Fix basic static checks comparing same commit with itself by jedcunningham · Pull Request #60194 · apache/airflow

jedcunningham · 2026-01-06T23:42:42Z

The prek command was comparing github.sha with itself, which checked no files. Now uses merge-base to check all files changed in the PR against the base branch.

This ensures static checks are actually ran in PRs like #60168.

The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

Copilot

Pull request overview

Fixes a bug in the basic-tests workflow where static checks were comparing the same commit (github.sha) with itself, resulting in no files being checked. The fix introduces a merge-base calculation to properly identify all changed files in a PR by comparing against the base branch.

Key changes:

Implements merge-base calculation to find the common ancestor between PR and base branch
Updates static check command to compare merge-base commit against HEAD instead of comparing github.sha with itself
Increases fetch depth and adds fallback logic to handle cases where merge-base isn't found in shallow clones

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

bugraoz93 · 2026-01-07T00:04:18Z

+ git fetch origin "$BASE_BRANCH" --depth=50
+
+ # Try to find merge-base, deepen if not found
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
+
+ if [ -z "$MERGE_BASE" ]; then
+ echo "Merge-base not found with depth=50, fetching more history..."
+ git fetch --deepen=450
+ git fetch origin "$BASE_BRANCH" --deepen=450
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
+ fi
+
+ echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
+ env:
+ BASE_BRANCH: ${{ github.base_ref || 'main' }}


Suggested change
git fetch origin "$BASE_BRANCH" --depth=50
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=50, fetching more history..."
git fetch --deepen=450
git fetch origin "$BASE_BRANCH" --deepen=450
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
git fetch origin "$BASE_BRANCH" --depth=$FETCH_DEPTH
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=$FETCH_DEPTH, fetching more history..."
git fetch --deepen=$DEEPEN
git fetch origin "$BASE_BRANCH" --deepen=$DEEPEN
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
FETCH_DEPTH: 50
DEEPEN: 450

Small nit, maybe we can make the values parameterised for both --depth and --deepen from env :)

amoghrajesh

Wow, thats a strange one

potiuk

Strange one .. I wonder when it happened and how it did not work (I am sure it worked before). But it can be done way simpler.

All you need it is to add missing ^ in --from-ref

At the moment you do this check - github.sha already contains merge commit - this is how pull-request workflow works.

https://docs.github.com/en/actions/reference/workflows-and-actions/events-that-trigger-workflows#pull_request

Note that GITHUB_SHA for this event is the last merge commit of the pull request merge branch. If you want to get the commit ID for the last commit to the head branch of the pull request, use github.event.pull_request.head.sha instead.

potiuk · 2026-01-07T08:39:15Z

Superseded by #60202

potiuk · 2026-01-07T08:42:11Z

This is precisely why we needed 2 depth - because in pull_request the github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased).

potiuk · 2026-01-07T09:20:57Z

I merged my fix and rebased #60168 to verify that it works (it should fail)

potiuk · 2026-01-07T09:21:11Z

Nice catch @jedcunningham BTW :)

potiuk · 2026-01-07T09:24:20Z

Yep. Nicely failed now https://github.com/apache/airflow/actions/runs/20776566467/job/59663769292?pr=60168

jedcunningham · 2026-01-07T18:52:40Z

Nice, I'd considered just adding "^" but somehow I confused myself thinking that would only go 1 commit up - that if we had many commits on the branch it'd only look at the last one. I guess it must squash? Anyways, glad its working.

edit:

github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased)

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

potiuk · 2026-01-07T22:07:27Z

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

Correct. This is actually what we very heavily bank on (this why I knew it by heart).
Basically whole selective checks code is based on the fact that we have "merge" commit passed as github.sha and it's first parent is base branch.

https://github.com/apache/airflow/blob/main/dev/breeze/src/airflow_breeze/commands/ci_commands.py#L163:

defget_changed_files(commit_ref: str|None) ->tuple[str, ...]:
ifcommit_refisNone:
return ()
cmd= [
"git",
"diff-tree",
"--no-commit-id",
"--name-only",
"-r",
commit_ref+"^",
commit_ref,
]
result=run_command(cmd, check=False, capture_output=True, text=True)
ifresult.returncode!=0:
get_console().print(
f"[warning] Error when running diff-tree command [/]\n{result.stdout}\n{result.stderr}"
)
return ()
changed_files=tuple(result.stdout.splitlines()) ifresult.stdoutelse ()
get_console().print("\n[info]Changed files:[/]\n")
get_console().print(changed_files)
get_console().print()
returnchanged_files

potiuk · 2026-01-07T22:18:43Z

BTW. There is also another interesting thing I learned with prek

There is a very nice shortcut to not have to find merge-base.

git diff main..your_pr_branch
git diff main...your_pr_branch

Spot the difference .... there are THREE dots in second case. And the ... is actually showing the changes in your branch from the merge-base

And ... (pun intended) there are few other gotchas: https://darekkay.com/blog/git-commit-ranges/#git-diff

For example git diff main...your-branch is not reverse of git diff your-branch...diff as you would suspect. The first one shows your changes since merge-base, the second shows SURPRISE -> all the changes in main (!) since merge-base 😱 😱 😱 😱 😱 😱 😱 😱 😱

Git is wild.

potiuk · 2026-01-07T22:20:40Z

So basically if you want to have prek to execute on "only your changes" from a branch you have PR to main with - do this (without running any of git merge-base thingies):

prek --from-ref maim

(we have it in our docs)

jedcunningham · 2026-01-06T23:42:42Z

The prek command was comparing github.sha with itself, which checked no files. Now uses merge-base to check all files changed in the PR against the base branch.

This ensures static checks are actually ran in PRs like #60168.

The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

Copilot

Pull request overview

Fixes a bug in the basic-tests workflow where static checks were comparing the same commit (github.sha) with itself, resulting in no files being checked. The fix introduces a merge-base calculation to properly identify all changed files in a PR by comparing against the base branch.

