[PATCH v2 0/3] Add array_view copy function and more

[PATCH v2 0/3] Add array_view copy function and more

[Simon Marchi]

This is a v2 of this patch and following (patch 1/4 of the original
series has been merged):

  https://sourceware.org/pipermail/gdb-patches/2021-November/183250.html

The changes are in patch "gdbsupport: add array_view copy function".
Unit tests were added, and the copy function is now implemented using
std::copy and std::copy_backwards.

Simon Marchi (3):
  gdbsupport: add array_view copy function
  gdb: make extract_integer take an array_view
  gdb: trivial changes to use array_view

 gdb/ada-lang.c                       |  17 ++---
 gdb/defs.h                           |  23 +++++-
 gdb/dwarf2/expr.c                    |   8 +--
 gdb/findvar.c                        |  35 +++++----
 gdb/frame.c                          |   6 +-
 gdb/regcache.c                       |  21 +++---
 gdb/rust-lang.c                      |  10 ++-
 gdb/unittests/array-view-selftests.c | 104 +++++++++++++++++++++++++++
 gdb/unittests/gmp-utils-selftests.c  |   2 +-
 gdb/valarith.c                       |  58 +++++++--------
 gdb/valops.c                         |  72 +++++++++----------
 gdb/value.c                          |  24 +++----
 gdbsupport/array-view.h              |  18 +++++
 13 files changed, 261 insertions(+), 137 deletions(-)

-- 
2.33.1

[PATCH v2 1/3] gdbsupport: add array_view copy function

[Simon Marchi]

From: Simon Marchi <simon.marchi@polymtl.ca>

An assertion was recently added to array_view::operator[] to ensure we
don't do out of bounds accesses.  However, when the array_view is copied
to or from using memcpy, it bypasses that safety.

To address this, add a `copy` free function that copies data from an
array view to another, ensuring that the destination and source array
views have the same size.  When copying to or from parts of an
array_view, we are expected to use gdb::array_view::slice, which does
its own bounds check.  With all that, any copy operation that goes out
of bounds should be caught by an assertion at runtime.

copy is implemented using std::copy (and std::copy_backward), which, at
least on libstdc++, appears to pick memmove when copying trivial data.
So in the end there shouldn't be much difference vs using a bare memcpy,
as we do right now.  When copying non-trivial data, std::copy simply
assigns each element in a loop, as expected.

To properly support overlapping ranges, we must use std::copy_backward
instead of std::copy if the destination array overlaps and starts after
the source array.

Change a few randomly selected spots to use the new function, to show
how it can be used.

Add a test for the new function, testing both with arrays of a trivial
type (int) and of a non-trivial type (foo).  Test non-overlapping
ranges as well as three kinds of overlapping ranges: source before dest,
dest before source, and dest == source.

Change-Id: Ibeaca04e0028410fd44ce82f72e60058d6230a03
---
 gdb/ada-lang.c                       |  17 ++---
 gdb/dwarf2/expr.c                    |   4 +-
 gdb/frame.c                          |   4 +-
 gdb/unittests/array-view-selftests.c | 104 +++++++++++++++++++++++++++
 gdb/valarith.c                       |  49 +++++++------
 gdb/valops.c                         |  44 +++++++-----
 gdb/value.c                          |  22 +++---
 gdbsupport/array-view.h              |  18 +++++
 8 files changed, 195 insertions(+), 67 deletions(-)

diff --git a/gdb/ada-lang.c b/gdb/ada-lang.c
index b84e10fd029..fa30dc9404d 100644
--- a/gdb/ada-lang.c
+++ b/gdb/ada-lang.c
@@ -2586,9 +2586,7 @@ ada_value_assign (struct value *toval, struct value *fromval)
       write_memory_with_notification (to_addr, buffer, len);
 
       val = value_copy (toval);
-      memcpy (value_contents_raw (val).data (),
-	      value_contents (fromval).data (),
-	      TYPE_LENGTH (type));
+      copy (value_contents_raw (val), value_contents (fromval));
       deprecated_set_value_type (val, type);
 
       return val;
@@ -4184,9 +4182,7 @@ ada_convert_actual (struct value *actual, struct type *formal_type0)
 
 	      actual_type = ada_check_typedef (value_type (actual));
 	      val = allocate_value (actual_type);
-	      memcpy ((char *) value_contents_raw (val).data (),
-		      (char *) value_contents (actual).data (),
-		      TYPE_LENGTH (actual_type));
+	      copy (value_contents_raw (val), value_contents (actual));
 	      actual = ensure_lval (val);
 	    }
 	  result = value_addr (actual);
@@ -8898,7 +8894,6 @@ ada_promote_array_of_integrals (struct type *type, struct value *val)
 {
   struct type *elt_type = TYPE_TARGET_TYPE (type);
   LONGEST lo, hi;
-  struct value *res;
   LONGEST i;
 
   /* Verify that both val and type are arrays of scalars, and
@@ -8914,16 +8909,16 @@ ada_promote_array_of_integrals (struct type *type, struct value *val)
   if (!get_array_bounds (type, &lo, &hi))
     error (_("unable to determine array bounds"));
 
-  res = allocate_value (type);
+  value *res = allocate_value (type);
+  gdb::array_view<gdb_byte> res_contents = value_contents_writeable (res);
 
   /* Promote each array element.  */
   for (i = 0; i < hi - lo + 1; i++)
     {
       struct value *elt = value_cast (elt_type, value_subscript (val, lo + i));
+      int elt_len = TYPE_LENGTH (elt_type);
 
-      memcpy ((value_contents_writeable (res).data ()
-	       + (i * TYPE_LENGTH (elt_type))),
-	      value_contents_all (elt).data (), TYPE_LENGTH (elt_type));
+      copy (res_contents.slice (elt_len * i, elt_len), value_contents_all (elt));
     }
 
   return res;
diff --git a/gdb/dwarf2/expr.c b/gdb/dwarf2/expr.c
index 652161955d5..e00a620406f 100644
--- a/gdb/dwarf2/expr.c
+++ b/gdb/dwarf2/expr.c
@@ -1037,8 +1037,8 @@ dwarf_expr_context::fetch_result (struct type *type, struct type *subobj_type,
 	    if (gdbarch_byte_order (arch) == BFD_ENDIAN_BIG)
 	      subobj_offset += n - max;
 
