MySQL/MariaDB: Unique Constraint on Branches doesn't respect casing: UQE_branch_s

[sebastian-sauer]

Description

In a repo with two branches feature/test_case and feature\Test_Case gitea throws a 500 exception when syncing branches as the unique constraint is violated.

SyncRepoBranches: Error 1062 (23000): Duplicate entry '<redacted_branch_name>' for key 'UQE_branch_s'

This error happened after an upgrade to 1.21.0 on an existing repo.

Workaround is to rename one of the branches.

Typically having this type of different branch names only based on casing is not needed (imho it's a bug ;) ) - but we shouldn't have a 500 page when accessing the repo.

Gitea Version

1.21.0

Can you reproduce the bug on the Gitea demo site?

No

Log Gist

No response

Screenshots

No response

Git Version

No response

Operating System

Linux

How are you running Gitea?

Official binary

Database

MySQL/MariaDB

[yorliuk]

I got the same error in Gitea log after updating to 1.21.
...dules/context/api.go:105:ServerError() [E] SyncRepoBranches: Error 1062 (23000): Duplicate entry '<BRANCH_NAME>' for key 'branch.UQE_branch_s'

But I can't find such branch in DB.
mysql Ver 8.0.28 for Linux on x86_64 (MySQL Community Server - GPL)

[silverwind]

Related: #25909

[sebastian-sauer]

Are we "allowed" to remove a unique constraint in a patch release?

[yorliuk]

I got the same error in Gitea log after updating to 1.21. ...dules/context/api.go:105:ServerError() [E] SyncRepoBranches: Error 1062 (23000): Duplicate entry '<BRANCH_NAME>' for key 'branch.UQE_branch_s'

But I can't find such branch in DB. mysql Ver 8.0.28 for Linux on x86_64 (MySQL Community Server - GPL)

It helped: #25909 (comment)

[silverwind]

Are we "allowed" to remove a unique constraint in a patch release?

Anything reasonable is possible, but I think the solution is to make the field case-sensitive.

[wxiaoguang]

For 1.21 (unlucky...) users, the workaround is to run the SQL manually:

ALTER TABLE branch MODIFY COLUMN `name` VARCHAR(255) CHARACTER SET utf8mb4 COLLATE utf8mb4_bin NOT NULL

It won't cause side-effect and won't conflict with the complete fix (if there is anything wrong, it could be revert to old COLLATE utf8mb4_general_ci)