On Tue, Nov 27, 2018 at 09:56:43AM +0100, Johannes Thumshirn wrote: > On 26/11/2018 19:37, David Sterba wrote: > > I though a transaction abort would be here, as the state is not > > consistent. Also I'm not sure what I as a user would get from such error > > message after calling link(). If the error handling in the error > > handling fails, there's not much left to do and the abort either > > happened earlier in the callees or is necessary here. > > OK, now you've lost me. Isn't that what my previous patch did? Yes (call abort in fail_dir_item:) but it also merged all the abort calls -- this is the part that I wanted to be updated.