-	    memcpy (value_contents_raw (retval).data (),
-		    value_contents_all (val).data () + subobj_offset, len);
+	    copy (value_contents_raw (retval),
+		  value_contents_all (val).slice (subobj_offset, len));
 	  }
 	  break;
 
diff --git a/gdb/frame.c b/gdb/frame.c
index 2a899fc494f..70b1247d15e 100644
--- a/gdb/frame.c
+++ b/gdb/frame.c
@@ -1362,9 +1362,9 @@ read_frame_register_unsigned (frame_info *frame, int regnum,
     {
       struct gdbarch *gdbarch = get_frame_arch (frame);
       enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);
-      int size = register_size (gdbarch, VALUE_REGNUM (regval));
 
-      *val = extract_unsigned_integer (value_contents (regval).data (), size,
+      gdb::array_view<const gdb_byte> contents = value_contents (regval);
+      *val = extract_unsigned_integer (contents.data (), contents.size (),
 				       byte_order);
       return true;
     }
diff --git a/gdb/unittests/array-view-selftests.c b/gdb/unittests/array-view-selftests.c
index 9df48db3912..a67be0534ac 100644
--- a/gdb/unittests/array-view-selftests.c
+++ b/gdb/unittests/array-view-selftests.c
@@ -546,6 +546,108 @@ run_tests ()
   }
 }
 
+template <typename T>
+void
+run_copy_test ()
+{
+  /* Test non-overlapping copy.  */
+  {
+    const std::vector<T> src_v = {1, 2, 3, 4};
+    std::vector<T> dest_v (4, -1);
+
+    SELF_CHECK (dest_v != src_v);
+    copy (gdb::array_view<T> (dest_v), gdb::array_view<const T> (src_v));
+    SELF_CHECK (dest_v == src_v);
+  }
+
+  /* Test overlapping copy, where the source is before the destination.  */
+  {
+    std::vector<T> vec = {1, 2, 3, 4, 5, 6, 7, 8};
+    gdb::array_view<T> v = vec;
+
+    copy (v.slice (2, 4),
+	  v.slice (1, 4));
+
+    std::vector<T> expected = {1, 2, 2, 3, 4, 5, 7, 8};
+    SELF_CHECK (vec == expected);
+  }
+
+  /* Test overlapping copy, where the source is after the destination.  */
+  {
+    std::vector<T> vec = {1, 2, 3, 4, 5, 6, 7, 8};
+    gdb::array_view<T> v = vec;
+
+    copy (v.slice (1, 4),
+	  v.slice (2, 4));
+
+    std::vector<T> expected = {1, 3, 4, 5, 6, 6, 7, 8};
+    SELF_CHECK (vec == expected);
+  }
+
+  /* Test overlapping copy, where the source is the same as the destination.  */
+  {
+    std::vector<T> vec = {1, 2, 3, 4, 5, 6, 7, 8};
+    gdb::array_view<T> v = vec;
+
+    copy (v.slice (2, 4),
+	  v.slice (2, 4));
+
+    std::vector<T> expected = {1, 2, 3, 4, 5, 6, 7, 8};
+    SELF_CHECK (vec == expected);
+  }
+}
+
+/* Class with a non-trivial copy assignment operator, used to test the
+   array_view copy function.  */
+struct foo
+{
+  /* Can be implicitly constructed from an int, such that we can use the same
+     templated test function to test against array_view<int> and
+     array_view<foo>.  */
+  foo (int n)
+    : n (n)
+  {}
+
+  /* Needed to avoid -Wdeprecated-copy-with-user-provided-copy error with
+     Clang.  */
+  foo (const foo &other) = default;
+
+  void operator= (const foo &other)
+  {
+    this->n = other.n;
+    this->n_assign_op_called++;
+  }
+
+  bool operator==(const foo &other) const
+  {
+    return this->n == other.n;
+  }
+
+  int n;
+
+  /* Number of times the assignment operator has been called.  */
+  static int n_assign_op_called;
+};
+
+int foo::n_assign_op_called = 0;
+
+/* Test the array_view copy free function.  */
+
+static void
+run_copy_tests ()
+{
+  /* Test with a trivial type.  */
+  run_copy_test<int> ();
+
+  /* Test with a non-trivial type.  */
+  foo::n_assign_op_called = 0;
+  run_copy_test<foo> ();
+
+  /* Make sure that for the non-trivial type foo, the assignment operator was
+     called an amount of times that makes sense.  */
+  SELF_CHECK (foo::n_assign_op_called == 16);
+}
+
 } /* namespace array_view_tests */
 } /* namespace selftests */
 
@@ -555,4 +657,6 @@ _initialize_array_view_selftests ()
 {
   selftests::register_test ("array_view",
 			    selftests::array_view_tests::run_tests);
+  selftests::register_test ("array_view-copy",
+			    selftests::array_view_tests::run_copy_tests);
 }
diff --git a/gdb/valarith.c b/gdb/valarith.c
index 140ef448137..ebf394df437 100644
--- a/gdb/valarith.c
+++ b/gdb/valarith.c
@@ -1531,7 +1531,7 @@ value_vector_widen (struct value *scalar_value, struct type *vector_type)
 {
   /* Widen the scalar to a vector.  */
   struct type *eltype, *scalar_type;
-  struct value *val, *elval;
+  struct value *elval;
   LONGEST low_bound, high_bound;
   int i;
 
@@ -1554,11 +1554,14 @@ value_vector_widen (struct value *scalar_value, struct type *vector_type)
       && !value_equal (elval, scalar_value))
     error (_("conversion of scalar to vector involves truncation"));
 
