Merge testing repo into this monorepo

[zhongkechen]

Issue #, if available:

Fix aws/aws-durable-execution-sdk-python-testing#226

Description of changes:

• To preserve the full history of testing repo, all the commits from that repo are included in this PR
  • so lint-commits failed as expected
• ECR workflow is not tested (we can test this when we do next release)

Review only the last 3 commits in this PR

• 3c7c79b
• 3399283

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[yaythomas]

we could keep the version history with some re-writing doing git-filter-repo.

We could use built-in filter branch git filter-branch but this is deprecated.

Instead, something like:

# Install git-filter-repo if you don't have it:
pip install --user git-filter-repo
# or: brew install git-filter-repo

mkdir -p /tmp/merge-testing
cd /tmp/merge-testing

In /tmp/merge-testing/ — clone both repos fresh

git clone https://github.com/aws/aws-durable-execution-sdk-python-testing.git testing-rewrite
git clone https://github.com/aws/aws-durable-execution-sdk-python.git monorepo

In /tmp/merge-testing/testing-rewrite/ — rewrite history

cd /tmp/merge-testing/testing-rewrite

# Drop testing's release tags so they don't collide with monorepo tags.
# Skip this line if you want to keep them.
git tag -l | xargs -r git tag -d

# The rewrite. Every commit's paths get prefixed with the target subdir.
git filter-repo --to-subdirectory-filter packages/aws-durable-execution-sdk-python-testing

# Sanity checks
git -P log --oneline | wc -l
# expect ~143

git -P log --name-only --pretty=format: | sort -u | head
# every line should start with packages/aws-durable-execution-sdk-python-testing/

In /tmp/merge-testing/monorepo/ — build the new branch

cd /tmp/merge-testing/monorepo

# Pull the rewritten testing history in as a local remote
git remote add testing ../testing-rewrite
git fetch testing

# Also fetch the existing PR branch so we can grab the workflow commit
git fetch origin pull/423/head:pr-423

# Build merge_testing as: testing's full history, replayed linearly on top of main
git checkout -B merge_testing testing/main
git rebase --onto origin/main --root

# Bring over the one real monorepo-side change (the workflow file)
git cherry-pick 3399283e

# Verify history-follow works end-to-end
git -P log --follow --oneline packages/aws-durable-execution-sdk-python-testing/README.md | tail
# Should show commits all the way back to testing's "Initial commit".

git -P log --oneline origin/main..HEAD | wc -l
# ~144 commits (143 testing + 1 workflow cherry-pick)

Still in the monorepo — push

# Replaces the existing PR branch. Everyone reviewing the PR will see the rebuilt history.
git push origin merge_testing --force-with-lease

The result: a linear merge_testing branch with all 143 testing commits replayed on top of main, every commit looking as if files always lived under packages/aws-durable-execution-sdk-python-testing/. No merge commit, and git
log --follow / git blame walk back to testing's initial commit.

[yaythomas]

see suggestion re keeping version history

[zhongkechen]

What was the real concern?

Drop testing's release tags so they don't collide with monorepo tags.
Skip this line if you want to keep them.
git tag -l | xargs -r git tag -d

tags are not be imported in this PR

The rewrite. Every commit's paths get prefixed with the target subdir.
git filter-repo --to-subdirectory-filter packages/aws-durable-execution-sdk-python-testing

renaming isn't necessary as the git history command works across rename properly

[yaythomas]

the PR as it stand right now with merge commits forbidden requires a squash which will abandon the history of testing library, and can't rebase due to conflicts. The important commits in this PR here effectively has
i) move commit
ii) "merge" commit
iii) workflow tweaks commit

goal is to 1) keep history (important) and 2) avoid merge commit (nice to have).

as far as I can see the options are:

A) rebase the commits you have here into main (but here the rename boundary remains, where git log <dir> stops at the import commit, but single-file git log --follow walks back via rename detection). Squash (i) move + (iii) workflow tweaks into one "chore: import testing repo" commit at the tip.
B) the filter-repo + rebase I outline above (a bunch of commands, but no conflct resolution, I think, because the paths don't overlap?), which will keep the individual file and dir history
C) enable merge commit on this repo (admin required) and then there will be the PR merge commit and inside that another merge commit (ii from above).

Even if we did go C, to avoid the "double" merge commit you'd pretty much have to do the work of A.

[zhongkechen]

Rebase will make the history unreadable. I'll probably abandon the history or create a merge commit. If merge by GitHub will create a new commit, I'll push directly to main branch