Key changes:

Implements merge-base calculation to find the common ancestor between PR and base branch
Updates static check command to compare merge-base commit against HEAD instead of comparing github.sha with itself
Increases fetch depth and adds fallback logic to handle cases where merge-base isn't found in shallow clones

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

bugraoz93 · 2026-01-07T00:04:18Z

+ git fetch origin "$BASE_BRANCH" --depth=50
+
+ # Try to find merge-base, deepen if not found
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
+
+ if [ -z "$MERGE_BASE" ]; then
+ echo "Merge-base not found with depth=50, fetching more history..."
+ git fetch --deepen=450
+ git fetch origin "$BASE_BRANCH" --deepen=450
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
+ fi
+
+ echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
+ env:
+ BASE_BRANCH: ${{ github.base_ref || 'main' }}


Suggested change
git fetch origin "$BASE_BRANCH" --depth=50
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=50, fetching more history..."
git fetch --deepen=450
git fetch origin "$BASE_BRANCH" --deepen=450
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
git fetch origin "$BASE_BRANCH" --depth=$FETCH_DEPTH
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=$FETCH_DEPTH, fetching more history..."
git fetch --deepen=$DEEPEN
git fetch origin "$BASE_BRANCH" --deepen=$DEEPEN
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
FETCH_DEPTH: 50
DEEPEN: 450

Small nit, maybe we can make the values parameterised for both --depth and --deepen from env :)

amoghrajesh

Wow, thats a strange one

potiuk

Strange one .. I wonder when it happened and how it did not work (I am sure it worked before). But it can be done way simpler.

All you need it is to add missing ^ in --from-ref

At the moment you do this check - github.sha already contains merge commit - this is how pull-request workflow works.

https://docs.github.com/en/actions/reference/workflows-and-actions/events-that-trigger-workflows#pull_request

Note that GITHUB_SHA for this event is the last merge commit of the pull request merge branch. If you want to get the commit ID for the last commit to the head branch of the pull request, use github.event.pull_request.head.sha instead.

potiuk · 2026-01-07T08:39:15Z

Superseded by #60202

potiuk · 2026-01-07T08:42:11Z

This is precisely why we needed 2 depth - because in pull_request the github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased).

potiuk · 2026-01-07T09:20:57Z

I merged my fix and rebased #60168 to verify that it works (it should fail)

potiuk · 2026-01-07T09:21:11Z

Nice catch @jedcunningham BTW :)

potiuk · 2026-01-07T09:24:20Z

Yep. Nicely failed now https://github.com/apache/airflow/actions/runs/20776566467/job/59663769292?pr=60168

jedcunningham · 2026-01-07T18:52:40Z

Nice, I'd considered just adding "^" but somehow I confused myself thinking that would only go 1 commit up - that if we had many commits on the branch it'd only look at the last one. I guess it must squash? Anyways, glad its working.

edit:

github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased)

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

potiuk · 2026-01-07T22:07:27Z

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

Correct. This is actually what we very heavily bank on (this why I knew it by heart).
Basically whole selective checks code is based on the fact that we have "merge" commit passed as github.sha and it's first parent is base branch.

https://github.com/apache/airflow/blob/main/dev/breeze/src/airflow_breeze/commands/ci_commands.py#L163:

defget_changed_files(commit_ref: str|None) ->tuple[str, ...]:
ifcommit_refisNone:
return ()
cmd= [
"git",
"diff-tree",
"--no-commit-id",
"--name-only",
"-r",
commit_ref+"^",
commit_ref,
]
result=run_command(cmd, check=False, capture_output=True, text=True)
ifresult.returncode!=0:
get_console().print(
f"[warning] Error when running diff-tree command [/]\n{result.stdout}\n{result.stderr}"
)
return ()
changed_files=tuple(result.stdout.splitlines()) ifresult.stdoutelse ()
get_console().print("\n[info]Changed files:[/]\n")
get_console().print(changed_files)
get_console().print()
returnchanged_files

potiuk · 2026-01-07T22:18:43Z

BTW. There is also another interesting thing I learned with prek

There is a very nice shortcut to not have to find merge-base.

git diff main..your_pr_branch
git diff main...your_pr_branch

Spot the difference .... there are THREE dots in second case. And the ... is actually showing the changes in your branch from the merge-base

And ... (pun intended) there are few other gotchas: https://darekkay.com/blog/git-commit-ranges/#git-diff

For example git diff main...your-branch is not reverse of git diff your-branch...diff as you would suspect. The first one shows your changes since merge-base, the second shows SURPRISE -> all the changes in main (!) since merge-base 😱 😱 😱 😱 😱 😱 😱 😱 😱

Git is wild.

potiuk · 2026-01-07T22:20:40Z

So basically if you want to have prek to execute on "only your changes" from a branch you have PR to main with - do this (without running any of git merge-base thingies):

prek --from-ref maim

(we have it in our docs)

jedcunningham · 2026-01-06T23:42:42Z

The prek command was comparing github.sha with itself, which checked no files. Now uses merge-base to check all files changed in the PR against the base branch.

This ensures static checks are actually ran in PRs like #60168.

The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

Copilot

Pull request overview

Fixes a bug in the basic-tests workflow where static checks were comparing the same commit (github.sha) with itself, resulting in no files being checked. The fix introduces a merge-base calculation to properly identify all changed files in a PR by comparing against the base branch.

Key changes:

Implements merge-base calculation to find the common ancestor between PR and base branch
Updates static check command to compare merge-base commit against HEAD instead of comparing github.sha with itself
Increases fetch depth and adds fallback logic to handle cases where merge-base isn't found in shallow clones

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

bugraoz93 · 2026-01-07T00:04:18Z

+ git fetch origin "$BASE_BRANCH" --depth=50
+
+ # Try to find merge-base, deepen if not found
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
+
+ if [ -z "$MERGE_BASE" ]; then
+ echo "Merge-base not found with depth=50, fetching more history..."
+ git fetch --deepen=450
+ git fetch origin "$BASE_BRANCH" --deepen=450
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
+ fi
+
+ echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
+ env:
+ BASE_BRANCH: ${{ github.base_ref || 'main' }}


Suggested change
git fetch origin "$BASE_BRANCH" --depth=50
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=50, fetching more history..."
git fetch --deepen=450
git fetch origin "$BASE_BRANCH" --deepen=450
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
git fetch origin "$BASE_BRANCH" --depth=$FETCH_DEPTH
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=$FETCH_DEPTH, fetching more history..."
git fetch --deepen=$DEEPEN
git fetch origin "$BASE_BRANCH" --deepen=$DEEPEN
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
FETCH_DEPTH: 50
DEEPEN: 450

Small nit, maybe we can make the values parameterised for both --depth and --deepen from env :)

amoghrajesh

Wow, thats a strange one

potiuk

Strange one .. I wonder when it happened and how it did not work (I am sure it worked before). But it can be done way simpler.