-  val = allocate_value (vector_type);
+  value *val = allocate_value (vector_type);
+  gdb::array_view<gdb_byte> val_contents = value_contents_writeable (val);
+  int elt_len = TYPE_LENGTH (eltype);
+
   for (i = 0; i < high_bound - low_bound + 1; i++)
     /* Duplicate the contents of elval into the destination vector.  */
-    memcpy (value_contents_writeable (val).data () + (i * TYPE_LENGTH (eltype)),
-	    value_contents_all (elval).data (), TYPE_LENGTH (eltype));
+    copy (val_contents.slice (i * elt_len, elt_len),
+	  value_contents_all (elval));
 
   return val;
 }
@@ -1569,7 +1572,6 @@ value_vector_widen (struct value *scalar_value, struct type *vector_type)
 static struct value *
 vector_binop (struct value *val1, struct value *val2, enum exp_opcode op)
 {
-  struct value *val, *tmp, *mark;
   struct type *type1, *type2, *eltype1, *eltype2;
   int t1_is_vec, t2_is_vec, elsize, i;
   LONGEST low_bound1, high_bound1, low_bound2, high_bound2;
@@ -1599,15 +1601,15 @@ vector_binop (struct value *val1, struct value *val2, enum exp_opcode op)
       || low_bound1 != low_bound2 || high_bound1 != high_bound2)
     error (_("Cannot perform operation on vectors with different types"));
 
-  val = allocate_value (type1);
-  mark = value_mark ();
+  value *val = allocate_value (type1);
+  gdb::array_view<gdb_byte> val_contents = value_contents_writeable (val);
+  value *mark = value_mark ();
   for (i = 0; i < high_bound1 - low_bound1 + 1; i++)
     {
-      tmp = value_binop (value_subscript (val1, i),
-			 value_subscript (val2, i), op);
-      memcpy (value_contents_writeable (val).data () + i * elsize,
-	      value_contents_all (tmp).data (),
-	      elsize);
+      value *tmp = value_binop (value_subscript (val1, i),
+				value_subscript (val2, i), op);
+      copy (val_contents.slice (i * elsize, elsize),
+	    value_contents_all (tmp));
      }
   value_free_to_mark (mark);
 
@@ -1888,7 +1890,7 @@ value_neg (struct value *arg1)
     return value_binop (value_zero (type, not_lval), arg1, BINOP_SUB);
   else if (type->code () == TYPE_CODE_ARRAY && type->is_vector ())
     {
-      struct value *tmp, *val = allocate_value (type);
+      struct value *val = allocate_value (type);
       struct type *eltype = check_typedef (TYPE_TARGET_TYPE (type));
       int i;
       LONGEST low_bound, high_bound;
@@ -1896,12 +1898,14 @@ value_neg (struct value *arg1)
       if (!get_array_bounds (type, &low_bound, &high_bound))
 	error (_("Could not determine the vector bounds"));
 
+      gdb::array_view<gdb_byte> val_contents = value_contents_writeable (val);
+      int elt_len = TYPE_LENGTH (eltype);
+
       for (i = 0; i < high_bound - low_bound + 1; i++)
 	{
-	  tmp = value_neg (value_subscript (arg1, i));
-	  memcpy ((value_contents_writeable (val).data ()
-		   + i * TYPE_LENGTH (eltype)),
-		  value_contents_all (tmp).data (), TYPE_LENGTH (eltype));
+	  value *tmp = value_neg (value_subscript (arg1, i));
+	  copy (val_contents.slice (i * elt_len, elt_len),
+		value_contents_all (tmp));
 	}
       return val;
     }
@@ -1931,7 +1935,6 @@ value_complement (struct value *arg1)
     val = value_from_longest (type, ~value_as_long (arg1));
   else if (type->code () == TYPE_CODE_ARRAY && type->is_vector ())
     {
-      struct value *tmp;
       struct type *eltype = check_typedef (TYPE_TARGET_TYPE (type));
       int i;
       LONGEST low_bound, high_bound;
@@ -1940,12 +1943,14 @@ value_complement (struct value *arg1)
 	error (_("Could not determine the vector bounds"));
 
       val = allocate_value (type);
+      gdb::array_view<gdb_byte> val_contents = value_contents_writeable (val);
+      int elt_len = TYPE_LENGTH (eltype);
+
       for (i = 0; i < high_bound - low_bound + 1; i++)
 	{
-	  tmp = value_complement (value_subscript (arg1, i));
-	  memcpy ((value_contents_writeable (val).data ()
-		   + i * TYPE_LENGTH (eltype)),
-		  value_contents_all (tmp).data (), TYPE_LENGTH (eltype));
+	  value *tmp = value_complement (value_subscript (arg1, i));
+	  copy (val_contents.slice (i * elt_len, elt_len),
+		value_contents_all (tmp));
 	}
     }
   else if (type->code () == TYPE_CODE_COMPLEX)
diff --git a/gdb/valops.c b/gdb/valops.c
index 9787cdbb513..106e12d3234 100644
--- a/gdb/valops.c
+++ b/gdb/valops.c
@@ -586,9 +586,12 @@ value_cast (struct type *type, struct value *arg2)
 	 sees a cast as a simple reinterpretation of the pointer's
 	 bits.  */
       if (code2 == TYPE_CODE_PTR)
-	longest = extract_unsigned_integer
-		    (value_contents (arg2).data (), TYPE_LENGTH (type2),
-		     type_byte_order (type2));
+	{
+	  gdb::array_view<const gdb_byte> contents = value_contents (arg2);
+	  longest = extract_unsigned_integer (contents.data (),
+					      contents.size (),
+					      type_byte_order (type2));
+	}
       else
 	longest = value_as_long (arg2);
       return value_from_longest (to_type, convert_to_boolean ?
@@ -954,18 +957,19 @@ value_one (struct type *type)
       struct type *eltype = check_typedef (TYPE_TARGET_TYPE (type1));
       int i;
       LONGEST low_bound, high_bound;
-      struct value *tmp;
 
