testing: high start up cost with high -count and no tests #49050

What version of Go are you using (go version)?

$ go version
go version devel +b7a85e0003 linux/amd64

Does this issue reproduce with the latest release?

Yes

What operating system and processor architecture are you using (go env)?

$ go env
GO111MODULE=""
GOARCH="amd64"
GOBIN=""
GOCACHE="/usr/local/google/home/mpratt/.cache/go-build"
GOENV="/usr/local/google/home/mpratt/.config/go/env"
GOEXE=""
GOEXPERIMENT=""
GOFLAGS=""
GOHOSTARCH="amd64"
GOHOSTOS="linux"
GOINSECURE=""
GOMODCACHE="/usr/local/google/home/mpratt/go/pkg/mod"
GONOPROXY=""
GONOSUMDB=""
GOOS="linux"
GOPATH="/usr/local/google/home/mpratt/go"
GOPRIVATE=""
GOPROXY="https://proxy.golang.org,direct"
GOROOT="/usr/lib/google-golang"
GOSUMDB="sum.golang.org"
GOTMPDIR=""
GOTOOLDIR="/usr/lib/google-golang/pkg/tool/linux_amd64"
GOVCS=""
GOVERSION="devel +b7a85e0003"
GCCGO="gccgo"
AR="ar"
CC="clang"
CXX="clang++"
CGO_ENABLED="1"
GOMOD="/usr/local/google/home/mpratt/src/scratch/go_bench/go.mod"
CGO_CFLAGS="-g -O2"
CGO_CPPFLAGS=""
CGO_CXXFLAGS="-g -O2"
CGO_FFLAGS="-g -O2"
CGO_LDFLAGS="-g -O2"
PKG_CONFIG="pkg-config"
GOGCCFLAGS="-fPIC -m64 -pthread -fno-caret-diagnostics -Qunused-arguments -fmessage-length=0 -fdebug-prefix-map=/tmp/go-build1003331103=/tmp/go-build -gno-record-gcc-switches"

What did you do?

Run a benchmark many, many times. (The intention here is to run benchmarks for hours)

https://play.golang.org/p/tKSzmTOK_-m

go test -bench=. -count=10000000

What did you expect to see?

Benchmarks start running immediately.

What did you see instead?

Nothing seems to happen for ~30s before the benchmarks start running. The issue appears to be that runTests runs its count loop 10000000 times, even though there are no tests to run.

One quick fix may be for runTests to exit immediately if len(tests) == 0, though I've distilled this issue down from my initial problem of a long time waiting for benchmarks to start in a package that has tests which I skip with -run=none. It would be nice to fix that case too. For that, perhaps the first pass through runTests records if it ran any tests. If not, the next iteration immediately breaks.