analyzer: fix ICE on pointer arithmetic with incomplete types [PR 93774]

PR analyzer/93774 reports an ICE in gfortran with -fanalyzer within
region_model::convert_byte_offset_to_array_index on a pointer of
incomplete type ("character(kind=1)[0:][1:0] * restrict").

This patch bulletproofs the routine against incomplete types, fixing
the ICE.

gcc/analyzer/ChangeLog:
	PR analyzer/93774
	* region-model.cc
	(region_model::convert_byte_offset_to_array_index): Use
	int_size_in_bytes before calling size_in_bytes, to gracefully fail
	on incomplete types.

gcc/testsuite/ChangeLog:
	PR analyzer/93774
	* gfortran.dg/analyzer/deferred_character_25.f90: New test,
	based on gfortran.dg/deferred_character_25.f90.

gcc/analyzer/ChangeLog

+2020-02-18  David Malcolm  <dmalcolm@redhat.com>
+
+	PR analyzer/93774
+	* region-model.cc
+	(region_model::convert_byte_offset_to_array_index): Use
+	int_size_in_bytes before calling size_in_bytes, to gracefully fail
+	on incomplete types.
+
 2020-02-17  David Malcolm  <dmalcolm@redhat.com>
 
 	PR analyzer/93775

gcc/analyzer/region-model.cc

@@ -6514,16 +6514,18 @@ region_model::convert_byte_offset_to_array_index (tree ptr_type,
 
       /* Arithmetic on void-pointers is a GNU C extension, treating the size
 	 of a void as 1.
-	 https://gcc.gnu.org/onlinedocs/gcc/Pointer-Arith.html
-
-	 Returning early for this case avoids a diagnostic from within the
-	 call to size_in_bytes.  */
+	 https://gcc.gnu.org/onlinedocs/gcc/Pointer-Arith.html  */
       if (TREE_CODE (elem_type) == VOID_TYPE)
 	return offset_sid;
 
-      /* This might not be a constant.  */
+      /* First, use int_size_in_bytes, to reject the case where we have an
+	 incomplete type, or a non-constant value.  */
+      HOST_WIDE_INT hwi_byte_size = int_size_in_bytes (elem_type);
+      if (hwi_byte_size > 0)
+	{
+	  /* Now call size_in_bytes to get the answer in tree form.  */
 	  tree byte_size = size_in_bytes (elem_type);
-
+	  gcc_assert (byte_size);
 	  /* Try to get a constant by dividing, ensuring that we're in a
 	     signed representation first.  */
 	  tree index
@@ -6533,6 +6535,7 @@ region_model::convert_byte_offset_to_array_index (tree ptr_type,
 	  if (index && TREE_CODE (index) == INTEGER_CST)
 	    return get_or_create_constant_svalue (index);
 	}
+    }
 
   /* Otherwise, we don't know the array index; generate a new unknown value.
      TODO: do we need to capture the relationship between two unknown

gcc/testsuite/ChangeLog

 2020-02-18  David Malcolm  <dmalcolm@redhat.com>
 
+	PR analyzer/93774
+	* gfortran.dg/analyzer/deferred_character_25.f90: New test,
+	based on gfortran.dg/deferred_character_25.f90.
+
+2020-02-18  David Malcolm  <dmalcolm@redhat.com>
+
 	PR analyzer/93779
 	* gfortran.dg/analyzer/pr88304-2.f90: New test, adapted from
 	gfortran.fortran-torture/compile/pr88304-2.f90

gcc/testsuite/gfortran.dg/analyzer/deferred_character_25.f90

+! { dg-do compile }
+! { dg-additional-options "-Wno-analyzer-too-complex" }
+
+! Copy of gfortran.dg/deferred_character_25.f90
+! as a regression test for ICE with -fanalyzer (PR analyzer/93774)
+
+PROGRAM TEST
+  IMPLICIT NONE
+  INTEGER, PARAMETER :: I = 3
+  character (len = i), parameter :: str(5) = ['abc','cde','fgh','ijk','lmn']
+
+  TYPE T
+    CHARACTER(LEN=:), ALLOCATABLE :: C(:)
+  END TYPE T
+  TYPE(T), TARGET :: S
+  CHARACTER (LEN=I), POINTER :: P(:)
+
+  ALLOCATE ( CHARACTER(LEN=I) :: S%C(5) )
+  s%c = str
+
+! This PR uncovered several problems associated with determining the
+! element length and indexing. Test fairly thoroughly!
+  if (SIZE(S%C, 1) .ne. 5) stop 1
+  if (LEN(S%C) .ne. 3) stop 2
+  if (any (s%c .ne. str)) stop 3
+  if (s%c(3) .ne. str(3)) stop 4
+  P => S%C
+  if (SIZE(p, 1) .ne. 5) stop 5
+  if (LEN(p) .ne. 3) stop 6
+  if (any (p .ne. str)) stop 7
+  if (p(5) .ne. str(5)) stop 8
+END PROGRAM TEST