       if (!get_array_bounds (type1, &low_bound, &high_bound))
 	error (_("Could not determine the vector bounds"));
 
       val = allocate_value (type);
+      gdb::array_view<gdb_byte> val_contents = value_contents_writeable (val);
+      int elt_len = TYPE_LENGTH (eltype);
+
       for (i = 0; i < high_bound - low_bound + 1; i++)
 	{
-	  tmp = value_one (eltype);
-	  memcpy ((value_contents_writeable (val).data ()
-		   + i * TYPE_LENGTH (eltype)),
-		  value_contents_all (tmp).data (), TYPE_LENGTH (eltype));
+	  value *tmp = value_one (eltype);
+	  copy (val_contents.slice (i * elt_len, elt_len),
+		value_contents_all (tmp));
 	}
     }
   else
@@ -1342,8 +1346,7 @@ value_assign (struct value *toval, struct value *fromval)
      implies the returned value is not lazy, even if TOVAL was.  */
   val = value_copy (toval);
   set_value_lazy (val, 0);
-  memcpy (value_contents_raw (val).data (), value_contents (fromval).data (),
-	  TYPE_LENGTH (type));
+  copy (value_contents_raw (val), value_contents (fromval));
 
   /* We copy over the enclosing type and pointed-to offset from FROMVAL
      in the case of pointer types.  For object types, the enclosing type
@@ -4030,10 +4033,13 @@ value_literal_complex (struct value *arg1,
   arg1 = value_cast (real_type, arg1);
   arg2 = value_cast (real_type, arg2);
 
-  memcpy (value_contents_raw (val).data (),
-	  value_contents (arg1).data (), TYPE_LENGTH (real_type));
-  memcpy (value_contents_raw (val).data () + TYPE_LENGTH (real_type),
-	  value_contents (arg2).data (), TYPE_LENGTH (real_type));
+  int len = TYPE_LENGTH (real_type);
+
+  copy (value_contents_raw (val).slice (0, len),
+	value_contents (arg1));
+  copy (value_contents_raw (val).slice (len, len),
+	value_contents (arg2));
+
   return val;
 }
 
@@ -4074,12 +4080,12 @@ cast_into_complex (struct type *type, struct value *val)
       struct type *val_real_type = TYPE_TARGET_TYPE (value_type (val));
       struct value *re_val = allocate_value (val_real_type);
       struct value *im_val = allocate_value (val_real_type);
+      int len = TYPE_LENGTH (val_real_type);
 
-      memcpy (value_contents_raw (re_val).data (),
-	      value_contents (val).data (), TYPE_LENGTH (val_real_type));
-      memcpy (value_contents_raw (im_val).data (),
-	      value_contents (val).data () + TYPE_LENGTH (val_real_type),
-	      TYPE_LENGTH (val_real_type));
+      copy (value_contents_raw (re_val),
+	    value_contents (val).slice (0, len));
+      copy (value_contents_raw (im_val),
+	    value_contents (val).slice (len, len));
 
       return value_literal_complex (re_val, im_val, type);
     }
diff --git a/gdb/value.c b/gdb/value.c
index 7649b029f91..51c9e02cc27 100644
--- a/gdb/value.c
+++ b/gdb/value.c
@@ -1344,9 +1344,13 @@ value_contents_copy_raw (struct value *dst, LONGEST dst_offset,
 					     TARGET_CHAR_BIT * length));
 
   /* Copy the data.  */
-  memcpy (value_contents_all_raw (dst).data () + dst_offset * unit_size,
-	  value_contents_all_raw (src).data () + src_offset * unit_size,
-	  length * unit_size);
+  gdb::array_view<gdb_byte> dst_contents
+    = value_contents_all_raw (dst).slice (dst_offset * unit_size,
+					  length * unit_size);
+  gdb::array_view<const gdb_byte> src_contents
+    = value_contents_all_raw (src).slice (src_offset * unit_size,
+					  length * unit_size);
+  copy (dst_contents, src_contents);
 
   /* Copy the meta-data, adjusted.  */
   src_bit_offset = src_offset * unit_size * HOST_CHAR_BIT;
@@ -1721,13 +1725,11 @@ value_copy (struct value *arg)
   val->stack = arg->stack;
   val->is_zero = arg->is_zero;
   val->initialized = arg->initialized;
+
   if (!value_lazy (val))
-    {
-      memcpy (value_contents_all_raw (val).data (),
-	      value_contents_all_raw (arg).data (),
-	      TYPE_LENGTH (value_enclosing_type (arg)));
+    copy (value_contents_all_raw (val),
+	  value_contents_all_raw (arg));
 
-    }
   val->unavailable = arg->unavailable;
   val->optimized_out = arg->optimized_out;
   val->parent = arg->parent;
@@ -1772,9 +1774,7 @@ value_non_lval (struct value *arg)
       struct type *enc_type = value_enclosing_type (arg);
       struct value *val = allocate_value (enc_type);
 
-      memcpy (value_contents_all_raw (val).data (),
-	      value_contents_all (arg).data (),
-	      TYPE_LENGTH (enc_type));
+      copy (value_contents_all_raw (val), value_contents_all (arg));
       val->type = arg->type;
       set_value_embedded_offset (val, value_embedded_offset (arg));
       set_value_pointed_to_offset (val, value_pointed_to_offset (arg));
diff --git a/gdbsupport/array-view.h b/gdbsupport/array-view.h
index edf66559e2d..8dd53b0b232 100644
--- a/gdbsupport/array-view.h
+++ b/gdbsupport/array-view.h
@@ -19,6 +19,7 @@
 #define COMMON_ARRAY_VIEW_H
 
 #include "traits.h"
+#include <algorithm>
 #include <type_traits>
 
 /* An array_view is an abstraction that provides a non-owning view
@@ -206,6 +207,23 @@ class array_view
   size_type m_size;
 };
 
