[WIP] cl: preserve linknames for main test variants

[visualfc]

Summary

Preserve preloaded //go:linkname entries for package main test variants.

Behavior

• A main.* target declared in ordinary package source is rewritten to the package import path when llgo builds the test variant.
• A main.* target declared in a _test.go source file is intentionally left unchanged. Test variants do not define main.* symbols, so this remains unsupported.
• Package-scoped linkname directives follow the same source-file distinction.

Tests

• GOCACHE=/tmp/llgo-bug-gocache go test ./cl -run TestParsePkgSyntaxCollectsLinknames -count=1
• GOCACHE=/tmp/llgo-bug-gocache go run ./cmd/llgo test -v ./cmd/demo/bug

History

The branch is a single commit: c7a592c59 cl: preserve linknames for main test variants.

[fennoai]

Review: fix main test linkname

This is a small, focused change that records an extra go:linkname alias under the raw import path for main packages compiled as test variants (go test emits symbols under the import path rather than the canonical main. prefix). The logic is correct, and the main.-prefix rewrite in testMainLinkname is safely anchored (targets like C.foo, example.com/pkg.X, or mainfoo.bar are untouched). No security or performance concerns — the added work is guarded so non-main packages short-circuit before any map access.

The findings below are maintainability and documentation/test-clarity suggestions; none block merge.

[visualfc]

Addressed the review suggestions in c7c50dd:

• centralized the raw-main-path predicate in rawMainPathOf
• renamed and documented the target rewrite helper
• split coverage between declaration-attached and cross-file floating //go:linkname directives

Verified with:

• GOCACHE=/tmp/llgo-bug-gocache go test ./cl -run TestParsePkgSyntaxCollectsLinknames -count=1
• GOCACHE=/tmp/llgo-bug-gocache go run ./cmd/llgo test -v ./cmd/demo/bug

[github-actions]

LLGo baseline benchmarks

f093bba768e4 | workflow run | long-term charts

Program measurements

Platform	Workload	File size	vs base	Build	vs base	Run	vs base
Linux	cprintf	19288 B	0 B / +0.0%	356.137 ms	-28.03 ms / -7.3% (better)	1.318 ms	+31.13 us / +2.4% (worse)
Linux	fmtprintf	1859320 B	0 B / +0.0%	2.853 s	+66.16 ms / +2.4% (worse)	3.261 ms	-131.9 us / -3.9% (better)
Linux	println	68776 B	0 B / +0.0%	356.535 ms	+4.079 ms / +1.2% (worse)	1.634 ms	-116.7 us / -6.7% (better)
macOS	cprintf	84672 B	0 B / +0.0%	628.239 ms	+34.5 ms / +5.8% (worse)	5.003 ms	-841.3 us / -14.4% (better)
macOS	fmtprintf	1892448 B	0 B / +0.0%	3.500 s	+98.56 ms / +2.9% (worse)	14.667 ms	-298.3 us / -2.0% (better)
macOS	println	121360 B	0 B / +0.0%	560.880 ms	-4.808 ms / -0.8% (better)	5.364 ms	-1.119 ms / -17.3% (better)

Core language and compiler benchmarks

Platform	Benchmark	ns/op	vs base
Linux	BenchmarkLookupPCRandom	13.240 ns/op	-0.01 ns/op / -0.1% (better)
Linux	BenchmarkMergeCompilerFlags	153.800 ns/op	+2 ns/op / +1.3% (worse)
Linux	BenchmarkMergeLinkerFlags	96.170 ns/op	+1.07 ns/op / +1.1% (worse)
Linux	BenchmarkChannelBuffered	33.980 ns/op	-0.03 ns/op / -0.1% (better)
Linux	BenchmarkChannelHandoff	30003 ns/op	+1181 ns/op / +4.1% (worse)
Linux	BenchmarkDefer	48.840 ns/op	+0.42 ns/op / +0.9% (worse)
Linux	BenchmarkDirectCall	1.557 ns/op	-0.001 ns/op / -0.1% (better)
Linux	BenchmarkGlobalRead	1.557 ns/op	-0.001 ns/op / -0.1% (better)
Linux	BenchmarkGlobalWrite	2.481 ns/op	+0.001 ns/op / +0.04032% (worse)
Linux	BenchmarkGoroutine	34787 ns/op	-682 ns/op / -1.9% (better)
Linux	BenchmarkInterfaceCall	8.141 ns/op	+0.042 ns/op / +0.5% (worse)
Linux	BenchmarkRuntimeGetG	2.180 ns/op	-0.001 ns/op / -0.04585% (better)
macOS	BenchmarkLookupPCRandom	13.540 ns/op	-4.23 ns/op / -23.8% (better)
macOS	BenchmarkMergeCompilerFlags	135.900 ns/op	-39.2 ns/op / -22.4% (better)
macOS	BenchmarkMergeLinkerFlags	97.610 ns/op	-44.29 ns/op / -31.2% (better)
macOS	BenchmarkChannelBuffered	25.350 ns/op	-7.93 ns/op / -23.8% (better)
macOS	BenchmarkChannelHandoff	8357 ns/op	-1198 ns/op / -12.5% (better)
macOS	BenchmarkDefer	39.440 ns/op	-7.46 ns/op / -15.9% (better)
macOS	BenchmarkDirectCall	1.149 ns/op	-0.106 ns/op / -8.4% (better)
macOS	BenchmarkGlobalRead	1.162 ns/op	-0.164 ns/op / -12.4% (better)
macOS	BenchmarkGlobalWrite	1.197 ns/op	-0.181 ns/op / -13.1% (better)
macOS	BenchmarkGoroutine	66942 ns/op	+898 ns/op / +1.4% (worse)
macOS	BenchmarkInterfaceCall	5.124 ns/op	-0.596 ns/op / -10.4% (better)
macOS	BenchmarkRuntimeGetG	2.707 ns/op	-0.643 ns/op / -19.2% (better)

Compared with cf91afb98fca measured in the same runner job.

[codecov]

Codecov Report

❌ Patch coverage is 96.29630% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
internal/build/build.go	66.66%	0 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!