Seemingly, we overlooked some issue when checking the debug patches. Compiling
octopus (tddft.org) with gfortran now gives an ICE:
$ gfortran -g tmp1.f90 tmp.f90
tmp.f90:2: internal compiler error: Segmentation fault
Valgrind shows (when compiling tmp.f90, tmp1.f90 is OK):
==8754== Invalid read of size 1
==8754== at 0xB5B660: htab_hash_string (hashtab.c:812)
==8754== by 0x54B12E: lookup_filename (dwarf2out.c:15573)
==8754== by 0x557B86: add_src_coords_attributes (dwarf2out.c:12442)
==8754== by 0x558B1A: add_name_and_src_coords_attributes (dwarf2out.c:12459)
==8754== by 0x560FC0: gen_decl_die (dwarf2out.c:15040)
==8754== by 0x563BD7: force_decl_die (dwarf2out.c:14940)
==8754== by 0x56559C: dwarf2out_imported_module_or_decl (dwarf2out.c:15344)
==8754== by 0x4A15DC: gfc_trans_use_stmts (trans-decl.c:3178)
==8754== by 0x4A180E: gfc_generate_module_vars (trans-decl.c:3385)
==8754== by 0x48F07C: gfc_generate_module_code (trans.c:1226)
Test files (note: There is no ICE if one puts the two files into a single
file).
---------- tmp1.f90 -------------
module c_pointer_m
use iso_c_binding, only : c_ptr, c_associated, c_null_ptr
implicit none
contains
subroutine set_null(ptr)
type(c_ptr), intent(out) :: ptr
ptr = c_null_ptr
end subroutine set_null
end module c_pointer_m
------------ tmp.f90 ---------------
module string_m
use c_pointer_m
implicit none
end module string_m
--
Summary: [4.4 Regression] ICE (segfault) with new debugging patch
Product: gcc
Version: 4.4.0
Status: UNCONFIRMED
Keywords: ice-on-valid-code
Severity: normal
Priority: P3
Component: debug
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: burnus at gcc dot gnu dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37287
