Hi, > I will try to implement something like this, and likely I will ask you > to test in your environment. Would this be OK?
Certainly. However, I am still investigating to pinpoint the problem more precisely. I am pretty sure now that file-attributes is not the problem. I artificially delayed stat on the remote end such that I was able to start different foreground actions with defined interleaving when file-attributes was called from a sentinel. It appears that Emacs is blocked completely during file-attributes, such that it is impossible for the user to trigger anything. This rules out the scenario that I described in the previous message for file-attributes. I haven't checked delete-file, but I have hints that file-truename is problematic. Maybe there is a simpler fix by ensuring that the remote actions inside file-attributes block Emacs? Best regards, Hendrik
