Re: is such exception handling approach good?
- From: George <George@xxxxxxxxxxxxxxxxxxxxxxxxx>
- Date: Wed, 26 Dec 2007 22:01:01 -0800
Thanks Ben,
new (old_address_pointer) (constructor parameter)
directly, without call old_address_pointer->destructor?
The C++ standard allows it.
It's not good practice.
Why it is not good code to call destructor at first before call placement
new operator? Any more descriptions please? :-)
regards,
George
"Ben Voigt [C++ MVP]" wrote:
.
"George" <George@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:F72C27EC-6EE7-4713-A0B9-07AD11F7FEEB@xxxxxxxxxxxxxxxx
Thanks Ben,
Note that the C++ standard allows you to reuse memory formerly occupied
by a
class object without the destructor running, unfortunate in my opinion.
You mean the placement new operator? I think it is always good practices
to
call the destructor pointer by some instance pointer, then use the same
memory address to allocate instance by use of the same memory.
You mean we can directly call,
new (old_address_pointer) (constructor parameter)
directly, without call old_address_pointer->destructor?
The C++ standard allows it.
It's not good practice.
regards,
George
"Ben Voigt [C++ MVP]" wrote:
"Larry Smith" <no_spam@xxxxxxxxxxx> wrote in message
news:Onq3SyjRIHA.1168@xxxxxxxxxxxxxxxxxxxxxxx
One question about your sample. If not using smart pointer, and if at
the
same time, the construction of m_Whatever is successful, but the
construction
of current instance is failed and exception is thrown -- throw
SomeException();
then, I think there will be memory leak since delete n_Whatever is not
invoked explicitly?
BTW: if the caller of the constructor MyClass() is smart enough, it
will
catch the exception and delete the member m_Whatever? Is that
possible?
The canonical example:
class MyClass
{
public:
MyClass() : m_Whatever(new Whatever)
{
// m_Whatever's destructor is still called
throw SomeException();
}
private:
std::auto_ptr<Whatever> m_Whatever;
};
The rule is that a destructor will be called only if its constructor
completed successfully. Note that we have two objects here however,
"MyClass" which someone will instantiate, and "m_Whatever" which is
created just before the body of the "MyClass" constructor starts
running.
Each of these objects has its own constructor. Note that by the time
the
body of the "MyClass" constructor starts running however, "m_Whatever"
has
already been fully created and initialized with a pointer to "Whatever"
(which it encapsulats internally). Its destructor is then guaranteed to
run based on my opening statement of this paragraph. Moreover, based on
the very same statement, when "SomeException" is thrown, the "MyClass"
constructor hasn't finished running so its destructor will *not* be
called
(though I haven't explicitly defined one). Only the destructor for
"m_Whatever" is called as noted and that automatically deletes
"Whatever".
You therefore have nothing to do to clean things up yourself. Just
catch
"SomeException" to handle the error and that's it.
Although for this particular case, it's probably better to have a
variable
of type Whatever and avoid the pointer completely.
Note that the C++ standard allows you to reuse memory formerly occupied
by a
class object without the destructor running, unfortunate in my opinion.
- Follow-Ups:
- Re: is such exception handling approach good?
- From: Ben Voigt [C++ MVP]
- Re: is such exception handling approach good?
- References:
- is such exception handling approach good?
- From: George
- Re: is such exception handling approach good?
- From: ajk
- Re: is such exception handling approach good?
- From: Larry Smith
- Re: is such exception handling approach good?
- From: ajk
- Re: is such exception handling approach good?
- From: Larry Smith
- Re: is such exception handling approach good?
- From: George
- Re: is such exception handling approach good?
- From: Larry Smith
- Re: is such exception handling approach good?
- From: Ben Voigt [C++ MVP]
- Re: is such exception handling approach good?
- From: George
- Re: is such exception handling approach good?
- From: Ben Voigt [C++ MVP]
- is such exception handling approach good?
- Prev by Date: Simple question about headers and malloc!
- Next by Date: Re: is such exception handling approach good?
- Previous by thread: Re: is such exception handling approach good?
- Next by thread: Re: is such exception handling approach good?
- Index(es):
Relevant Pages
|