Returning an error message from a semantic action

Semantic actions are really powerful.
But their semantic is very difficult to figure out for non-js (or
non-SPARQL) gurus.
Especially error messages such as:
  eval of { js_var_p2 = _.o; alert(js_var_p2); return js_var_p1 ==
js_var_p2; } rejected [[
    <http://ex.example/#p2> <http://www.w3.org/2001/XMLSchema#string> %js{
js_var_p2 = _.o; alert(js_var_p2); return js_var_p1 == js_var_p2; %}
matched by <Issue1> <http://ex.example/#p2> "p2".
    ]]
are too cryptic w.r.t the semantic of the action.

How could semantic actions return explicit error messages, in addition to
the true/false boolean ?