All you need it is to add missing ^ in --from-ref

At the moment you do this check - github.sha already contains merge commit - this is how pull-request workflow works.

https://docs.github.com/en/actions/reference/workflows-and-actions/events-that-trigger-workflows#pull_request

Note that GITHUB_SHA for this event is the last merge commit of the pull request merge branch. If you want to get the commit ID for the last commit to the head branch of the pull request, use github.event.pull_request.head.sha instead.

potiuk · 2026-01-07T08:39:15Z

Superseded by #60202

potiuk · 2026-01-07T08:42:11Z

This is precisely why we needed 2 depth - because in pull_request the github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased).

potiuk · 2026-01-07T09:20:57Z

I merged my fix and rebased #60168 to verify that it works (it should fail)

potiuk · 2026-01-07T09:21:11Z

Nice catch @jedcunningham BTW :)

potiuk · 2026-01-07T09:24:20Z

Yep. Nicely failed now https://github.com/apache/airflow/actions/runs/20776566467/job/59663769292?pr=60168

jedcunningham · 2026-01-07T18:52:40Z

Nice, I'd considered just adding "^" but somehow I confused myself thinking that would only go 1 commit up - that if we had many commits on the branch it'd only look at the last one. I guess it must squash? Anyways, glad its working.

edit:

github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased)

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

potiuk · 2026-01-07T22:07:27Z

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

Correct. This is actually what we very heavily bank on (this why I knew it by heart).
Basically whole selective checks code is based on the fact that we have "merge" commit passed as github.sha and it's first parent is base branch.

https://github.com/apache/airflow/blob/main/dev/breeze/src/airflow_breeze/commands/ci_commands.py#L163:

defget_changed_files(commit_ref: str|None) ->tuple[str, ...]:
ifcommit_refisNone:
return ()
cmd= [
"git",
"diff-tree",
"--no-commit-id",
"--name-only",
"-r",
commit_ref+"^",
commit_ref,
]
result=run_command(cmd, check=False, capture_output=True, text=True)
ifresult.returncode!=0:
get_console().print(
f"[warning] Error when running diff-tree command [/]\n{result.stdout}\n{result.stderr}"
)
return ()
changed_files=tuple(result.stdout.splitlines()) ifresult.stdoutelse ()
get_console().print("\n[info]Changed files:[/]\n")
get_console().print(changed_files)
get_console().print()
returnchanged_files

potiuk · 2026-01-07T22:18:43Z

BTW. There is also another interesting thing I learned with prek

There is a very nice shortcut to not have to find merge-base.

git diff main..your_pr_branch
git diff main...your_pr_branch

Spot the difference .... there are THREE dots in second case. And the ... is actually showing the changes in your branch from the merge-base

And ... (pun intended) there are few other gotchas: https://darekkay.com/blog/git-commit-ranges/#git-diff

For example git diff main...your-branch is not reverse of git diff your-branch...diff as you would suspect. The first one shows your changes since merge-base, the second shows SURPRISE -> all the changes in main (!) since merge-base 😱 😱 😱 😱 😱 😱 😱 😱 😱

Git is wild.

potiuk · 2026-01-07T22:20:40Z

So basically if you want to have prek to execute on "only your changes" from a branch you have PR to main with - do this (without running any of git merge-base thingies):

prek --from-ref maim

(we have it in our docs)

jedcunningham · 2026-01-06T23:42:42Z

The prek command was comparing github.sha with itself, which checked no files. Now uses merge-base to check all files changed in the PR against the base branch.

This ensures static checks are actually ran in PRs like #60168.

The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

Copilot

Pull request overview

Fixes a bug in the basic-tests workflow where static checks were comparing the same commit (github.sha) with itself, resulting in no files being checked. The fix introduces a merge-base calculation to properly identify all changed files in a PR by comparing against the base branch.

Key changes:

Implements merge-base calculation to find the common ancestor between PR and base branch
Updates static check command to compare merge-base commit against HEAD instead of comparing github.sha with itself
Increases fetch depth and adds fallback logic to handle cases where merge-base isn't found in shallow clones

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

bugraoz93 · 2026-01-07T00:04:18Z

+ git fetch origin "$BASE_BRANCH" --depth=50
+
+ # Try to find merge-base, deepen if not found
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
+
+ if [ -z "$MERGE_BASE" ]; then
+ echo "Merge-base not found with depth=50, fetching more history..."
+ git fetch --deepen=450
+ git fetch origin "$BASE_BRANCH" --deepen=450
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
+ fi
+
+ echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
+ env:
+ BASE_BRANCH: ${{ github.base_ref || 'main' }}


Suggested change
git fetch origin "$BASE_BRANCH" --depth=50
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=50, fetching more history..."
git fetch --deepen=450
git fetch origin "$BASE_BRANCH" --deepen=450
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
git fetch origin "$BASE_BRANCH" --depth=$FETCH_DEPTH
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=$FETCH_DEPTH, fetching more history..."
git fetch --deepen=$DEEPEN
git fetch origin "$BASE_BRANCH" --deepen=$DEEPEN
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
FETCH_DEPTH: 50
DEEPEN: 450

Small nit, maybe we can make the values parameterised for both --depth and --deepen from env :)

amoghrajesh

Wow, thats a strange one

potiuk

Strange one .. I wonder when it happened and how it did not work (I am sure it worked before). But it can be done way simpler.

All you need it is to add missing ^ in --from-ref

At the moment you do this check - github.sha already contains merge commit - this is how pull-request workflow works.

https://docs.github.com/en/actions/reference/workflows-and-actions/events-that-trigger-workflows#pull_request

Note that GITHUB_SHA for this event is the last merge commit of the pull request merge branch. If you want to get the commit ID for the last commit to the head branch of the pull request, use github.event.pull_request.head.sha instead.

potiuk · 2026-01-07T08:39:15Z

Superseded by #60202

potiuk · 2026-01-07T08:42:11Z

This is precisely why we needed 2 depth - because in pull_request the github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased).

potiuk · 2026-01-07T09:20:57Z

I merged my fix and rebased #60168 to verify that it works (it should fail)

potiuk · 2026-01-07T09:21:11Z

Nice catch @jedcunningham BTW :)

potiuk · 2026-01-07T09:24:20Z

Yep. Nicely failed now https://github.com/apache/airflow/actions/runs/20776566467/job/59663769292?pr=60168

jedcunningham · 2026-01-07T18:52:40Z

Nice, I'd considered just adding "^" but somehow I confused myself thinking that would only go 1 commit up - that if we had many commits on the branch it'd only look at the last one. I guess it must squash? Anyways, glad its working.

edit:

github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased)

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

potiuk · 2026-01-07T22:07:27Z

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

Correct. This is actually what we very heavily bank on (this why I knew it by heart).
Basically whole selective checks code is based on the fact that we have "merge" commit passed as github.sha and it's first parent is base branch.

