Could we run less predictive frontend pipelines?
Context
I noticed that 100% of backend predictive tests (i.e. rspec predictive) also run frontend predictive tests (i.e. jest predictive), causing longer predictive pipelines.
Goal
Investigate whether we could bring the number of predictive frontend pipelines to less than 100%.
For instance, do we need to run predictive frontend tests on code-backstage-patterns, which seems quite a broad pattern?
# .code-patterns + .backstage-patterns
.code-backstage-patterns: &code-backstage-patterns
- ".{eslintrc.yml,eslintignore,gitattributes,nvmrc,prettierrc,stylelintrc,yamllint}"
- ".browserslistrc"
- ".stylelintrc"
- "{,ee/,jh/}{app,bin,config,db,elastic,generator_templates,gems,haml_lint,lib,locale,public,scripts,storybook,symbol,vendor}/**/*"
- "{,jh/}Gemfile{,.lock}"
- "{package.json,yarn.lock}"
- "*_VERSION"
- "babel.config.js"
- "config.ru"
- "Dockerfile.assets"
- "jest.config.{base,integration,unit}.js"
- "keeps/**/*"
- "Rakefile"
- "tests.yml"
- "vendor/assets/**/*"
# Auto-generated files
- "doc/api/graphql/reference/*"
# CI changes
- ".gitlab-ci.yml"
- ".gitlab/ci/**/*"
# Mapped patterns (see tests.yml)
- "data/whats_new/*.yml"
- "doc/index.md"
# Backstage changes
- "Dangerfile"
- "danger/**/*"
- "{,ee/,jh/}fixtures/**/*"
- "{,ee/,jh/}rubocop/**/*"
- "{,ee/,jh/}spec/**/*"
- "{,spec/}tooling/**/*"
Could we instead have a similar to .backend-patterns, but for frontend?
Important
We do not want to miss Jest Predictive failures.
Edited by David Dieulivol