Re: Patch for fields of `struct buffer'

[Stefan Monnier <monnier@IRO.UMontreal.CA>]

>> This issue arises now within the single Emacs thread, and we already
>> handle it.
>     Not really, because we never rewind, we only unwind.

> Isn't rewinding the reverse of unwinding?
> If we have got unwinding right (or close enough),
> I'd expect that makes it easy to do the rewinding right.

I don't share your optimism, although it is true that with the advantage
of the 20 years experience getting unwinding to work right, we may be
able to get the rewinding to work right in less time.


        Stefan


PS: BTW, talking about concurrency, I was recently reminded of an
article I read not that long ago which advocated the use of yield-points
instead of transactions (or rather, to define transactions as "what
happens between two yields").  Part of the argument was that the usual
syntactic nesting traditionally imposed by transactions can make some
things really cumbersome, but another aspect was that it is safer since
a missing yield only incurs a performance loss rather than
a race-condition.
It sounds like that might be a good approach for Elisp, but I can't find
this article again and my memory of it is rather vague, so I may be
imagining some of the benefits, and I definitely do not remember what
were the pitfalls.  If that rings a bell for anyone...