13.11.2 Unchecked Storage Deallocation
[
Unchecked
storage deallocation of an object designated by a value of an access
type is achieved by a call to an instance of the generic procedure Unchecked_Deallocation.]
Static Semantics
The following language-defined
generic library procedure exists:
{
AI05-0229-1}
generic
type Object(<>)
is limited private;
type Name
is access Object;
procedure Ada.Unchecked_Deallocation(X :
in out Name)
with Convention => Intrinsic;
pragma Preelaborate(Ada.Unchecked_Deallocation);
Reason: {
AI05-0229-1}
The aspect Convention implies that the attribute Access is not allowed
for instances of Unchecked_Deallocation.
Legality Rules
{
AI05-0157-1}
A call on an instance of Unchecked_Deallocation is illegal if the actual
access type of the instance is a type for which the Storage_Size has
been specified by a static expression with value zero or is defined by
the language to be zero.
In addition to the places
where Legality Rules normally apply (see
12.3),
this rule applies also in the private part of an instance of a generic
unit.
Discussion: This rule is the same as
the rule for
allocators.
We could have left the last sentence out, as a call to Unchecked_Deallocation
cannot occur in a specification as it is a procedure call, but we left
it for consistency and to avoid future maintenance hazards.
Dynamic Semantics
Given an instance of
Unchecked_Deallocation declared as follows:
procedure Free is
new Ada.Unchecked_Deallocation(
object_subtype_name, access_to_variable_subtype_name);
Procedure Free has
the following effect:
1.
After executing Free(X), the value of X is null.
2.
Free(X), when X is already equal to null, has no effect.
3.
{
AI95-00416-01}
{
AI05-0107-1}
Free(X), when X is not equal to
null first performs finalization
of the object designated by X (and any coextensions of the object —
see
3.10.2), as described in
7.6.1.
It then deallocates the storage occupied by the object designated by
X (and any coextensions). If the storage pool is a user-defined object,
then the storage is deallocated by calling Deallocate as described in
13.11. There is one exception: if the object
being freed contains tasks, the object might not be deallocated.
Ramification: {
AI05-0107-1}
Free calls only the specified Deallocate procedure to do deallocation.
{
AI95-00416-01}
{
AI12-0148-1}
After
the finalization step of Free(X), the object designated by X, and any
subcomponents (and coextensions) thereof, no longer exist; their storage
can be reused for other purposes.
Bounded (Run-Time) Errors
It
is a bounded error to free a discriminated, unterminated task object.
The possible consequences are:
Reason: This is an error because the
task might refer to its discriminants, and the discriminants might be
deallocated by freeing the task object.
No exception is raised.
Program_Error
or Tasking_Error is raised at the point of the deallocation.
Program_Error
or Tasking_Error is raised in the task the next time it references any
of the discriminants.
Implementation Note: This last case presumes
an implementation where the task references its discriminants indirectly,
and the pointer is nulled out when the task object is deallocated.
In the first two cases, the storage for the discriminants
(and for any enclosing object if it is designated by an access discriminant
of the task) is not reclaimed prior to task termination.
Ramification: The storage might never
be reclaimed.
{
AI12-0148-1}
An access value that designates a nonexistent object is called a
dangling
reference.
Discussion: These can result from use
of Unchecked_Deallocation, Unchecked_Deallocate_Subpool, and attribute
Unchecked_Access. Bad results from Unchecked_Conversion and from stream-oriented
attributes are abnormal by
13.9.1, which
is stronger and thus takes precedence.
{
AI12-0148-1}
[If a dangling reference is dereferenced (implicitly or explicitly),
execution is erroneous (see below).] If there is no explicit or implicit
dereference, then it is a bounded error
to evaluate
an expression whose result is a dangling reference. If the error is detected,
either Constraint_Error or Program_Error is raised.
Otherwise,
execution proceeds normally, but with the possibility that the access
value designates some other existing object.
Reason: If a dangling reference is compared
with another access value, a result of either True or False is allowed.
We need to allow this so that simple implementations of access values
(for instance, as a bare address) can work if the memory in question
is reused. (The formal definition of access equality is that it returns
True if both access values designate the same object; that can never
be True if one of the values is a dangling reference, and the other is
not, but both values could refer to the same memory.) Membership tests
that do not involve an implicit dereference generally do not depend on
the access value at all.
We allow Constraint_Error to be raised here
so that dangling reference and null pointer checks can be combined into
a single check. If different exceptions are required, then the checks
have to be made separately - but there's little semantic difference (neither
designate a usable object).
Ramification: If a dangling reference
is assigned into an object, including being passed to a formal parameter,
that object also contains a dangling reference afterwards.
