Re: {file} "locked by ... pid NNN", but there is no process NNN

[Eli Zaretskii <eliz@gnu.org>]

> Date: Tue, 19 Jan 2016 02:40:21 -0800 (PST)
> From: djc <peter.kaiser@gmail.com>
> Injection-Date: Tue, 19 Jan 2016 10:40:22 +0000
> 
> The situation was too complex and too intermittent to simplify easily for diagnosis, but I think I know generally what's happening: emacs is failing to delete its own lock of a file that I was working on once, but either (1) whose buffer has been killed or (2) where the lock has survived exiting and re-opening emacs.

This never happens to me, except if my system crashes in the middle of
editing a file.  I have just searched my entire filesystem for
user-lock files, and found none.  This is my main development machine
where a large Emacs session runs for weeks on end, in which I edit
hundreds of files.  If something like that was happening even with a
small probability, I should have had several such files around.  But I
don't.

> On Windows, emacs also doesn't necessarily release Windows handles of files and directories for which it no longer has any associated buffer.  I've seen this repeatedly.

In what Emacs version is that, in what version of Windows, and with
what kind of filesystem (NTFS, FAT32, network)?  I have here Emacs
24.5 on XPSP3 with NTFS volumes, which in this case runs for 11 days
and 9 hours, and Process Explorer shows it holds a handle for a single
directory -- my home directory -- and none of the almost 200 files
this session visits.

Of course, if you can at least describe under which circumstances
these problems tend to happen, please do.  And please report that
using "M-x report-emacs-bug", not here.

Thanks.