+/* Copy the contents referenced by the array view SRC to the array view DEST.
+
+   The two array views must have the same length and element size.
+
+   The copy is done using memcpy, so both element types must be trivially
+   copyable (this is checked by poison.h).  */
+
+template <typename T, typename U>
+void copy (gdb::array_view<T> dest, gdb::array_view<U> src)
+{
+  gdb_assert (dest.size () == src.size ());
+  if (dest.data () <= src.data ())
+    std::copy(src.begin (), src.end (), dest.begin ());
+  else
+    std::copy_backward (src.begin (), src.end (), dest.end ());
+}
+
 /* Compare LHS and RHS for (deep) equality.  That is, whether LHS and
    RHS have the same sizes, and whether each pair of elements of LHS
    and RHS at the same position compares equal.  */
-- 
2.33.1

[PATCH v2 2/3] gdb: make extract_integer take an array_view

[Simon Marchi]

From: Simon Marchi <simon.marchi@polymtl.ca>

I think it would make sense for extract_integer, extract_signed_integer
and extract_unsigned_integer to take an array_view.  This way, when we
extract an integer, we can validate that we don't overflow the buffer
passed by the caller (e.g. ask to extract a 4-byte integer but pass a
2-byte buffer).

 - Change extract_integer to take an array_view
 - Add overloads of extract_signed_integer and extract_unsigned_integer
   that take array_views.  Keep the existing versions so we don't
   need to change all callers, but make them call the array_view
   versions.

This shortens some places like:

  result = extract_unsigned_integer (value_contents (result_val).data (),
				     TYPE_LENGTH (value_type (result_val)),
				     byte_order);

into

  result = extract_unsigned_integer (value_contents (result_val), byte_order);

value_contents returns an array view that is of length
`TYPE_LENGTH (value_type (result_val))` already, so the length is
implicitly communicated through the array view.

Change-Id: Ic1c1f98c88d5c17a8486393af316f982604d6c95
---
 gdb/defs.h                          | 23 ++++++++++++++++---
 gdb/dwarf2/expr.c                   |  4 +---
 gdb/findvar.c                       | 35 ++++++++++++++---------------
 gdb/regcache.c                      | 21 +++++++----------
 gdb/unittests/gmp-utils-selftests.c |  2 +-
 5 files changed, 47 insertions(+), 38 deletions(-)

diff --git a/gdb/defs.h b/gdb/defs.h
index f7e09eca9db..3b6a0e63905 100644
--- a/gdb/defs.h
+++ b/gdb/defs.h
@@ -63,6 +63,7 @@
 
 #include "gdbsupport/host-defs.h"
 #include "gdbsupport/enum-flags.h"
+#include "gdbsupport/array-view.h"
 
 /* Scope types enumerator.  List the types of scopes the compiler will
    accept.  */
@@ -500,20 +501,36 @@ enum symbol_needs_kind
 /* In findvar.c.  */
 
 template<typename T, typename = RequireLongest<T>>
-T extract_integer (const gdb_byte *addr, int len, enum bfd_endian byte_order);
+T extract_integer (gdb::array_view<const gdb_byte>, enum bfd_endian byte_order);
+
+static inline LONGEST
+extract_signed_integer (gdb::array_view<const gdb_byte> buf,
+			enum bfd_endian byte_order)
+{
+  return extract_integer<LONGEST> (buf, byte_order);
+}
 
 static inline LONGEST
 extract_signed_integer (const gdb_byte *addr, int len,
 			enum bfd_endian byte_order)
 {
-  return extract_integer<LONGEST> (addr, len, byte_order);
+  return extract_signed_integer (gdb::array_view<const gdb_byte> (addr, len),
+				 byte_order);
+}
+
+static inline ULONGEST
+extract_unsigned_integer (gdb::array_view<const gdb_byte> buf,
+			  enum bfd_endian byte_order)
+{
+  return extract_integer<ULONGEST> (buf, byte_order);
 }
 
 static inline ULONGEST
 extract_unsigned_integer (const gdb_byte *addr, int len,
 			  enum bfd_endian byte_order)
 {
-  return extract_integer<ULONGEST> (addr, len, byte_order);
+  return extract_unsigned_integer (gdb::array_view<const gdb_byte> (addr, len),
+				   byte_order);
 }
 
 extern int extract_long_unsigned_integer (const gdb_byte *, int,
diff --git a/gdb/dwarf2/expr.c b/gdb/dwarf2/expr.c
index e00a620406f..e308a50d5f5 100644
--- a/gdb/dwarf2/expr.c
+++ b/gdb/dwarf2/expr.c
@@ -1157,9 +1157,7 @@ dwarf_expr_context::fetch_address (int n)
   ULONGEST result;
 
   dwarf_require_integral (value_type (result_val));
-  result = extract_unsigned_integer (value_contents (result_val).data (),
-				     TYPE_LENGTH (value_type (result_val)),
-				     byte_order);
+  result = extract_unsigned_integer (value_contents (result_val), byte_order);
 
   /* For most architectures, calling extract_unsigned_integer() alone
      is sufficient for extracting an address.  However, some
diff --git a/gdb/findvar.c b/gdb/findvar.c
index f7e632809d0..a0031d2dadd 100644
--- a/gdb/findvar.c
+++ b/gdb/findvar.c
@@ -48,14 +48,11 @@ you lose
 
 template<typename T, typename>
 T
-extract_integer (const gdb_byte *addr, int len, enum bfd_endian byte_order)
+extract_integer (gdb::array_view<const gdb_byte> buf, enum bfd_endian byte_order)
 {
   typename std::make_unsigned<T>::type retval = 0;
-  const unsigned char *p;
-  const unsigned char *startaddr = addr;
-  const unsigned char *endaddr = startaddr + len;
 