Discussion: For equality and membership
operations on composite types, this applies to any parts that are access
types, as these operations are created based on the operations of the
components (which triggers the bounded error). For other operations on
composite types, the bounded error is not triggered. For instance, an
assignment of a composite object with a subcomponent that is a dangling
reference has to work normally; no exception can be raised, but the target
object will have a subcomponent that is a dangling references, and a
(direct) use of that subcomponent is again a bounded error. This is similar
to the way that assignments of invalid subcomponents are handled (see
13.9.1).
Erroneous Execution
{
AI05-0033-1}
{
AI05-0262-1}
Evaluating a name that denotes
a nonexistent object, or a protected subprogram or subprogram renaming
whose associated object (if any) is nonexistent, is erroneous. The execution
of a call to an instance of Unchecked_Deallocation is erroneous if the
object was created other than by an
allocator
for an access type whose pool is Name'Storage_Pool.
Reason: {
AI05-0033-1}
{
AI05-0262-1}
The part about a protected subprogram is intended to cover the case of
an access-to-protected-subprogram where the associated object has been
deallocated. The part about a subprogram renaming is intended to cover
the case of a renaming of a prefixed view where the prefix object has
been deallocated, or the case of a renaming of an entry or protected
subprogram where the associated task or protected object has been deallocated.
Ramification: {
AI05-0157-1}
This text does not cover the case of a name that contains a null access
value, as
null does not denote an object (rather than denoting
a nonexistent object).
Implementation Advice
For a standard storage pool, Free should actually
reclaim the storage.
Implementation Advice: For a standard
storage pool, an instance of Unchecked_Deallocation should actually reclaim
the storage.
Ramification: {
AI95-00114-01}
This is not a testable property, since we do not know how much storage
is used by a given pool element, nor whether fragmentation can occur.
{
AI05-0157-1}
A call on an instance of Unchecked_Deallocation with a nonnull access
value should raise Program_Error if the actual access type of the instance
is a type for which the Storage_Size has been specified to be zero or
is defined by the language to be zero.
Implementation Advice: A call on an instance
of Unchecked_Deallocation with a nonnull access value should raise Program_Error
if the actual access type of the instance is a type for which the Storage_Size
has been specified to be zero or is defined by the language to be zero.
Discussion: If the call is not illegal
(as in a generic body), we recommend that it raise Program_Error. Since
the execution of this call is erroneous (any allocator from the pool
will have raised Storage_Error, so the nonnull access value must have
been allocated from a different pool or be a stack-allocated object),
we can't require any behavior — anything at all would be a legitimate
implementation.
30 The rules here that refer to Free apply
to any instance of Unchecked_Deallocation.
31 Unchecked_Deallocation cannot be instantiated
for an access-to-constant type. This is implied by the rules of
12.5.4.
Wording Changes from Ada 95
{
AI95-00416-01}
The rules for coextensions are clarified (mainly by adding that term).
In theory, this reflects no change from Ada 95 (coextensions existed
in Ada 95, they just didn't have a name).
Wording Changes from Ada 2005
{
AI05-0033-1}
Correction: Added a rule that using an access-to-protected-subprogram
is erroneous if the associated object no longer exists. It is hard to
imagine an alternative meaning here, and this has no effect on correct
programs.
{
AI05-0107-1}
Correction: Moved the requirements on an implementation-generated
call to Deallocate to
13.11, in order to
put all of the rules associated with implementation-generated calls to
Allocate and Deallocate together.
{
AI05-0157-1}
Correction: Added wording so that calling an instance of Unchecked_Deallocation
is treated similarly to
allocators
for access types where
allocators
would be banned.
Inconsistencies With Ada 2012
{
AI12-0148-1}
Corrigendum: Defined a "dangling reference",
and specified that a dangling reference might designate some other existing
object. This allows simple implementations of access values and reuse
of object memory after deallocation. In prior versions of Ada, "="
between a dangling reference and an access to an existing object has
to return False, even if the existing object and the object designated
by the dangling reference are allocated in the same memory. A program
that depended upon that could break with this revised rule. However,
as a practical matter, almost all Ada implementations use simple implementations
of access types that do not meet that requirement. So such a program
would not work (consistently) on most Ada implementations; thus the change
shouldn't break any existing programs - it just aligns the Standard with
actual practice.
{
AI12-0148-1}
A side effect of this change is to allow an Ada implementation to detect
dangling references in more places. This does not require any Ada implementation
to change, and if the implementation does change, it just means that
errors will be detected earlier.
Wording Changes from Ada 2012
{
AI12-0148-1}
Corrigendum: Clarified that deallocated objects cease to exist
after finalization but before Deallocate is called. This is necessary
to prevent erroneous execution from being triggered by the rules in
13.11
in the time between the end of finalization and the end of the call to
the instance of Unchecked_Deallocation.
Ada 2005 and 2012 Editions sponsored in part by Ada-Europe