https://github.com/apache/airflow/blob/main/dev/breeze/src/airflow_breeze/commands/ci_commands.py#L163:

defget_changed_files(commit_ref: str|None) ->tuple[str, ...]:
ifcommit_refisNone:
return ()
cmd= [
"git",
"diff-tree",
"--no-commit-id",
"--name-only",
"-r",
commit_ref+"^",
commit_ref,
]
result=run_command(cmd, check=False, capture_output=True, text=True)
ifresult.returncode!=0:
get_console().print(
f"[warning] Error when running diff-tree command [/]\n{result.stdout}\n{result.stderr}"
)
return ()
changed_files=tuple(result.stdout.splitlines()) ifresult.stdoutelse ()
get_console().print("\n[info]Changed files:[/]\n")
get_console().print(changed_files)
get_console().print()
returnchanged_files

potiuk · 2026-01-07T22:18:43Z

BTW. There is also another interesting thing I learned with prek

There is a very nice shortcut to not have to find merge-base.

git diff main..your_pr_branch
git diff main...your_pr_branch

Spot the difference .... there are THREE dots in second case. And the ... is actually showing the changes in your branch from the merge-base

And ... (pun intended) there are few other gotchas: https://darekkay.com/blog/git-commit-ranges/#git-diff

For example git diff main...your-branch is not reverse of git diff your-branch...diff as you would suspect. The first one shows your changes since merge-base, the second shows SURPRISE -> all the changes in main (!) since merge-base 😱 😱 😱 😱 😱 😱 😱 😱 😱

Git is wild.

potiuk · 2026-01-07T22:20:40Z

So basically if you want to have prek to execute on "only your changes" from a branch you have PR to main with - do this (without running any of git merge-base thingies):

prek --from-ref maim

(we have it in our docs)

jedcunningham · 2026-01-06T23:42:42Z

The prek command was comparing github.sha with itself, which checked no files. Now uses merge-base to check all files changed in the PR against the base branch.

This ensures static checks are actually ran in PRs like #60168.

The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

Copilot

Pull request overview

Fixes a bug in the basic-tests workflow where static checks were comparing the same commit (github.sha) with itself, resulting in no files being checked. The fix introduces a merge-base calculation to properly identify all changed files in a PR by comparing against the base branch.

Key changes:

Implements merge-base calculation to find the common ancestor between PR and base branch
Updates static check command to compare merge-base commit against HEAD instead of comparing github.sha with itself
Increases fetch depth and adds fallback logic to handle cases where merge-base isn't found in shallow clones

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

bugraoz93 · 2026-01-07T00:04:18Z

+ git fetch origin "$BASE_BRANCH" --depth=50
+
+ # Try to find merge-base, deepen if not found
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
+
+ if [ -z "$MERGE_BASE" ]; then
+ echo "Merge-base not found with depth=50, fetching more history..."
+ git fetch --deepen=450
+ git fetch origin "$BASE_BRANCH" --deepen=450
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
+ fi
+
+ echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
+ env:
+ BASE_BRANCH: ${{ github.base_ref || 'main' }}


Suggested change
git fetch origin "$BASE_BRANCH" --depth=50
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=50, fetching more history..."
git fetch --deepen=450
git fetch origin "$BASE_BRANCH" --deepen=450
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
git fetch origin "$BASE_BRANCH" --depth=$FETCH_DEPTH
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=$FETCH_DEPTH, fetching more history..."
git fetch --deepen=$DEEPEN
git fetch origin "$BASE_BRANCH" --deepen=$DEEPEN
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
FETCH_DEPTH: 50
DEEPEN: 450

Small nit, maybe we can make the values parameterised for both --depth and --deepen from env :)

amoghrajesh

Wow, thats a strange one

potiuk

Strange one .. I wonder when it happened and how it did not work (I am sure it worked before). But it can be done way simpler.

All you need it is to add missing ^ in --from-ref

At the moment you do this check - github.sha already contains merge commit - this is how pull-request workflow works.

https://docs.github.com/en/actions/reference/workflows-and-actions/events-that-trigger-workflows#pull_request

Note that GITHUB_SHA for this event is the last merge commit of the pull request merge branch. If you want to get the commit ID for the last commit to the head branch of the pull request, use github.event.pull_request.head.sha instead.

potiuk · 2026-01-07T08:39:15Z

Superseded by #60202

potiuk · 2026-01-07T08:42:11Z

This is precisely why we needed 2 depth - because in pull_request the github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased).

potiuk · 2026-01-07T09:20:57Z

I merged my fix and rebased #60168 to verify that it works (it should fail)

potiuk · 2026-01-07T09:21:11Z

Nice catch @jedcunningham BTW :)

potiuk · 2026-01-07T09:24:20Z

Yep. Nicely failed now https://github.com/apache/airflow/actions/runs/20776566467/job/59663769292?pr=60168

jedcunningham · 2026-01-07T18:52:40Z

Nice, I'd considered just adding "^" but somehow I confused myself thinking that would only go 1 commit up - that if we had many commits on the branch it'd only look at the last one. I guess it must squash? Anyways, glad its working.

edit:

github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased)

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

potiuk · 2026-01-07T22:07:27Z

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

Correct. This is actually what we very heavily bank on (this why I knew it by heart).
Basically whole selective checks code is based on the fact that we have "merge" commit passed as github.sha and it's first parent is base branch.

https://github.com/apache/airflow/blob/main/dev/breeze/src/airflow_breeze/commands/ci_commands.py#L163:

defget_changed_files(commit_ref: str|None) ->tuple[str, ...]:
ifcommit_refisNone:
return ()
cmd= [
"git",
"diff-tree",
"--no-commit-id",
"--name-only",
"-r",
commit_ref+"^",
commit_ref,
]
result=run_command(cmd, check=False, capture_output=True, text=True)
ifresult.returncode!=0:
get_console().print(
f"[warning] Error when running diff-tree command [/]\n{result.stdout}\n{result.stderr}"
)
return ()
changed_files=tuple(result.stdout.splitlines()) ifresult.stdoutelse ()
get_console().print("\n[info]Changed files:[/]\n")
get_console().print(changed_files)
get_console().print()
returnchanged_files

potiuk · 2026-01-07T22:18:43Z

BTW. There is also another interesting thing I learned with prek

There is a very nice shortcut to not have to find merge-base.

git diff main..your_pr_branch
git diff main...your_pr_branch

Spot the difference .... there are THREE dots in second case. And the ... is actually showing the changes in your branch from the merge-base

And ... (pun intended) there are few other gotchas: https://darekkay.com/blog/git-commit-ranges/#git-diff

For example git diff main...your-branch is not reverse of git diff your-branch...diff as you would suspect. The first one shows your changes since merge-base, the second shows SURPRISE -> all the changes in main (!) since merge-base 😱 😱 😱 😱 😱 😱 😱 😱 😱

Git is wild.

potiuk · 2026-01-07T22:20:40Z

So basically if you want to have prek to execute on "only your changes" from a branch you have PR to main with - do this (without running any of git merge-base thingies):

prek --from-ref maim

(we have it in our docs)

jedcunningham · 2026-01-06T23:42:42Z

The prek command was comparing github.sha with itself, which checked no files. Now uses merge-base to check all files changed in the PR against the base branch.

