On 01/03/2016 03:21 PM, Paul Eggert wrote:
> John Wiegley wrote:
>> I'd like to know what others think (maybe they are ignoring this
>> thread now,
>> so I've changed the subject). But I feel like there is some
>> convergence now.
> 
> I didn't understand all the details of the proposal. That being said,
> I'm not sure I like the idea of losing my Emacs session merely because
> some regular expression got too long. Sure, Emacs would create a bunch
> of autosave files that I can recover file data from, but if my session
> has been going for hours or days it'll have quite a bit of state that
> isn't in any buffer associated with a file, and I'd rather not lose that
> state. So although it'd be fine to have async-signal-safe autosave as an
> option, as a user I'd like to also retain the option for Emacs to keep
> doing the same old unsafe stuff that it's been doing for decades, at
> least until after we get in-the-field experience with the new approach.

As I've mentioned elsewhere, I'm perfectly happy making the regular
expression code a special case, since we own it and know what it does. I
agree that Emacs absolutely shouldn't crash on any Lisp input to
re-search-forward, looking-at, etc.