Re: Flex/Bison standardized diagnostics

[Hans Aberg]

At 16:10 +0100 2002/01/07, Akim Demaille wrote:
>Hans> Then tuning it together with Bison's location feature would be
>Hans> suitable, because then the parser would neither record anything
>Hans> but the essentials needed at any eventual error time.
>
>I would like to emphasize that Bison's main feature is the
>_propagation_ of locations (whatever their structure/contents).  It
>turns out it provides a default structure for locations,

I am well aware of that:

> but that's
>only history.

Then the default locations feature might be changed so that it is a
structure that contains the information needed for the other computations
used in various applications.

This might make that old Bison feature usable again. This was in fact one
of the starting points of my reasoning.

>  The core feature is simply that there is a third stack,
>and a user settable default action for them.

I merged the stacks into a single one, and then the locations is just an
addition to the semantic value, with a special type of default rule (the
location of $$ is bounded by the $n).

  Hans Aberg