-  if (len > (int) sizeof (T))
+  if (buf.size () > (int) sizeof (T))
     error (_("\
 That operation is not available on integers of more than %d bytes."),
 	   (int) sizeof (T));
@@ -64,36 +61,38 @@ That operation is not available on integers of more than %d bytes."),
      the least significant.  */
   if (byte_order == BFD_ENDIAN_BIG)
     {
-      p = startaddr;
+      size_t i = 0;
+
       if (std::is_signed<T>::value)
 	{
 	  /* Do the sign extension once at the start.  */
-	  retval = ((LONGEST) * p ^ 0x80) - 0x80;
-	  ++p;
+	  retval = ((LONGEST) buf[i] ^ 0x80) - 0x80;
+	  ++i;
 	}
-      for (; p < endaddr; ++p)
-	retval = (retval << 8) | *p;
+      for (; i < buf.size (); ++i)
+	retval = (retval << 8) | buf[i];
     }
   else
     {
-      p = endaddr - 1;
+      ssize_t i = buf.size () - 1;
+
       if (std::is_signed<T>::value)
 	{
 	  /* Do the sign extension once at the start.  */
-	  retval = ((LONGEST) * p ^ 0x80) - 0x80;
-	  --p;
+	  retval = ((LONGEST) buf[i] ^ 0x80) - 0x80;
+	  --i;
 	}
-      for (; p >= startaddr; --p)
-	retval = (retval << 8) | *p;
+      for (; i >= 0; --i)
+	retval = (retval << 8) | buf[i];
     }
   return retval;
 }
 
 /* Explicit instantiations.  */
-template LONGEST extract_integer<LONGEST> (const gdb_byte *addr, int len,
+template LONGEST extract_integer<LONGEST> (gdb::array_view<const gdb_byte> buf,
 					   enum bfd_endian byte_order);
-template ULONGEST extract_integer<ULONGEST> (const gdb_byte *addr, int len,
-					     enum bfd_endian byte_order);
+template ULONGEST extract_integer<ULONGEST>
+  (gdb::array_view<const gdb_byte> buf, enum bfd_endian byte_order);
 
 /* Sometimes a long long unsigned integer can be extracted as a
    LONGEST value.  This is done so that we can print these values
diff --git a/gdb/regcache.c b/gdb/regcache.c
index 8457284c12a..b3ecba2fbe6 100644
--- a/gdb/regcache.c
+++ b/gdb/regcache.c
@@ -620,15 +620,12 @@ template<typename T, typename>
 enum register_status
 readable_regcache::raw_read (int regnum, T *val)
 {
-  gdb_byte *buf;
-  enum register_status status;
-
   assert_regnum (regnum);
-  buf = (gdb_byte *) alloca (m_descr->sizeof_register[regnum]);
-  status = raw_read (regnum, buf);
+  size_t len = m_descr->sizeof_register[regnum];
+  gdb_byte *buf = (gdb_byte *) alloca (len);
+  register_status status = raw_read (regnum, buf);
   if (status == REG_VALID)
-    *val = extract_integer<T> (buf,
-			       m_descr->sizeof_register[regnum],
+    *val = extract_integer<T> ({buf, len},
 			       gdbarch_byte_order (m_descr->gdbarch));
   else
     *val = 0;
@@ -772,14 +769,12 @@ template<typename T, typename>
 enum register_status
 readable_regcache::cooked_read (int regnum, T *val)
 {
-  enum register_status status;
-  gdb_byte *buf;
-
   gdb_assert (regnum >= 0 && regnum < m_descr->nr_cooked_registers);
-  buf = (gdb_byte *) alloca (m_descr->sizeof_register[regnum]);
-  status = cooked_read (regnum, buf);
+  size_t len = m_descr->sizeof_register[regnum];
+  gdb_byte *buf = (gdb_byte *) alloca (len);
+  register_status status = cooked_read (regnum, buf);
   if (status == REG_VALID)
-    *val = extract_integer<T> (buf, m_descr->sizeof_register[regnum],
+    *val = extract_integer<T> ({buf, len},
 			       gdbarch_byte_order (m_descr->gdbarch));
   else
     *val = 0;
diff --git a/gdb/unittests/gmp-utils-selftests.c b/gdb/unittests/gmp-utils-selftests.c
index f40666e4cd5..e48bb39166e 100644
--- a/gdb/unittests/gmp-utils-selftests.c
+++ b/gdb/unittests/gmp-utils-selftests.c
@@ -299,7 +299,7 @@ write_and_extract (T val, size_t buf_len, enum bfd_endian byte_order)
   gdb_byte *buf = (gdb_byte *) alloca (buf_len);
   v.write ({buf, buf_len}, byte_order, !std::is_signed<T>::value);
 
-  return extract_integer<T> (buf, buf_len, byte_order);
+  return extract_integer<T> ({buf, buf_len}, byte_order);
 }
 
 /* Test the gdb_mpz::write method over a reasonable range of values.
-- 
2.33.1

[PATCH v2 3/3] gdb: trivial changes to use array_view

[Simon Marchi]

From: Simon Marchi <simon.marchi@polymtl.ca>

Change a few relatively obvious spots using value contents to propagate
the use array_view a bit more.

Change-Id: I5338a60986f06d5969fec803d04f8423c9288a15
---
 gdb/frame.c     |  2 +-
 gdb/rust-lang.c | 10 ++++------
 gdb/valarith.c  |  9 +++------
 gdb/valops.c    | 28 ++++++++++------------------
 gdb/value.c     |  2 +-
 5 files changed, 19 insertions(+), 32 deletions(-)

diff --git a/gdb/frame.c b/gdb/frame.c
index 70b1247d15e..268397259de 100644
--- a/gdb/frame.c
+++ b/gdb/frame.c
@@ -1261,7 +1261,7 @@ frame_unwind_register_value (frame_info *next_frame, int regnum)
 	  else
 	    {
 	      int i;
-	      const gdb_byte *buf = value_contents (value).data ();
+	      gdb::array_view<const gdb_byte> buf = value_contents (value);
 
