, yeah, that's what I'm implying. The Emacs
autosave mechanism doesn't overwrite the user's files with the contents of a buffer (very much intentionally), but real-auto-save does: see e.g. <https://github.com/ChillarAnand/real-auto-save/blob/master/real-auto-save.el#L84
> (assuming this is the right thing, which it looks like it is).
Undo still works across real saves, of course, but if you do a real save and them Emacs crashes or the power goes out, or, y'know, any of the things you might have been avoiding saving for... you're stuffed.
This is more problematic yet for those of us (like, well, you and me) who work on things have to be in consistent states for things like compilers and interpreters to accept them. I dunno about you, but I'm all the time compiling things that already exist in a modified intermediate uncompilable state in some buffer somewhere: finding that Emacs has silently overwritten the compilable copy with the copy in the buffer would be ferociously annoying. (Maybe I am more annoyed than most by this sort of thing, but people have suggested making auto-save work the way real-auto-save does on the emacs development list in the past, and the pushback was pretty fierce. Mind you, the pushback against everything
is fierce on that list. :) )