GitLab

  This removes the need for the exception [Accept].
Modified the table back-end to include the number of start productions
  as part of the generated tables.

I see a 5% performance gain, from 2.0 to 1.9.
If we could remove the tests altogether, we would go down to 1.8.

as [MenhirLib.Convert] has already committed to this type.

This requires storing a triple in [env.token], which becomes [env.triple].
The ratio table/code seems to go from 2.0 up to 2.1.

that it does not write [env.current]. This was easy, as the write
to [env.current] was in fact unobservable and could be replaced
with a write to a local reference [current]. At the same time,
changed the local recursive function [pop] to a loop, for greater
readability.

for updating [env.stack]. Instead, it returns the new stack.
This corresponds to Frédéric's "Isolate side-effects from table-backend generated code".

git-svn-id: svn+ssh://scm.gforge.inria.fr/svnroot/menhir@320 0f8b5475-4b4e-0410-85a8-ee3154a6bfe7