 	      fprintf_unfiltered (&debug_file, " bytes=");
 	      fprintf_unfiltered (&debug_file, "[");
diff --git a/gdb/rust-lang.c b/gdb/rust-lang.c
index e5a404187be..9ab8fbe30c0 100644
--- a/gdb/rust-lang.c
+++ b/gdb/rust-lang.c
@@ -1317,9 +1317,8 @@ eval_op_rust_struct_anon (struct type *expect_type, struct expression *exp,
 
       if (rust_enum_p (type))
 	{
-	  gdb::array_view<const gdb_byte> view (value_contents (lhs).data (),
-						TYPE_LENGTH (type));
-	  type = resolve_dynamic_type (type, view, value_address (lhs));
+	  type = resolve_dynamic_type (type, value_contents (lhs),
+				       value_address (lhs));
 
 	  if (rust_empty_enum_p (type))
 	    error (_("Cannot access field %d of empty enum %s"),
@@ -1380,9 +1379,8 @@ eval_op_rust_structop (struct type *expect_type, struct expression *exp,
   struct type *type = value_type (lhs);
   if (type->code () == TYPE_CODE_STRUCT && rust_enum_p (type))
     {
-      gdb::array_view<const gdb_byte> view (value_contents (lhs).data (),
-					    TYPE_LENGTH (type));
-      type = resolve_dynamic_type (type, view, value_address (lhs));
+      type = resolve_dynamic_type (type, value_contents (lhs),
+				   value_address (lhs));
 
       if (rust_empty_enum_p (type))
 	error (_("Cannot access field %s of empty enum %s"),
diff --git a/gdb/valarith.c b/gdb/valarith.c
index ebf394df437..ac995a5cbb0 100644
--- a/gdb/valarith.c
+++ b/gdb/valarith.c
@@ -927,12 +927,10 @@ fixed_point_binop (struct value *arg1, struct value *arg2, enum exp_opcode op)
 	  type2 = type1;
 	}
 
-      v1.read_fixed_point (gdb::make_array_view (value_contents (arg1).data (),
-						 TYPE_LENGTH (type1)),
+      v1.read_fixed_point (value_contents (arg1),
 			   type_byte_order (type1), type1->is_unsigned (),
 			   type1->fixed_point_scaling_factor ());
-      v2.read_fixed_point (gdb::make_array_view (value_contents (arg2).data (),
-						 TYPE_LENGTH (type2)),
+      v2.read_fixed_point (value_contents (arg2),
 			   type_byte_order (type2), type2->is_unsigned (),
 			   type2->fixed_point_scaling_factor ());
     }
@@ -942,8 +940,7 @@ fixed_point_binop (struct value *arg1, struct value *arg2, enum exp_opcode op)
       value *fp_val = allocate_value (type1);
 
       fp.write_fixed_point
-	(gdb::make_array_view (value_contents_raw (fp_val).data (),
-			       TYPE_LENGTH (type1)),
+	(value_contents_raw (fp_val),
 	 type_byte_order (type1),
 	 type1->is_unsigned (),
 	 type1->fixed_point_scaling_factor ());
diff --git a/gdb/valops.c b/gdb/valops.c
index 106e12d3234..86a1da11ec3 100644
--- a/gdb/valops.c
+++ b/gdb/valops.c
@@ -351,9 +351,8 @@ value_to_gdb_mpq (struct value *value)
 		  || is_fixed_point_type (type));
 
       gdb_mpz vz;
-      vz.read (gdb::make_array_view (value_contents (value).data (),
-				     TYPE_LENGTH (type)),
-	       type_byte_order (type), type->is_unsigned ());
+      vz.read (value_contents (value), type_byte_order (type),
+	       type->is_unsigned ());
       mpq_set_z (result.val, vz.val);
 
       if (is_fixed_point_type (type))
@@ -392,8 +391,7 @@ value_cast_to_fixed_point (struct type *to_type, struct value *from_val)
   /* Finally, create the result value, and pack the unscaled value
      in it.  */
   struct value *result = allocate_value (to_type);
-  unscaled.write (gdb::make_array_view (value_contents_raw (result).data (),
-					TYPE_LENGTH (to_type)),
+  unscaled.write (value_contents_raw (result),
 		  type_byte_order (to_type),
 		  to_type->is_unsigned ());
 
@@ -554,11 +552,10 @@ value_cast (struct type *type, struct value *arg2)
 	{
 	  gdb_mpq fp_val;
 
-	  fp_val.read_fixed_point
-	    (gdb::make_array_view (value_contents (arg2).data (),
-				   TYPE_LENGTH (type2)),
-	     type_byte_order (type2), type2->is_unsigned (),
-	     type2->fixed_point_scaling_factor ());
+	  fp_val.read_fixed_point (value_contents (arg2),
+				   type_byte_order (type2),
+				   type2->is_unsigned (),
+				   type2->fixed_point_scaling_factor ());
 
 	  struct value *v = allocate_value (to_type);
 	  target_float_from_host_double (value_contents_raw (v).data (),
@@ -1259,14 +1256,9 @@ value_assign (struct value *toval, struct value *fromval)
 					   value_contents (fromval).data ());
 	      }
 	    else
-	      {
-		gdb::array_view<const gdb_byte> contents
-		  = gdb::make_array_view (value_contents (fromval).data (),
-					  TYPE_LENGTH (type));
-		put_frame_register_bytes (frame, value_reg,
-					  value_offset (toval),
-					  contents);
-	      }
+	      put_frame_register_bytes (frame, value_reg,
+					value_offset (toval),
+					value_contents (fromval));
 	  }
 
 	gdb::observers::register_changed.notify (frame, value_reg);
diff --git a/gdb/value.c b/gdb/value.c
index 51c9e02cc27..fbe4a2a491d 100644
--- a/gdb/value.c
+++ b/gdb/value.c
@@ -4011,7 +4011,7 @@ value_fetch_lazy_register (struct value *val)
       else
 	{
 	  int i;
-	  const gdb_byte *buf = value_contents (new_val).data ();
+	  gdb::array_view<const gdb_byte> buf = value_contents (new_val);
 
 	  if (VALUE_LVAL (new_val) == lval_register)
 	    fprintf_unfiltered (&debug_file, " register=%d",
-- 
2.33.1

Re: [PATCH v2 1/3] gdbsupport: add array_view copy function

[Pedro Alves]

Hi!

