[tree-ssa] diagnosing switch errors

[Jeff Sturm]

On Thu, 30 Oct 2003, Paul Brook wrote:
> On Thursday 30 October 2003 6:05 pm, Richard Henderson wrote:
> > So I went looking to see what Java and Fortran do.  And I can't
> > see that we check for duplicate case labels at all.  Am I correct
> > in this, or have I missed something?
>
> This is done much earlier on by gfortran. See fortran/resolve.c
> (resolve_select/check_case_overlap). By the time we generate trees for the
> switch we already know all the cases are valid. Fortran needs to be able to
> do switch statements for character (ie. string) variables, so checking at
> the GENERIC/GIMPLE level probably wouldn't gain us much.

Similarly, Java needs to handle duplicate case labels before emitting
bytecode, so any checking done in gimplify is redundant.

Jeff