teardown deletes repository before Sidekiq worker
This happens frequently on development setups, in the teardown of functional tests:
message: No such file or directory @ rb_sysopen - /home/gracinet/heptapod/hdk/release-0.29/repositories/@hashed/53/5f/535fa30d7e25dd8a49f1536779734ec8286108d115da5045d77f3b4185d8f790.hg/.hg/hgrc
backtrace:
lib/gitlab/mercurial/hg_git_repository.rb:1596:in `readlines'
lib/gitlab/mercurial/hg_git_repository.rb:1596:in `rpath_to_namespace_if_hgrc_included'
lib/gitlab/mercurial/hg_git_repository.rb:221:in `hg_rename'
lib/gitlab/mercurial/hg_git_repository.rb:214:in `rename'
lib/gitlab/shell.rb:97:in `mv_repository'
app/services/repositories/base_service.rb:23:in `mv_repository'
app/services/repositories/destroy_service.rb:13:in `execute'
app/services/projects/destroy_service.rb:91:in `remove_repository'
app/services/projects/destroy_service.rb:55:in `trash_project_repositories!'
app/services/projects/destroy_service.rb:129:in `destroy_project_related_records'
app/services/projects/destroy_service.rb:123:in `attempt_destroy'
app/services/projects/destroy_service.rb:35:in `execute'
app/workers/project_destroy_worker.rb:17:in `perform'
This not new – I've dismissed it as an inconsistency due to deletion – but it should be a real bug.