This ensures static checks are actually ran in PRs like #60168.

The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

Copilot

Pull request overview

Fixes a bug in the basic-tests workflow where static checks were comparing the same commit (github.sha) with itself, resulting in no files being checked. The fix introduces a merge-base calculation to properly identify all changed files in a PR by comparing against the base branch.

Key changes:

Implements merge-base calculation to find the common ancestor between PR and base branch
Updates static check command to compare merge-base commit against HEAD instead of comparing github.sha with itself
Increases fetch depth and adds fallback logic to handle cases where merge-base isn't found in shallow clones

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

bugraoz93 · 2026-01-07T00:04:18Z

+ git fetch origin "$BASE_BRANCH" --depth=50
+
+ # Try to find merge-base, deepen if not found
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
+
+ if [ -z "$MERGE_BASE" ]; then
+ echo "Merge-base not found with depth=50, fetching more history..."
+ git fetch --deepen=450
+ git fetch origin "$BASE_BRANCH" --deepen=450
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
+ fi
+
+ echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
+ env:
+ BASE_BRANCH: ${{ github.base_ref || 'main' }}


Suggested change
git fetch origin "$BASE_BRANCH" --depth=50
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=50, fetching more history..."
git fetch --deepen=450
git fetch origin "$BASE_BRANCH" --deepen=450
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
git fetch origin "$BASE_BRANCH" --depth=$FETCH_DEPTH
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=$FETCH_DEPTH, fetching more history..."
git fetch --deepen=$DEEPEN
git fetch origin "$BASE_BRANCH" --deepen=$DEEPEN
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
FETCH_DEPTH: 50
DEEPEN: 450

Small nit, maybe we can make the values parameterised for both --depth and --deepen from env :)

amoghrajesh

Wow, thats a strange one

potiuk

Strange one .. I wonder when it happened and how it did not work (I am sure it worked before). But it can be done way simpler.

All you need it is to add missing ^ in --from-ref

At the moment you do this check - github.sha already contains merge commit - this is how pull-request workflow works.

https://docs.github.com/en/actions/reference/workflows-and-actions/events-that-trigger-workflows#pull_request

Note that GITHUB_SHA for this event is the last merge commit of the pull request merge branch. If you want to get the commit ID for the last commit to the head branch of the pull request, use github.event.pull_request.head.sha instead.

potiuk · 2026-01-07T08:39:15Z

Superseded by #60202

potiuk · 2026-01-07T08:42:11Z

This is precisely why we needed 2 depth - because in pull_request the github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased).

potiuk · 2026-01-07T09:20:57Z

I merged my fix and rebased #60168 to verify that it works (it should fail)

potiuk · 2026-01-07T09:21:11Z

Nice catch @jedcunningham BTW :)

potiuk · 2026-01-07T09:24:20Z

Yep. Nicely failed now https://github.com/apache/airflow/actions/runs/20776566467/job/59663769292?pr=60168

jedcunningham · 2026-01-07T18:52:40Z

Nice, I'd considered just adding "^" but somehow I confused myself thinking that would only go 1 commit up - that if we had many commits on the branch it'd only look at the last one. I guess it must squash? Anyways, glad its working.

edit:

github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased)

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

potiuk · 2026-01-07T22:07:27Z

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

Correct. This is actually what we very heavily bank on (this why I knew it by heart).
Basically whole selective checks code is based on the fact that we have "merge" commit passed as github.sha and it's first parent is base branch.

https://github.com/apache/airflow/blob/main/dev/breeze/src/airflow_breeze/commands/ci_commands.py#L163:

defget_changed_files(commit_ref: str|None) ->tuple[str, ...]:
ifcommit_refisNone:
return ()
cmd= [
"git",
"diff-tree",
"--no-commit-id",
"--name-only",
"-r",
commit_ref+"^",
commit_ref,
]
result=run_command(cmd, check=False, capture_output=True, text=True)
ifresult.returncode!=0:
get_console().print(
f"[warning] Error when running diff-tree command [/]\n{result.stdout}\n{result.stderr}"
)
return ()
changed_files=tuple(result.stdout.splitlines()) ifresult.stdoutelse ()
get_console().print("\n[info]Changed files:[/]\n")
get_console().print(changed_files)
get_console().print()
returnchanged_files

potiuk · 2026-01-07T22:18:43Z

BTW. There is also another interesting thing I learned with prek

There is a very nice shortcut to not have to find merge-base.

git diff main..your_pr_branch
git diff main...your_pr_branch

Spot the difference .... there are THREE dots in second case. And the ... is actually showing the changes in your branch from the merge-base

And ... (pun intended) there are few other gotchas: https://darekkay.com/blog/git-commit-ranges/#git-diff

For example git diff main...your-branch is not reverse of git diff your-branch...diff as you would suspect. The first one shows your changes since merge-base, the second shows SURPRISE -> all the changes in main (!) since merge-base 😱 😱 😱 😱 😱 😱 😱 😱 😱

Git is wild.

potiuk · 2026-01-07T22:20:40Z

So basically if you want to have prek to execute on "only your changes" from a branch you have PR to main with - do this (without running any of git merge-base thingies):

prek --from-ref maim

(we have it in our docs)

jedcunningham · 2026-01-06T23:42:42Z

The prek command was comparing github.sha with itself, which checked no files. Now uses merge-base to check all files changed in the PR against the base branch.

This ensures static checks are actually ran in PRs like #60168.

The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

Copilot

Pull request overview

Fixes a bug in the basic-tests workflow where static checks were comparing the same commit (github.sha) with itself, resulting in no files being checked. The fix introduces a merge-base calculation to properly identify all changed files in a PR by comparing against the base branch.

