`Test_MigrateFromGiteaToGitea` fails when gitea.com is unavailable

[silverwind]

Written by Claude on behalf of @silverwind.

Description

Test_MigrateFromGiteaToGitea in tests/integration/migrate_test.go fails intermittently in CI with a timeout, causing unrelated PRs to show red CI.

Example failure

From PR #36858 CI run (test-sqlite job):

+++ Test_MigrateFromGiteaToGitea is a slow test (run: 2m18.102875547s, flush: 87.112µs)
--- FAIL: Test_MigrateFromGiteaToGitea (138.10s)
    testlogger.go:62: 2026/03/08 07:33:40 HTTPRequest [W] router: slow      POST /api/v1/repos/migrate for test-mock:12345, elapsed 3038.6ms @ repo/migrate.go:36(repo.Migrate)
    testlogger.go:62: 2026/03/08 07:35:53 HTTPRequest [I] router: completed POST /api/v1/repos/migrate for test-mock:12345, 500 Internal Server Error in 136206.5ms @ repo/migrate.go:36(repo.Migrate)
    migrate_test.go:153: Response:  {"message":"","url":"http://localhost:3003/api/swagger"}

The migration call takes over 2 minutes and returns a 500 error with an empty message.

Impact

This causes false-negative CI results on unrelated PRs, making it harder to identify real test failures.

[silverwind]

I think the root cause is that gitea.com is currently unavailable and the test depends on repo https://gitea.com/gitea/test_repo.git. We should add a mock for it.

[silverwind]

Still an issue it seems: https://github.com/go-gitea/gitea/actions/runs/22822297451/job/66196881289

I think we ought to mock this. We can't have tests depend on external systems.

[wxiaoguang]

Yes, it needs to be mocked.

The problem is that the site is not "completely" down. Sometimes it responds, then the first "GET" check occasionally passes, but most requests still fail.

[silverwind]

I'm checking regarding a mock.