In Stash 3.7, code was added which automatically reenables gc.auto when a repository has no forks. The intent was to undo the GC changes made when a repository is forked when the last fork was deleted, and that works well.

However, some instances have repositories which trigger GC frequently, or for which GC frequently fails (usually due to pull request refs being updated). For such instances, the code to automatically reenable gc.auto effectively eliminates the simplest way an administrator could work around the error.

Rather than just always reenabling gc.auto when a repository has no forks, instead we should reenable it only when gc.pruneexpire=never is also set. Our forking code always sets both, so we can use that as a reliable mechanism for determining whether we're reverting fork GC changes or a repository administrator has disabled auto GC.