Key changes:

Implements merge-base calculation to find the common ancestor between PR and base branch
Updates static check command to compare merge-base commit against HEAD instead of comparing github.sha with itself
Increases fetch depth and adds fallback logic to handle cases where merge-base isn't found in shallow clones

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

bugraoz93 · 2026-01-07T00:04:18Z

+ git fetch origin "$BASE_BRANCH" --depth=50
+
+ # Try to find merge-base, deepen if not found
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
+
+ if [ -z "$MERGE_BASE" ]; then
+ echo "Merge-base not found with depth=50, fetching more history..."
+ git fetch --deepen=450
+ git fetch origin "$BASE_BRANCH" --deepen=450
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
+ fi
+
+ echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
+ env:
+ BASE_BRANCH: ${{ github.base_ref || 'main' }}


Suggested change
git fetch origin "$BASE_BRANCH" --depth=50
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=50, fetching more history..."
git fetch --deepen=450
git fetch origin "$BASE_BRANCH" --deepen=450
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
git fetch origin "$BASE_BRANCH" --depth=$FETCH_DEPTH
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=$FETCH_DEPTH, fetching more history..."
git fetch --deepen=$DEEPEN
git fetch origin "$BASE_BRANCH" --deepen=$DEEPEN
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
FETCH_DEPTH: 50
DEEPEN: 450

Small nit, maybe we can make the values parameterised for both --depth and --deepen from env :)

amoghrajesh

Wow, thats a strange one

potiuk

Strange one .. I wonder when it happened and how it did not work (I am sure it worked before). But it can be done way simpler.

All you need it is to add missing ^ in --from-ref

At the moment you do this check - github.sha already contains merge commit - this is how pull-request workflow works.

https://docs.github.com/en/actions/reference/workflows-and-actions/events-that-trigger-workflows#pull_request

Note that GITHUB_SHA for this event is the last merge commit of the pull request merge branch. If you want to get the commit ID for the last commit to the head branch of the pull request, use github.event.pull_request.head.sha instead.

potiuk · 2026-01-07T08:39:15Z

Superseded by #60202

potiuk · 2026-01-07T08:42:11Z

This is precisely why we needed 2 depth - because in pull_request the github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased).

potiuk · 2026-01-07T09:20:57Z

I merged my fix and rebased #60168 to verify that it works (it should fail)

potiuk · 2026-01-07T09:21:11Z

Nice catch @jedcunningham BTW :)

potiuk · 2026-01-07T09:24:20Z

Yep. Nicely failed now https://github.com/apache/airflow/actions/runs/20776566467/job/59663769292?pr=60168

jedcunningham · 2026-01-07T18:52:40Z

Nice, I'd considered just adding "^" but somehow I confused myself thinking that would only go 1 commit up - that if we had many commits on the branch it'd only look at the last one. I guess it must squash? Anyways, glad its working.

edit:

github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased)

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

potiuk · 2026-01-07T22:07:27Z

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

Correct. This is actually what we very heavily bank on (this why I knew it by heart).
Basically whole selective checks code is based on the fact that we have "merge" commit passed as github.sha and it's first parent is base branch.

https://github.com/apache/airflow/blob/main/dev/breeze/src/airflow_breeze/commands/ci_commands.py#L163:

defget_changed_files(commit_ref: str|None) ->tuple[str, ...]:
ifcommit_refisNone:
return ()
cmd= [
"git",
"diff-tree",
"--no-commit-id",
"--name-only",
"-r",
commit_ref+"^",
commit_ref,
]
result=run_command(cmd, check=False, capture_output=True, text=True)
ifresult.returncode!=0:
get_console().print(
f"[warning] Error when running diff-tree command [/]\n{result.stdout}\n{result.stderr}"
)
return ()
changed_files=tuple(result.stdout.splitlines()) ifresult.stdoutelse ()
get_console().print("\n[info]Changed files:[/]\n")
get_console().print(changed_files)
get_console().print()
returnchanged_files

potiuk · 2026-01-07T22:18:43Z

BTW. There is also another interesting thing I learned with prek

There is a very nice shortcut to not have to find merge-base.

git diff main..your_pr_branch
git diff main...your_pr_branch

Spot the difference .... there are THREE dots in second case. And the ... is actually showing the changes in your branch from the merge-base

And ... (pun intended) there are few other gotchas: https://darekkay.com/blog/git-commit-ranges/#git-diff

For example git diff main...your-branch is not reverse of git diff your-branch...diff as you would suspect. The first one shows your changes since merge-base, the second shows SURPRISE -> all the changes in main (!) since merge-base 😱 😱 😱 😱 😱 😱 😱 😱 😱

Git is wild.

potiuk · 2026-01-07T22:20:40Z

So basically if you want to have prek to execute on "only your changes" from a branch you have PR to main with - do this (without running any of git merge-base thingies):

prek --from-ref maim

(we have it in our docs)

jedcunningham · 2026-01-06T23:42:42Z

The prek command was comparing github.sha with itself, which checked no files. Now uses merge-base to check all files changed in the PR against the base branch.

This ensures static checks are actually ran in PRs like #60168.

The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

Copilot

Pull request overview

Fixes a bug in the basic-tests workflow where static checks were comparing the same commit (github.sha) with itself, resulting in no files being checked. The fix introduces a merge-base calculation to properly identify all changed files in a PR by comparing against the base branch.

Key changes:

Implements merge-base calculation to find the common ancestor between PR and base branch
Updates static check command to compare merge-base commit against HEAD instead of comparing github.sha with itself
Increases fetch depth and adds fallback logic to handle cases where merge-base isn't found in shallow clones

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

bugraoz93 · 2026-01-07T00:04:18Z

+ git fetch origin "$BASE_BRANCH" --depth=50
+
+ # Try to find merge-base, deepen if not found
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
+
+ if [ -z "$MERGE_BASE" ]; then
+ echo "Merge-base not found with depth=50, fetching more history..."
+ git fetch --deepen=450
+ git fetch origin "$BASE_BRANCH" --deepen=450
+ MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
+ fi
+
+ echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
+ env:
+ BASE_BRANCH: ${{ github.base_ref || 'main' }}


Suggested change
git fetch origin "$BASE_BRANCH" --depth=50
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=50, fetching more history..."
git fetch --deepen=450
git fetch origin "$BASE_BRANCH" --deepen=450
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
git fetch origin "$BASE_BRANCH" --depth=$FETCH_DEPTH
# Try to find merge-base, deepen if not found
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD 2>/dev/null || echo "")
if [ -z "$MERGE_BASE" ]; then
echo "Merge-base not found with depth=$FETCH_DEPTH, fetching more history..."
git fetch --deepen=$DEEPEN
git fetch origin "$BASE_BRANCH" --deepen=$DEEPEN
MERGE_BASE=$(git merge-base "origin/$BASE_BRANCH" HEAD)
fi
echo "sha=${MERGE_BASE}" >> $GITHUB_OUTPUT
env:
BASE_BRANCH: ${{ github.base_ref || 'main' }}
FETCH_DEPTH: 50
DEEPEN: 450

Small nit, maybe we can make the values parameterised for both --depth and --deepen from env :)

amoghrajesh

Wow, thats a strange one

potiuk

Strange one .. I wonder when it happened and how it did not work (I am sure it worked before). But it can be done way simpler.

