[Bug ipa/97346] New: IPA reference reference_vars_to_consider leaks

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97346

            Bug ID: 97346
           Summary: IPA reference reference_vars_to_consider leaks
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: ipa
          Assignee: unassigned at gcc dot gnu.org
          Reporter: rguenth at gcc dot gnu.org
                CC: marxin at gcc dot gnu.org
  Target Milestone: ---

==20551== 96 (8 direct, 88 indirect) bytes in 1 blocks are definitely lost in
loss record 535 of 730
==20551==    at 0x4C2E94F: operator new(unsigned long) (in
/usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
==20551==    by 0x14A6E6B: void vec_alloc<tree_node*>(vec<tree_node*, va_heap,
vl_ptr>*&, unsigned int) (vec.h:1567)
==20551==    by 0x24BB41F: ipa_init() (ipa-reference.c:461)
==20551==    by 0x24BB896: generate_summary() (ipa-reference.c:597)
==20551==    by 0x24BC18A: propagate() (ipa-reference.c:768)
==20551==    by 0x24BD88C: (anonymous
namespace)::pass_ipa_reference::execute(function*) (ipa-reference.c:1299)
==20551==    by 0x121A924: execute_one_pass(opt_pass*) (passes.c:2509)
==20551==    by 0x121B893: execute_ipa_pass_list(opt_pass*) (passes.c:2936)
==20551==    by 0xC6ED1D: ipa_passes() (cgraphunit.c:2700)

the use of this global is quite odd, with two allocations, one release
site using just 'delete' and one conditional freeing the vector but
unconditionally NULLing the pointer to it.

I wonder if the array should be made local to a few functions to make
lifetime more obvious.