Skip to content

Go Pool 40

Cervantes edited this page Jun 17, 2026 · 3 revisions

Go Pool 40

Tracking issue: https://github.com/cervantesh/cervo-mutant/issues/13

This page summarizes the anti-fitting repository pool used to keep CervoMutant from overfitting to Cobra.

Source documents:

  • docs/evaluations/go-repo-pool-40.md
  • docs/evaluations/go-repo-pool-40.json
  • docs/evaluations/2026-05-26-go-pool-40-smoke.md

Smoke Result

Metric Count
Repositories in manifest 40
Shallow clone succeeded 40
Baseline go test <target> passed within timeout 23
Baseline timed out at 30s 4
CervoMutant dry-run passed 36
Dry-run failed 4

Selected 12 Mutation Sample

Repo Lane Domain Mutants Killed Survived Not covered Score
cobra tuning cli 25 10 14 1 41.67%
pflag validation cli 25 12 5 8 70.59%
moby holdout containers 25 0 1 24 0.00%
hugo holdout static-site 25 19 2 4 90.48%
prometheus holdout observability 25 5 0 20 100.00%
terraform holdout iac 25 11 3 11 78.57%
grpc-go validation networking 25 14 1 10 93.33%
echo validation web 25 18 2 5 90.00%
logrus tuning logging 25 10 4 10 66.67%
validator tuning validation 25 0 0 25 0.00%
decimal tuning numeric - - - - timeout
gjson validation parser 25 20 5 0 80.00%

Conclusion: Cobra is not representative enough to tune alone. The first sample already shows large differences in coverage shape, survivor rate, runtime, and domain behavior.

Clone this wiki locally