Why is this a problem? You said "allow empty" so shouldn't it allow the action???

Why is this a problem? You said "allow empty" so shouldn't it allow the action???

Since we're checking if the content of the two branches is different, adding the "allow-empty" flag introduces the option to merge branch_a branch_a --allow-empty.
Although it can technically work, it's something we'd rather prevent from happening (mainly to prevent user mistakes, I think).

I have a question about desired output of lakectl when trying to merge a branch to itself.

Currently, my PR changes the server only, and the output of lakectl is:

update branch main: same branch validation error  
400 Bad Request

Should the "400 Bad Request" line appear? it seems to expose implementation details unnecessarily, especially since we already have a descriptive error message.

I checked the output in a similar case where the server returns an error response - trying to make an empty commit without the "allow empty" flag. In this case the output also shows the Bad Request line.

commit: no changes  
400 Bad Request

I believe the output should be consistent in both cases, but from the same reason I think this line should not appear on both.
What do you think?

Should the "400 Bad Request" line appear? it seems to expose implementation details unnecessarily, especially since we already have a descriptive error message.

Personally I believe that it should: unlike unformatted error text, an http code is an unambiguous specification.

If you disagree, we could discuss on Slack or on a separate issue.

I talked to @ozkatz , here's his suggestion: do what Git does in this situation. I'll check and update here.

What Git does

This is actually impossible on Git! Git fails any merge from an ancestor of the branch HEAD, with the message "Already up to date.". This includes merging the branch HEAD itself to the branch

For instance:

❯ git log
commit bcf5a09671562551797fa747c43d9f86e5b30aef (HEAD -> main)
Author: Ariel Shaqed (Scolnicov) <[email protected]>
Date:   Thu Dec 5 19:57:07 2024 +0200

[...]

commit b5aa1a1b17a8bfb4d55f37e1d8a0395fb96773d5
Author: Ariel Shaqed (Scolnicov) <[email protected]>
Date:   Thu Dec 5 19:45:40 2024 +0200

[...]
❯ git merge main~
Already up to date.
❯ git merge b5aa
Already up to date.
❯ git merge main
Already up to date.

Proposal

Do what Git does: forbid the merge if a commit is an ancestor of the branch head. This has multiple advantages:

• Does what Git does.
• Prevent merging a branch HEAD to itself.
• Prevent other empty merges.
• Continue to treat a merge as going from a commit to a branch.