summaryrefslogtreecommitdiff
path: root/src/context
diff options
context:
space:
mode:
authorMorgan Deters <mdeters@gmail.com>2010-09-01 21:02:07 +0000
committerMorgan Deters <mdeters@gmail.com>2010-09-01 21:02:07 +0000
commit3bfa95fd5044257e41b5276dea2e0f3ac036f457 (patch)
treee2fd4bbfacf1d3d2ba1331dd814ccc2bc647c0f2 /src/context
parentd8c37e4d1cbe1c99b1b5fc250085419b9de4e513 (diff)
added documentation, closes bug 97
Diffstat (limited to 'src/context')
-rw-r--r--src/context/cdo.h6
-rw-r--r--src/context/context.h49
2 files changed, 55 insertions, 0 deletions
diff --git a/src/context/cdo.h b/src/context/cdo.h
index 20c512fbc..ba4b368a8 100644
--- a/src/context/cdo.h
+++ b/src/context/cdo.h
@@ -85,6 +85,9 @@ public:
* This version takes an argument that specifies whether this CDO<>
* was itself allocated in context memory. If it was, it is linked
* with the current scope rather than scope 0.
+ *
+ * WARNING: Read the notes in src/context/context.h on "Gotchas when
+ * allocating contextual objects with non-standard allocators."
*/
CDO(bool allocatedInCMM, Context* context) :
ContextObj(allocatedInCMM, context) {
@@ -111,6 +114,9 @@ public:
* This version takes an argument that specifies whether this CDO<>
* was itself allocated in context memory. If it was, it is linked
* with the current scope rather than scope 0.
+ *
+ * WARNING: Read the notes in src/context/context.h on "Gotchas when
+ * allocating contextual objects with non-standard allocators."
*/
CDO(bool allocatedInCMM, Context* context, const T& data) :
ContextObj(allocatedInCMM, context) {
diff --git a/src/context/context.h b/src/context/context.h
index 20e84d7e5..9a561abb6 100644
--- a/src/context/context.h
+++ b/src/context/context.h
@@ -268,7 +268,9 @@ public:
* This is an abstract base class from which all objects that are
* context-dependent should inherit. For any data structure that you
* want to have be automatically backtracked, do the following:
+ *
* 1. Create a class for the data structure that inherits from ContextObj
+ *
* 2. Implement save()
* The role of save() is to create a new ContexObj object that contains
* enough information to restore the object to its current state, even if
@@ -283,6 +285,7 @@ public:
* fact, any clean-up work on a saved object must be done by restore().
* This is because the destructor is never called explicitly on saved
* objects.
+ *
* 3. Implement restore()
* The role of restore() is, given the ContextObj returned by a previous
* call to save(), to restore the current object to the state it was in
@@ -291,8 +294,10 @@ public:
* automatically. Ideally, restore() should not have to free any memory
* as any memory allocated by save() should have been done using the
* ContextMemoryManager (see item 2 above).
+ *
* 4. In the subclass implementation, any time the state is about to be
* changed, first call makeCurrent().
+ *
* 5. In the subclass implementation, the destructor should call destroy(),
* which repeatedly calls restore() until the object is restored to context
* level 0. Note, however, that if there is additional cleanup required at
@@ -301,6 +306,47 @@ public:
* cannot be in the ContextObj destructor is that it needs to call the
* subclass-specific restore() method in order to properly clean up saved
* copies.
+ *
+ * GOTCHAS WHEN ALLOCATING CONTEXTUAL OBJECTS WITH NON-STANDARD ALLOCATORS
+ *
+ * Be careful if you intend to allocate ContextObj in (for example)
+ * ContextMemoryManager memory. ContextMemoryManager doesn't call the
+ * destructors of contained objects, which means the objects aren't
+ * properly unregistered from the Context (as in point #5, above).
+ * This can be remedied by allocating the ContextObj in the "top
+ * scope" rather than the "bottom scope" (which is what the
+ * "allocatedInCMM" flag to the special constructor in ContextObj
+ * does).
+ *
+ * But why allocate in CMM if it's so complicated? There's a big
+ * advantage: you don't have to track the lifetime of the ContextObj.
+ * The object simply ceases to exist when the Context is popped. Thus
+ * you can create an object in context memory, and if you stow
+ * pointers to it only in context memory as well, all is cleaned up
+ * for you when the scope pops. Here's a list of gotchas:
+ *
+ * 1. For data structures intended solely to be allocated in context
+ * memory, privately declare (but don't define) an operator
+ * new(size_t) and destructor (as currently in the Link class, in
+ * src/theory/uf/ecdata.h).
+ *
+ * 2. For data structures that may or may not be allocated in context
+ * memory, and are designed to be that way (esp. if they contain
+ * ContextObj instances), they should be heavily documented --
+ * especially the destructor, since it _may_or_may_not_be_called_.
+ *
+ * 3. There's also an issue for generic code -- some class Foo<T>
+ * might be allocated in context memory, and that might normally be
+ * fine, but if T is a ContextObj this requires certain care.
+ *
+ * 4. Note that certain care is required for ContextObj inside of data
+ * structures. If the (enclosing) data structure can be allocated
+ * in CMM, that means the (enclosed) ContextObj can be, even if it
+ * was not designed to be allocated in that way. In many
+ * instances, it may be required that data structures enclosing
+ * ContextObj be parameterized with a similar "bool allocatedInCMM"
+ * argument as the special constructor in this class (and pass it
+ * on to all ContextObj instances).
*/
class ContextObj {
/**
@@ -484,6 +530,9 @@ public:
* specifies whether this ContextObj is itself allocated in context
* memory. If it is, it's invalid below the current scope level, so
* we don't put it in scope 0.
+ *
+ * WARNING: Read the notes above on "Gotchas when allocating
+ * contextual objects with non-standard allocators."
*/
ContextObj(bool allocatedInCMM, Context* context);
generated by cgit on debian on lair
contact matthew@masot.net with questions or feedback