This PR refactors the onboarding PR logic into its own subdirectory and redesigns it to use plain markdown instead of html. Plain markdown in PR bodies is a requirement for adding VSTS Support ( #1049 ).
This refactor means that createPr now takes `labels` as a new parameter so that the GitLab API can include them with the MR creation request, so save one API call. This also fixes#985 as a result.
99% of the time we should not Renovate forked repositories. Previously, we skipped onboarding them unless a renovate.json was present. However, that meant that if someone forked a repository where the source was using Renovate, then the fork gets renovated automatically too. After this change, forked repositories need to set renovateFork=true in their renovate.json
Closes#940
BREAKING CHANGE: Forked repositories now need to configure renovateFork=true in renovate.json
This is instead of autodetecting lib or app during onbiarding. This means some package files might be all pinned while others in the same monorepo can be not. This also simplifies the onboarding logic.
This commonly applies to monorepos where the same dependency may be present in multiple.json files. Instead of using the group name to describe the PR, it uses the dependency name.
Closes#848
This is a failsafe against Renovate bugs and potentially also mistakes on the user’s part. Before deleting/pruning any branches, Renovate will rename the PR’s title so that it should not block future PRs.
Closes#989
* refactor: replace github code search with getTree/manual search (#944)
Previously Renovate was using the GitHub code search API once for each filename we are looking for. Instead, we now retrieve a list of files in the repository and cache it so it can be reused for filename matches.
Closes#939
* fix: use recursive=true
* fix: reset fileList for every repo (github)
* feat: implement findFilePaths on gitlab (#945)
* refactor: do not check manually for root package.json
* fix: reset fileList for every repo (gitlab)
* fix tests
Previously Renovate was using the GitHub code search API once for each filename we are looking for. Instead, we now retrieve a list of files in the repository and cache it so it can be reused for filename matches.
Closes#939
* Revert "feat: support docker and meteor onboarding without npm"
This reverts commit 437145b901.
# Conflicts:
# test/workers/repository/onboarding.spec.js
* chore: downgrade npm cache message from debug to trace