Why?And for what purpose? I cannot understand the following words: // verify that types are complete for increased safety thanks
On Friday, September 20, 2002, at 06:34 PM, zhangwusheng wrote:
Why?And for what purpose? I cannot understand the following words:
// verify that types are complete for increased safety
The rules in the C++ standard are that if you delete using a pointer in a context where the type of the object pointed to is not completely defined, the object will be deleted, but the destructor will not be called. Many people think that's a particularly bad rule, because the code compiles and runs, but does the wrong thing. Since it contains a typedef that includes sizeof(T), this function will fail to compile if the type T is not complete. This means that the checked delete function template will only compile if the destructor will be called. That way the compiler will detect a programming mistake that would otherwise result in deleting the object without calling the destructor. This is likely to save time that would otherwise be spent debugging problems caused by destructors that are not called. A lot of us are hoping that like a future version of the C++ standard will make delete itself behave like this. -- Darin
At 01:30 AM 9/22/2002, Darin Adler wrote:
The rules in the C++ standard are that if you delete using a pointer in a context where the type of the object pointed to is not completely defined, the object will be deleted, but the destructor will not be called. Many people think that's a particularly bad rule, because the code compiles and runs, but does the wrong thing.
Since it contains a typedef that includes sizeof(T), this function will fail to compile if the type T is not complete. This means that the checked delete function template will only compile if the destructor will be called. That way the compiler will detect a programming mistake that would otherwise result in deleting the object without calling the destructor. This is likely to save time that would otherwise be spent debugging problems caused by destructors that are not called.
Darin, Nice description of the problem, and the solution.
A lot of us are hoping that like a future version of the C++ standard will make delete itself behave like this.
While I don't following the C++ committee's Evolution Working Group closely, I don't recall actually seeing a proposal the above. If not, maybe you should consider writing a paper. Cheers, --Beman
From: "Beman Dawes"
A lot of us are hoping that like a future version of the C++ standard will make delete itself behave like this.
[fail at compile time on incomplete types]
While I don't following the C++ committee's Evolution Working Group closely, I don't recall actually seeing a proposal the above. If not, maybe you should consider writing a paper.
I find the other alternative - making delete "just do the right thing" and invoke the proper destructor for incomplete types - more "interesting". :-)
Thanks again.
What puzzled me is the noun "complete type" and
"incompltete type",I don't catch it at first glance.
After reading the holy standard 3.9, I understand .
thanks.
"Darin Adler"
On Friday, September 20, 2002, at 06:34 PM, zhangwusheng wrote:
Why?And for what purpose? I cannot understand the following words:
// verify that types are complete for increased safety
The rules in the C++ standard are that if you delete using a pointer in a context where the type of the object pointed to is not completely defined, the object will be deleted, but the destructor will not be called. Many people think that's a particularly bad rule, because the code compiles and runs, but does the wrong thing.
Since it contains a typedef that includes sizeof(T), this function will fail to compile if the type T is not complete. This means that the checked delete function template will only compile if the destructor will be called. That way the compiler will detect a programming mistake that would otherwise result in deleting the object without calling the destructor. This is likely to save time that would otherwise be spent debugging problems caused by destructors that are not called.
A lot of us are hoping that like a future version of the C++ standard will make delete itself behave like this.
-- Darin
participants (4)
-
Beman Dawes -
Darin Adler -
Peter Dimov -
zhangwusheng