How does test coverage work in Go? Explain covermode, -coverpkg, and coverage for integration tests.

The toolchain rewrites source to increment counters per basic block.

go test -coverprofile=cover.out ./...
go tool cover -func=cover.out      # per-function percentages
go tool cover -html=cover.out      # annotated source

go test -covermode=atomic -coverpkg=./... ./...

• -covermode: set (was it run?), count (how many times), atomic (count, safe under concurrency — the default when -race is on).
• -coverpkg=./...: by default coverage counts only the package under test; this counts code in other packages exercised by the tests too (e.g. handlers tested via an e2e package).

Integration coverage (Go 1.20+): build a real binary with coverage instrumentation, run it, and collect data:

go build -cover -o app ./cmd/app
GOCOVERDIR=./covdata ./app --run-scenario   # writes counter files on exit
go tool covdata percent -i=./covdata
go tool covdata textfmt -i=./covdata -o=cover.out

Interviewer is looking for: coverage is a tool for finding untested code, not a quality target — 100% line coverage says nothing about asserting the right behavior or edge cases. Error branches and concurrency paths are what usually lack coverage. Mutation testing or fuzzing find what coverage misses.