All you need it is to add missing ^ in --from-ref

At the moment you do this check - github.sha already contains merge commit - this is how pull-request workflow works.

https://docs.github.com/en/actions/reference/workflows-and-actions/events-that-trigger-workflows#pull_request

Note that GITHUB_SHA for this event is the last merge commit of the pull request merge branch. If you want to get the commit ID for the last commit to the head branch of the pull request, use github.event.pull_request.head.sha instead.

potiuk · 2026-01-07T08:39:15Z

Superseded by #60202

potiuk · 2026-01-07T08:42:11Z

This is precisely why we needed 2 depth - because in pull_request the github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased).

potiuk · 2026-01-07T09:20:57Z

I merged my fix and rebased #60168 to verify that it works (it should fail)

potiuk · 2026-01-07T09:21:11Z

Nice catch @jedcunningham BTW :)

potiuk · 2026-01-07T09:24:20Z

Yep. Nicely failed now https://github.com/apache/airflow/actions/runs/20776566467/job/59663769292?pr=60168

jedcunningham · 2026-01-07T18:52:40Z

Nice, I'd considered just adding "^" but somehow I confused myself thinking that would only go 1 commit up - that if we had many commits on the branch it'd only look at the last one. I guess it must squash? Anyways, glad its working.

edit:

github.sha is merge and it's parent is the commit that was in the target branch when the merge was created (i.e. last time the PR was rebased)

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

potiuk · 2026-01-07T22:07:27Z

Ah, a merge commit. Confusing since it has 2 parents. Apparently the "first" parent is the commit from the base branch, which is why this works. TIL.

Correct. This is actually what we very heavily bank on (this why I knew it by heart).
Basically whole selective checks code is based on the fact that we have "merge" commit passed as github.sha and it's first parent is base branch.

https://github.com/apache/airflow/blob/main/dev/breeze/src/airflow_breeze/commands/ci_commands.py#L163:

defget_changed_files(commit_ref: str|None) ->tuple[str, ...]:
ifcommit_refisNone:
return ()
cmd= [
"git",
"diff-tree",
"--no-commit-id",
"--name-only",
"-r",
commit_ref+"^",
commit_ref,
]
result=run_command(cmd, check=False, capture_output=True, text=True)
ifresult.returncode!=0:
get_console().print(
f"[warning] Error when running diff-tree command [/]\n{result.stdout}\n{result.stderr}"
)
return ()
changed_files=tuple(result.stdout.splitlines()) ifresult.stdoutelse ()
get_console().print("\n[info]Changed files:[/]\n")
get_console().print(changed_files)
get_console().print()
returnchanged_files

potiuk · 2026-01-07T22:18:43Z

BTW. There is also another interesting thing I learned with prek

There is a very nice shortcut to not have to find merge-base.

git diff main..your_pr_branch
git diff main...your_pr_branch

Spot the difference .... there are THREE dots in second case. And the ... is actually showing the changes in your branch from the merge-base

And ... (pun intended) there are few other gotchas: https://darekkay.com/blog/git-commit-ranges/#git-diff

For example git diff main...your-branch is not reverse of git diff your-branch...diff as you would suspect. The first one shows your changes since merge-base, the second shows SURPRISE -> all the changes in main (!) since merge-base 😱 😱 😱 😱 😱 😱 😱 😱 😱

Git is wild.

potiuk · 2026-01-07T22:20:40Z

So basically if you want to have prek to execute on "only your changes" from a branch you have PR to main with - do this (without running any of git merge-base thingies):

prek --from-ref maim

(we have it in our docs)

Fix basic static checks comparing same commit with itself
b9f20ea
The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

jedcunningham requested review from amoghrajesh, ashb, bugraoz93, gopidesupavan, jason810496, jscheffl, kaxil and potiuk as code owners January 6, 2026 23:42

boring-cyborgBot added area:dev-tools backport-to-v3-1-test labels Jan 6, 2026

jedcunningham requested a review from Copilot January 6, 2026 23:43

CopilotAI reviewed Jan 6, 2026
View reviewed changes

Comment thread.github/workflows/basic-tests.yml
Comment thread.github/workflows/basic-tests.yml

bugraoz93 approved these changes Jan 7, 2026
View reviewed changes

amoghrajesh approved these changes Jan 7, 2026
View reviewed changes

potiuk reviewed Jan 7, 2026
View reviewed changes

potiuk closed this Jan 7, 2026

jedcunningham deleted the fix_inthewild_sorting branch January 7, 2026 18:52

Fix basic static checks comparing same commit with itself
b9f20ea
The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

jedcunningham requested review from amoghrajesh, ashb, bugraoz93, gopidesupavan, jason810496, jscheffl, kaxil and potiuk as code owners January 6, 2026 23:42

boring-cyborgBot added area:dev-tools backport-to-v3-1-test labels Jan 6, 2026

jedcunningham requested a review from Copilot January 6, 2026 23:43

CopilotAI reviewed Jan 6, 2026
View reviewed changes

Comment thread.github/workflows/basic-tests.yml
Comment thread.github/workflows/basic-tests.yml

bugraoz93 approved these changes Jan 7, 2026
View reviewed changes

amoghrajesh approved these changes Jan 7, 2026
View reviewed changes

potiuk reviewed Jan 7, 2026
View reviewed changes

potiuk closed this Jan 7, 2026

jedcunningham deleted the fix_inthewild_sorting branch January 7, 2026 18:52

Fix basic static checks comparing same commit with itself
b9f20ea
The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

jedcunningham requested review from amoghrajesh, ashb, bugraoz93, gopidesupavan, jason810496, jscheffl, kaxil and potiuk as code owners January 6, 2026 23:42

boring-cyborgBot added area:dev-tools backport-to-v3-1-test labels Jan 6, 2026

jedcunningham requested a review from Copilot January 6, 2026 23:43

CopilotAI reviewed Jan 6, 2026
View reviewed changes

Comment thread.github/workflows/basic-tests.yml
Comment thread.github/workflows/basic-tests.yml

