another concurrent PR #2511 , in favor of this one for the cleaner factoring.

Blocker before merge: test_needs_raw_fallback drops the raw fallback whenever every listed failure carries detail, but never checks the list is complete against summary.failed.

Fix: keep the raw fallback when the structured section is numerically incomplete, add a failed_tests.len() < summary.failed clause to the condition.

[aeppling]

Hey, this look good to me, thanks for contributing to RTK :)