Improve branch cache logic

[rarkins]

Describe the proposed change(s).

The branch cache should have this function, which is called at the start of branch processing each run: syncBranchCache(branchName, branchSha, baseBranchName, baseBranchSha)

It's logic will be like this:

• If baseBranchName has changed, delete isModified
• If branchSha has changed, delete isModified, isConflicted, isBehindBaseBranch, branchFingerprint
• If baseBranchSha has changed, delete isConflicted, isBehindBaseBranch
• Now sync the fields (or create the cache for the branch if it didn't exist)

During branch processing we can call setBranchModified(branchName, isModified), setBranchConflicted, setBranchBehindBase, setBranchFingerprint. These don’t need SHAs because we assume they were sync’d earlier in the branch processing.

Towards the end of branch processing, if we have pushed a commit (either branch creation or branch updating) then we call this function: setBranchCommit(branchName, baseBranchName, branchSha, baseBranchSha, branchFingerprint). It should:

• If a new branch, set all fields
• If an existing branch, update all fields
• Set parentSha to equal baseBranchSha
• Set isModified, isConflicted and isBehindBaseBranch to false

Note: parentSha should not be reset during syncBranchCache() - only during setBranchCommit().

We should remove all git-level caching. It's OK for us to "lose" such existing cache values for one run while we populate the branch cache using the new logic.

[rarkins]

This should also fix #17528 because during syncBranchCache() the value for isX will all be reset once a new commit is detected.

[RahulGautamSingh]

Should we also invalidate isModified when config.gitIgnoredAuthors change?

[rarkins]

Yes, but that should be a rare edge case!

[renovate-release]

🎉 This issue has been resolved in version 32.219.0 🎉

The release is available on:

• GitHub release
• 32.219.0

Your semantic-release bot 📦🚀