bug#13149: 24.3.50; Emacs thinks file was changed outside Emacs, but it was not

[Dmitry Gutov]

On 18.01.2013 1:14, Paul Eggert wrote:
> On 01/17/13 02:32, Dmitry Gutov wrote:
> > I think we need this to work without requiring the user to customize a variable.
>
> I agree.  But I would rather avoid having Emacs assume
> that file system time stamps can be off by several seconds,
> since that'll cause Emacs to miss changes that it should report.

I'd like to report that after shutdown, reboot and whatever, my 
yesterday's build of emacs-24 on Ubuntu is again buggy on vboxsf. But it 
still works fine in the cifs-mounted folder, so if all else fails, I 
guess we can still revert the commit I mentioned previously and 
recommend people to avoid vboxsf. Maybe then change that code 
piece-by-piece to see what causes the breakage.

> Does the following patch help?  It should be applied anyway,
> since it closes a minor race even on non-buggy file systems.
> I'm hoping that it helps to work around your bug by using
> fstat both times, instead of fstat in one place and stat
> in another, since the file system bug seems to be that
> fstat and stat disagree.

Alas, no dice. It didn't help neither with vboxsf, nor with cifs.