Re: The speed of Emacs 21

[Eli Zaretskii]

On Tue, 31 Oct 2000, John Wiegley wrote:

> > Yes, this is turned on by default to hunt an elusive memory-related
> > bug.  It certainly slows down Emacs significantly.  However, it does
> > not explain the massive consing that John reports (assuming that the
> > previous versions didn't do all that consing).
> 
> If you try the e21 from around Jun/Aug, it did not have these speed
> problems at all.

Did you actually see that there was much less consing in that Jun/Aug
version?

The slow-down due to checking strings is expected and I also see it.
However, if we want to catch that bug, we will probably have to live
with it through the pretest (I expect it to be turned off in the
release version).  It's the massive consing that I don't understand; I
think there were no changes since the summer that could explain that.
In fact, some changes were introduced to reduce consing.