On 2021-11-18 21:29, Simon Marchi via Gdb-patches wrote:
>  
> +/* Copy the contents referenced by the array view SRC to the array view DEST.
> +
> +   The two array views must have the same length and element size.
> +
> +   The copy is done using memcpy, so both element types must be trivially
> +   copyable (this is checked by poison.h).  */

This sentence about memcpy is stale.

> +
> +template <typename T, typename U>
> +void copy (gdb::array_view<T> dest, gdb::array_view<U> src)
> +{
> +  gdb_assert (dest.size () == src.size ());
> +  if (dest.data () <= src.data ())
> +    std::copy(src.begin (), src.end (), dest.begin ());

Missing space before parens.

> +  else
> +    std::copy_backward (src.begin (), src.end (), dest.end ());
> +}

Three points below.

#1 - overlap triggering undefined behavior

If SRC and DEST are exactly the same, as in:

  /* Test overlapping copy, where the source is the same as the destination.  */
  {
    std::vector<T> vec = {1, 2, 3, 4, 5, 6, 7, 8};
    gdb::array_view<T> v = vec;

    copy (v.slice (2, 4),
	  v.slice (2, 4));

    std::vector<T> expected = {1, 2, 3, 4, 5, 6, 7, 8};
    SELF_CHECK (vec == expected);
  }

Isn't copy invoking undefined behavior?  In this case, we end up
in std::copy, which states, in 25.3.1 in the C++11 standard:

 "Requires: result shall not be in the range [first,last)."

In the case in question, result == first, so it is in range.


#2 - argument order

You don't mention this explicitly in the commit log, so I'll raise it:

It's sad that memcpy and std::copy have source/dest positions reversed.

Off-hand, I'd think it is less confusing if a function named "copy", same
name as the C++ standard function, follows the convention of std::copy.

#3 - slicing

The function as is allows slicing.  I.e., this compiles and passes runtime testing:

  {
    struct A { int i; };
    struct B : A { int j; };

    A a[2];
    B b[2];

    gdb::array_view<A> v1 = a;
    gdb::array_view<B> v2 = b;

    copy (v1, v2);
  }

Do we want that?  I'd think not off-hand.

Pedro Alves

Re: [PATCH v2 1/3] gdbsupport: add array_view copy function

[Simon Marchi]

On 2021-11-19 13:11, Pedro Alves wrote:
> Hi!
> 
> On 2021-11-18 21:29, Simon Marchi via Gdb-patches wrote:
>>  
>> +/* Copy the contents referenced by the array view SRC to the array view DEST.
>> +
>> +   The two array views must have the same length and element size.
>> +
>> +   The copy is done using memcpy, so both element types must be trivially
>> +   copyable (this is checked by poison.h).  */
> 
> This sentence about memcpy is stale.

Fixed.

>> +
>> +template <typename T, typename U>
>> +void copy (gdb::array_view<T> dest, gdb::array_view<U> src)
>> +{
>> +  gdb_assert (dest.size () == src.size ());
>> +  if (dest.data () <= src.data ())
>> +    std::copy(src.begin (), src.end (), dest.begin ());
> 
> Missing space before parens.

Fixed.

>> +  else
>> +    std::copy_backward (src.begin (), src.end (), dest.end ());
>> +}
> 
> Three points below.
> 
> #1 - overlap triggering undefined behavior
> 
> If SRC and DEST are exactly the same, as in:
> 
>   /* Test overlapping copy, where the source is the same as the destination.  */
>   {
>     std::vector<T> vec = {1, 2, 3, 4, 5, 6, 7, 8};
>     gdb::array_view<T> v = vec;
> 
>     copy (v.slice (2, 4),
> 	  v.slice (2, 4));
> 
>     std::vector<T> expected = {1, 2, 3, 4, 5, 6, 7, 8};
>     SELF_CHECK (vec == expected);
>   }
> 
> Isn't copy invoking undefined behavior?  In this case, we end up
> in std::copy, which states, in 25.3.1 in the C++11 standard:
> 
>  "Requires: result shall not be in the range [first,last)."
> 
> In the case in question, result == first, so it is in range.

Oh, right.  It seem like we can't use copy nor copy_backward if the
ranges are exactly overlapping.  Anyway, there's not much use to do the
copy if the source == the destination.  So I'll change it to this, would
that work?

  if (dest.data () < src.data ())
    std::copy (src.begin (), src.end (), dest.begin ());
  else if (dest.data () > src.data ())
    std::copy_backward (src.begin (), src.end (), dest.end ());

> #2 - argument order
> 
> You don't mention this explicitly in the commit log, so I'll raise it:
> 
> It's sad that memcpy and std::copy have source/dest positions reversed.
> 
> Off-hand, I'd think it is less confusing if a function named "copy", same
> name as the C++ standard function, follows the convention of std::copy.

Since I was mostly replacing some memcpys, I indeed mimicked memcpy.
But I'll change it to (src, dest).

> #3 - slicing
> 
> The function as is allows slicing.  I.e., this compiles and passes runtime testing:
> 
>   {
>     struct A { int i; };
>     struct B : A { int j; };
> 
>     A a[2];
>     B b[2];
> 
>     gdb::array_view<A> v1 = a;
>     gdb::array_view<B> v2 = b;
> 
>     copy (v1, v2);
>   }
> 
> Do we want that?  I'd think not off-hand.

I tried it, and we fall in the implementation of std::copy that does a
for loop over all items and justs uses the assignment operator (in other
words, the implementation for non-memcpyable types).  So I think the
behavior will be correct.

Simon