Re: JIT stealth font-lock tuning parameters

[Stefan Monnier]

>     > Maybe that parameter should be reduced on a slow machine.
> 
>     I'm not sure it would help.
> 
> Would you please try it and see?  Try it with syntax-ppss, since that
> is what we will be using, and that has the best chance of actually
> improving this.

I'm afraid I can't.  I don't see those pauses except for the
ones that are reduced by syntax-ppss (which I have indeed tested and
confirmed that it tends to reduce their frequency).
Note that the worst case is not improved: if font-lock requires
the parse-partial-sexp-state at (point) and the syntax-ppss cache
is empty, there's nothing we can do, really.  But that worst case
is extremely unlikely to happen from the stealthy part of jit-lock.

Another line of attack is a local hack (that I haven't cleaned up
yet) which adds a variable `quit-on-input' which forces any input
to cause a quit.
This way you can run a very lengthy computation inside a
`with-quit-on-input' and any input will interrupt the computation
without having to poll with `sit-for'.

This could completely eliminate those pauses, but it will require
some work to clean up the `quit-on-input' thing and then to decide
how/when quit should be allowed during jit-locking.


        Stefan