[Bug tree-optimization/13741] [3.3/3.4 Regression] [tree-ssa] Variable-length arrays broke in gcc 3.x

["reichelt at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>]

------- Additional Comments From reichelt at gcc dot gnu dot org  2004-06-23 14:47 -------
Fixed on mainline by Richard Kenner's patch

http://gcc.gnu.org/ml/gcc-cvs/2004-06/msg00881.html

But there are two open questions:

a) Will this be fixed on the 3.3 and/or 3.4 branch?

   I'd say no, since
   * The fix is tree-ssa specific and can't be backported.
   * The testcase involves nested functions which happen to be broken
     in many cases anyway.
   * It's not really a regression. The testcase in comment #2 did
     compile with 2.95.3, but Andrew's example in comment #15 did not
     (I get a tree checking failure with checking enabled).

b) Are we going to have a testcase in the testsuite?

   Here's a reduced version of Andrew's example that I'd propose as
   a candidate:

   =========================
   void foo(int i)
   {
       typedef struct
       {
           char c[i];
           int  j;
       } A;

       void bar(A *a)
       {
           a->j=0;
       }
   }
   =========================


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|tree-ssa                    |3.3.4 3.4.1
      Known to work|2.95.3                      |3.5.0
            Summary|[3.3/3.4/3.5 Regression]    |[3.3/3.4 Regression] [tree-
                   |[tree-ssa] Variable-length  |ssa] Variable-length arrays
                   |arrays broke in gcc 3.x     |broke in gcc 3.x


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13741