[RFA] internal-error using '@' (repeat) operator on array of dynamic objects

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

[RFA] internal-error using '@' (repeat) operator on array of dynamic objects

Joel Brobecker
Hello,

Using the following Ada declarations (the same as in
gdb.ada/dyn_stride.exp)...

      subtype Small_Type is Integer range L .. U;
      type Record_Type (I : Small_Type := L) is record
         S : String (1 .. I);
      end record;
      type Array_Type is array (Integer range <>) of Record_Type;

      A1 : Array_Type :=
        (1 => (I => U, S => (others => ASCII.NUL)),
         2 => (I => 1, S => "A"),
         3 => (I => 2, S => "AB"));

... where "L" and "U" are variables, trying to apply the repeat
operator to "A1(1)" yields to an internal error:

  | (gdb) print a1(1)@3
  | $5 = /[...]/gdbtypes.c:4883: internal-error: type* copy_type(const type*):
  | Assertion `TYPE_OBJFILE_OWNED (type)' failed.

What happens first is that the ada-lang module evaluated the "A1(1)"
sub-expression returning a structure where "I" (one of the fields
in that structure) has a type which is dynamic, because it is
a range type whose bounds are not statically known.

Next, we apply the repeat ('@') operator, which is done via
allocate_repeat_value, which creates an array type with the correct
bounds to associate to our value, by calling lookup_array_range_type:

  | struct type *
  | lookup_array_range_type (struct type *element_type,
  |                          LONGEST low_bound, LONGEST high_bound)
  | {
  |   struct gdbarch *gdbarch = get_type_arch (element_type);
  |   struct type *index_type = builtin_type (gdbarch)->builtin_int;
  |   struct type *range_type
  |     = create_static_range_type (NULL, index_type, low_bound, high_bound);
  |
  |   return create_array_type (NULL, element_type, range_type);
  | }

As we can see, this creates an array type whose index type is
always owned by the gdbarch. This is where the problem lies.

Next, we use that type to construct a struct value. That value
then gets passed to the valprint module, which then checks
whether our object is dynamic or not. And because field "I" above
had a dynamic range type, we end up determining by association
that the artificial repeat array itself is also dynamic. So
we attempt to resolve the type, which leads to trying to copying
that type. And because the artifical array created by
lookup_array_range_type has an index which is not objfile-owned,
we trip the assertion.

This patch fixes the issue by enhancing lookup_array_range_type
to create an index type which has the same owner as the element
type.

gdb/ChangeLog:

        * gdbtypes.c (lookup_array_range_type): Make sure the array's
        index type is objfile-owned if the element type is as well.

gdb/testsuite/ChangeLog:

        * testsuite/gdb.ada/dyn_stride.exp: Add "print a1(1)@3" test.

Tested on x86_64-linux. OK to push?

Thanks,
--
Joel

---
 gdb/gdbtypes.c                       | 13 +++++++++----
 gdb/testsuite/gdb.ada/dyn_stride.exp | 19 +++++++++++++++++++
 2 files changed, 28 insertions(+), 4 deletions(-)

diff --git a/gdb/gdbtypes.c b/gdb/gdbtypes.c
index 7ba62df..0a30223 100644
--- a/gdb/gdbtypes.c
+++ b/gdb/gdbtypes.c
@@ -1194,10 +1194,15 @@ struct type *
 lookup_array_range_type (struct type *element_type,
  LONGEST low_bound, LONGEST high_bound)
 {
-  struct gdbarch *gdbarch = get_type_arch (element_type);
-  struct type *index_type = builtin_type (gdbarch)->builtin_int;
-  struct type *range_type
-    = create_static_range_type (NULL, index_type, low_bound, high_bound);
+  struct type *index_type;
+  struct type *range_type;
+
+  if (TYPE_OBJFILE_OWNED (element_type))
+    index_type = objfile_type (TYPE_OWNER (element_type).objfile)->builtin_int;
+  else
+    index_type = builtin_type (get_type_arch (element_type))->builtin_int;
+  range_type = create_static_range_type (NULL, index_type,
+ low_bound, high_bound);
 
   return create_array_type (NULL, element_type, range_type);
 }
diff --git a/gdb/testsuite/gdb.ada/dyn_stride.exp b/gdb/testsuite/gdb.ada/dyn_stride.exp
index 0267ca1..dae5106 100644
--- a/gdb/testsuite/gdb.ada/dyn_stride.exp
+++ b/gdb/testsuite/gdb.ada/dyn_stride.exp
@@ -39,3 +39,22 @@ gdb_test "print A1(3)" \
 
 gdb_test "print A1(1..3)" \
          "\\(\\(i => 0, s => \"\"\\), \\(i => 1, s => \"A\"\\), \\(i => 2, s => \"AB\"\\)\\)"
+
+# Test the use of the "repeat" operator (@).
+#
+# Note that, in this case, the repeat operator makes little sense
+# and is NOT equivalent to requesting an array slice. In the case
+# of "print a1(1)@3", the size of each element in the array is
+# variable from element to element.  So, when asked to repeat
+# one element of the array a number of times, you're not guaranteed
+# to get the same answer as in a slice; while the slice will
+# take into account the array stride, the repeat operator uses
+# the size of the object being repeated as its stride, which
+# is often not the same. So, in the test below, the output for
+# the second and third element is not entirely predictable, because
+# it reads part of their contents from a memory region which has
+# an undefined value (the "holes" between the each actual element
+# of the array).
+
+gdb_test "print a1(1)@3" \
+         " = \\(\\(i => 0, s => \"\"\\), \\(i => -?$decimal, s => .*\\), \\(i => -?$decimal, s => .*\\)\\)"
--
2.1.4