bugraoz93 approved these changes Jan 7, 2026
View reviewed changes

amoghrajesh approved these changes Jan 7, 2026
View reviewed changes

potiuk reviewed Jan 7, 2026
View reviewed changes

potiuk closed this Jan 7, 2026

jedcunningham deleted the fix_inthewild_sorting branch January 7, 2026 18:52

Fix basic static checks comparing same commit with itself
b9f20ea
The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

jedcunningham requested review from amoghrajesh, ashb, bugraoz93, gopidesupavan, jason810496, jscheffl, kaxil and potiuk as code owners January 6, 2026 23:42

boring-cyborgBot added area:dev-tools backport-to-v3-1-test labels Jan 6, 2026

jedcunningham requested a review from Copilot January 6, 2026 23:43

CopilotAI reviewed Jan 6, 2026
View reviewed changes

Comment thread.github/workflows/basic-tests.yml
Comment thread.github/workflows/basic-tests.yml

bugraoz93 approved these changes Jan 7, 2026
View reviewed changes

amoghrajesh approved these changes Jan 7, 2026
View reviewed changes

potiuk reviewed Jan 7, 2026
View reviewed changes

potiuk closed this Jan 7, 2026

jedcunningham deleted the fix_inthewild_sorting branch January 7, 2026 18:52

Fix basic static checks comparing same commit with itself
b9f20ea
The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

jedcunningham requested review from amoghrajesh, ashb, bugraoz93, gopidesupavan, jason810496, jscheffl, kaxil and potiuk as code owners January 6, 2026 23:42

boring-cyborgBot added area:dev-tools backport-to-v3-1-test labels Jan 6, 2026

jedcunningham requested a review from Copilot January 6, 2026 23:43

CopilotAI reviewed Jan 6, 2026
View reviewed changes

Comment thread.github/workflows/basic-tests.yml
Comment thread.github/workflows/basic-tests.yml

bugraoz93 approved these changes Jan 7, 2026
View reviewed changes

amoghrajesh approved these changes Jan 7, 2026
View reviewed changes

potiuk reviewed Jan 7, 2026
View reviewed changes

potiuk closed this Jan 7, 2026

jedcunningham deleted the fix_inthewild_sorting branch January 7, 2026 18:52

Fix basic static checks comparing same commit with itself
b9f20ea
The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

jedcunningham requested review from amoghrajesh, ashb, bugraoz93, gopidesupavan, jason810496, jscheffl, kaxil and potiuk as code owners January 6, 2026 23:42

boring-cyborgBot added area:dev-tools backport-to-v3-1-test labels Jan 6, 2026

jedcunningham requested a review from Copilot January 6, 2026 23:43

CopilotAI reviewed Jan 6, 2026
View reviewed changes

Comment thread.github/workflows/basic-tests.yml
Comment thread.github/workflows/basic-tests.yml

bugraoz93 approved these changes Jan 7, 2026
View reviewed changes

amoghrajesh approved these changes Jan 7, 2026
View reviewed changes

potiuk reviewed Jan 7, 2026
View reviewed changes

potiuk closed this Jan 7, 2026

jedcunningham deleted the fix_inthewild_sorting branch January 7, 2026 18:52

Fix basic static checks comparing same commit with itself
b9f20ea
The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

jedcunningham requested review from amoghrajesh, ashb, bugraoz93, gopidesupavan, jason810496, jscheffl, kaxil and potiuk as code owners January 6, 2026 23:42

boring-cyborgBot added area:dev-tools backport-to-v3-1-test labels Jan 6, 2026

jedcunningham requested a review from Copilot January 6, 2026 23:43

CopilotAI reviewed Jan 6, 2026
View reviewed changes

Comment thread.github/workflows/basic-tests.yml
Comment thread.github/workflows/basic-tests.yml

bugraoz93 approved these changes Jan 7, 2026
View reviewed changes

amoghrajesh approved these changes Jan 7, 2026
View reviewed changes

potiuk reviewed Jan 7, 2026
View reviewed changes

potiuk closed this Jan 7, 2026

jedcunningham deleted the fix_inthewild_sorting branch January 7, 2026 18:52

Fix basic static checks comparing same commit with itself
b9f20ea
The prek command was comparing `github.sha` with itself, which checked no files. Now uses `merge-base` to check all files changed in the PR against the base branch.

jedcunningham requested review from amoghrajesh, ashb, bugraoz93, gopidesupavan, jason810496, jscheffl, kaxil and potiuk as code owners January 6, 2026 23:42

boring-cyborgBot added area:dev-tools backport-to-v3-1-test labels Jan 6, 2026

jedcunningham requested a review from Copilot January 6, 2026 23:43

CopilotAI reviewed Jan 6, 2026
View reviewed changes

Comment thread.github/workflows/basic-tests.yml
Comment thread.github/workflows/basic-tests.yml

bugraoz93 approved these changes Jan 7, 2026
View reviewed changes

amoghrajesh approved these changes Jan 7, 2026
View reviewed changes

potiuk reviewed Jan 7, 2026
View reviewed changes

potiuk closed this Jan 7, 2026

jedcunningham deleted the fix_inthewild_sorting branch January 7, 2026 18:52

Uh oh!

Conversation

jedcunningham commented Jan 6, 2026

Uh oh!

CopilotAI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

bugraoz93Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

bugraoz93Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

amoghrajesh left a comment

Choose a reason for hiding this comment

Uh oh!

potiuk left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

potiuk commented Jan 7, 2026

Uh oh!

potiuk commented Jan 7, 2026• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

potiuk commented Jan 7, 2026

Uh oh!

potiuk commented Jan 7, 2026

Uh oh!

potiuk commented Jan 7, 2026

Uh oh!

jedcunningham commented Jan 7, 2026• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

potiuk commented Jan 7, 2026

Uh oh!

potiuk commented Jan 7, 2026• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

potiuk commented Jan 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Uh oh!

Conversation

jedcunningham commented Jan 6, 2026

Uh oh!

CopilotAI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

bugraoz93Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

bugraoz93Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

amoghrajesh left a comment

Choose a reason for hiding this comment

Uh oh!

potiuk left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

potiuk commented Jan 7, 2026

Uh oh!

potiuk commented Jan 7, 2026• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

potiuk left a comment •
edited
Loading

potiuk commented Jan 7, 2026•
edited
Loading

jedcunningham commented Jan 7, 2026•
edited
Loading

potiuk commented Jan 7, 2026•
edited
Loading

potiuk left a comment •
edited
Loading

potiuk commented Jan 7, 2026•
edited
Loading

jedcunningham commented Jan 7, 2026•
edited
Loading

potiuk commented Jan 7, 2026•
edited
Loading

potiuk left a comment •
edited
Loading

potiuk commented Jan 7, 2026•
edited
Loading