[antlr-interest] Why is ANTLR getting lost?

[Jim Idle]

That rule should really be simplified, but try that 3.1 snapshot.

Jim



> -----Original Message-----
> From: shmuel siegel [mailto:antlr at shmuelhome.mine.nu]
> Sent: Wednesday, February 20, 2008 6:04 AM
> 

> I can't tell you why but I can tell you that the dfa predictor is
> failing on
>         (addExpr[true] EQ) => addExpr[true] EQ addExpr[false]
>       | (number LE addExpr[true] LE) => lb=number LE addExpr[false] LE
> ub=number
>       | (addExpr[true] LE) => ae+=addExpr[true] LE ae+=addExpr[false]
> 
> Note that the second alternative satisfies the conditions of the
> condition of the third alternative. If you eliminate the third
> alternative, the rule works. The rule also works if you change the
> third
> alternative to be different then the second, say by adding LE before
> addExpr. I suspect that something is wrong with the dfa predictor.
> 
>