test/goroot: remove stale bug388 expectation

[MeteorsLiu]

Removes the stale Go 1.26 not-applicable entry for fixedbugs/bug388.go after #2274 made cmd/compile syntax diagnostics authoritative.

The implementation includes:

• Removes fixedbugs/bug388.go from test/goroot/notapplicable.yaml.
• Relies on the loader behavior merged in fix: prefer compiler syntax diagnostics #2274 and introduces no production code changes.
• Verifies the case with the Go 1.26.5 GOROOT errorcheck runner and the expectation configuration tests.

This keeps the GOROOT expectations aligned with current LLGo behavior and prevents a passing case from being reported as an unexpected success.

[fennoai]

Review: suppress type errors after syntax errors

Focused, correct change. Verified that the diagnostic suppression cannot cause a failed build to appear successful: buildSSAPkgs returns fmt.Errorf("cannot build SSA for packages") whenever len(errs) > 0, and package pass/fail is driven by IllTyped (set in internal/packages/load.go whenever len(lpkg.Errors) > 0), independent of which diagnostics are printed. Suppression is correctly scoped per-package (hasParseError is reset each errPkg iteration), and a pure type-error package still prints its diagnostics. No security or performance concerns (this runs only on the error path).

Two maintainability suggestions inline. Nothing blocking.

• The added gopackages alias is justified — the existing internal/packages alias re-exports Error/Package types but not the ParseError/TypeError kind constants. (Optionally, re-exporting those two constants from internal/packages would let this file keep a single packages identifier.)

[fennoai]

Two small maintainability points on this block:

1. The first loop can collapse into a single pass — slices is already imported, so hasParseError := slices.ContainsFunc(errPkg.Errors, func(e packages.Error) bool { return e.Kind == gopackages.ParseError }) avoids iterating errPkg.Errors twice.
2. Silently dropping TypeError output when a ParseError is present is a deliberate decision (type errors are cascading noise once a file fails to parse). A one-line comment explaining the rationale would prevent a future reader from mistaking it for accidental error swallowing.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[github-actions]

LLGo baseline benchmarks

5787cd3ec27a | workflow run | long-term charts

Program measurements

Platform	Workload	File size	vs base	Build	vs base	Run	vs base
Linux	cprintf	18456 B	+0.0%	285.548 ms	-1.8% (better)	1.283 ms	-3.5% (better)
Linux	fmtprintf	1829968 B	+0.0%	2.978 s	+0.3% (worse)	3.086 ms	-1.0% (better)
Linux	println	68008 B	+0.0%	289.121 ms	+0.6% (worse)	1.574 ms	-0.7% (better)
macOS	cprintf	84672 B	+0.0%	365.997 ms	-49.7% (better)	4.917 ms	-31.0% (better)
macOS	fmtprintf	1869328 B	+0.0%	5.181 s	+85.1% (worse)	16.491 ms	+27.4% (worse)
macOS	println	121200 B	+0.0%	361.533 ms	-44.5% (better)	3.953 ms	-24.9% (better)

Core language and compiler benchmarks

Platform	Benchmark	ns/op	vs base
Linux	BenchmarkLookupPCRandom	13.310 ns/op	+0.2% (worse)
Linux	BenchmarkMergeCompilerFlags	150.800 ns/op	-0.2% (better)
Linux	BenchmarkMergeLinkerFlags	94.310 ns/op	-0.4% (better)
Linux	BenchmarkChannelBuffered	33.650 ns/op	-0.2% (better)
Linux	BenchmarkChannelHandoff	26707 ns/op	-1.2% (better)
Linux	BenchmarkDefer	49.230 ns/op	+0.0%
Linux	BenchmarkDirectCall	1.556 ns/op	+0.0%
Linux	BenchmarkGlobalRead	1.557 ns/op	+0.1% (worse)
Linux	BenchmarkGlobalWrite	2.477 ns/op	-0.1% (better)
Linux	BenchmarkGoroutine	41302 ns/op	+32.1% (worse)
Linux	BenchmarkInterfaceCall	8.095 ns/op	-0.0% (better)
Linux	BenchmarkRuntimeGetG	2.181 ns/op	+0.0% (worse)
macOS	BenchmarkLookupPCRandom	12.800 ns/op	-2.2% (better)
macOS	BenchmarkMergeCompilerFlags	167.600 ns/op	+30.6% (worse)
macOS	BenchmarkMergeLinkerFlags	91.760 ns/op	+4.6% (worse)
macOS	BenchmarkChannelBuffered	24.260 ns/op	-29.0% (better)
macOS	BenchmarkChannelHandoff	7557 ns/op	-24.2% (better)
macOS	BenchmarkDefer	29.710 ns/op	-25.7% (better)
macOS	BenchmarkDirectCall	1.088 ns/op	-2.1% (better)
macOS	BenchmarkGlobalRead	1.292 ns/op	-4.0% (better)
macOS	BenchmarkGlobalWrite	1.179 ns/op	-3.6% (better)
macOS	BenchmarkGoroutine	28837 ns/op	-49.7% (better)
macOS	BenchmarkInterfaceCall	4.957 ns/op	-6.0% (better)
macOS	BenchmarkRuntimeGetG	2.502 ns/op	-14.4% (better)

Compared with 6670dae3884d measured in the same runner job.

[MeteorsLiu]

Superseded by #2265, which now includes the fixedbugs/bug388.go expectation cleanup together with the filtered go list diagnostic handling.