Compiling semi_empirical_int_ana.f90 of CP2k (cf. PR 29975) gives an ICE:
(with 4.3.0 20070526 on x86_64-unknown-linux-gnu)
gfortran -c -O3 -ftree-vectorize -ffast-math -march=opteron semi_empirical_int_ana.f90
semi_empirical_int_ana.f90: In function 'dterep_ana':
semi_empirical_int_ana.f90:2321: internal compiler error: Segmentation fault
==3451== Conditional jump or move depends on uninitialised value(s)
==3451== at 0x7A34F8: vrp_evaluate_conditional_warnv (tree-vrp.c:4814)
==3451== by 0x7A3B99: vrp_evaluate_conditional (tree-vrp.c:4946)
==3451== by 0x76515B: substitute_and_fold (tree-ssa-propagate.c:1131)
==3451== by 0x79EAD8: vrp_finalize (tree-vrp.c:5925)
==3451== by 0x7A03BA: execute_vrp (tree-vrp.c:6011)
==3451== by 0x609B40: execute_one_pass (passes.c:1067)
==3451== by 0x609CFB: execute_pass_list (passes.c:1119)
==3451== by 0x609D0D: execute_pass_list (passes.c:1120)
==3451== by 0x6D3B27: tree_rest_of_compilation (tree-optimize.c:406)
==3451== by 0x8129CF: cgraph_expand_function (cgraphunit.c:1086)
==3451== by 0x814E71: cgraph_optimize (cgraphunit.c:1155)
==3451== by 0x466F9C: gfc_be_parse_file (f95-lang.c:307)
Created attachment 13614 [details]
test file (2540 lines)
Test case, I failed to reduce the size.
It worked with 2007-05-14-r124708 and fails with 2007-05-15-r124736.
The patch causing (or exposing) the ICE is:
r124727 | rakdver | 2007-05-14 23:58:42 +0200 (Mon, 14 May 2007) | 46 lines
Zdenek, could you have a look?
I'd think this looks like sth for ian.
I think this is fixed by this patch:
I am currently testing that patch.
Patch committed to mainline as revision 12503. The test case works for me.
*** Bug 32521 has been marked as a duplicate of this bug. ***