From Newsgroup: comp.lang.c++

I recently used C++11's delegating constructors for the first time.
I thought that when I call a delegating constructor and I thow an
exception from the constructor that calls the delegating constructor
I've to call the destructor manually. But it's not like that:

#include <iostream>

using namespace std;

struct S
{
S() { cout << "S::S()" << endl; }
S( int ) : S() { throw invalid_argument( "sads" ); }
~S() { cout << "S::~S()" << endl; }
};

int main()
{
try
{
S s( 123 );
}
catch( ... )
{
}
}

Prints:

S::S()
S::~S()
--- Synchronet 3.20a-Linux NewsLink 1.114

From Newsgroup: comp.lang.c++

On 01/10/2024 10:44, Bonita Montero wrote:

I recently used C++11's delegating constructors for the first time.
I thought that when I call a delegating constructor and I thow an
exception from the constructor that calls the delegating constructor
I've to call the destructor manually. But it's not like that:

When you have called the delegating constructor, the object is fully constructed. The rest of the code comes after that, on the constructed
object - so the destructor will be called as normal if there is an
exception during the rest of the constructor.

If you want the standard reference, it is in paragraph 4 of
"[except.ctor]", at least in C++20 :

"""
If the compound-statement of the function-body of a delegating
constructor for an object exits via an exception, the object’s
destructor is invoked. Such destruction is sequenced before entering a
handler of the function-try-block of a delegating constructor for that
object, if any.
"""

#include <iostream>

using namespace std;

struct S
{
    S() { cout << "S::S()" << endl; }
    S( int ) : S() { throw invalid_argument( "sads" ); }
    ~S() { cout << "S::~S()" << endl; }
};

int main()
{
    try
    {
        S s( 123 );
    }
    catch( ... )
    {
    }
}

Prints:

S::S()
S::~S()

--- Synchronet 3.20a-Linux NewsLink 1.114

From Newsgroup: comp.lang.c++

On 10/01/24 1:44 AM, Bonita Montero wrote:

I recently used C++11's delegating constructors for the first time.
I thought that when I call a delegating constructor and I thow an
exception from the constructor that calls the delegating constructor
I've to call the destructor manually. But it's not like that:

Um... Why would this even be a question? Why would the idea of "calling
the destructor manually" even enter consideration?
--
Best regards,
Andrey
--- Synchronet 3.20a-Linux NewsLink 1.114

From Newsgroup: comp.lang.c++

Am 01.10.2024 um 16:21 schrieb Andrey Tarasevich:

Um... Why would this even be a question? Why would the idea of "calling
the destructor manually" even enter consideration?

Because I've never used delegating constructors so far and I didn't
know where the object entered the state with complete construction.
--- Synchronet 3.20a-Linux NewsLink 1.114

From Newsgroup: comp.lang.c++

Bonita Montero writes:

Am 01.10.2024 um 16:21 schrieb Andrey Tarasevich:

Um... Why would this even be a question? Why would the idea of "calling the
destructor manually" even enter consideration?

Because I've never used delegating constructors so far and I didn't
know where the object entered the state with complete construction.

If the object already "entered the state with complete construction", then
an explicit destructor results in UB when the destructor gets called again, for real-sies. It is fundamental to C++ that (barring immaterial
distractions like memory leaks) C++ will invoke the destructor, guaranteed, whenever a fully-constructed object goes bye-bye.

So, if the object was not fully constructed you will end up calling a destructor on an object that was never fully constructed. And if it was
fully constructed, this ends up calling the destructor twice. So, either
way, it makes no sense.

--- Synchronet 3.20a-Linux NewsLink 1.114

From Newsgroup: comp.lang.c++

Am 02.10.2024 um 01:48 schrieb Sam:

If the object already "entered the state with complete construction",
then an explicit destructor results in UB when the destructor gets
called again, ...

Because the constructor delegating to another constructor does also
a part of the initialization I thought I've to call the destructor
manually.


--- Synchronet 3.20a-Linux NewsLink 1.114

From Newsgroup: comp.lang.c++

On 10/01/24 7:23 AM, Bonita Montero wrote:

Am 01.10.2024 um 16:21 schrieb Andrey Tarasevich:

Um... Why would this even be a question? Why would the idea of
"calling the destructor manually" even enter consideration?

Because I've never used delegating constructors so far and I didn't
know where the object entered the state with complete construction.

[except.ctor] says

4 If the compound-statement of the function-body of a delegating constructor for an object exits via an exception, the object's
destructor is invoked. [...]

https://timsong-cpp.github.io/cppwp/except.ctor#4

I.e. once the the target constructor has finished its work, the object
is considered to be fully constructed, and it is considered to be OK to
invoke its "full" destructor. So, if the delegating constructor body
throws, the entire object's destructor is invoked (automatically).

This is different from situation when regular constructor throws. In
that case individual sub-object destructors are invoked one by one.

But in either case, the destructor(s) is(are) invoked automatically. No
need to do it manually.
--
Best regards,
Andrey


--- Synchronet 3.20a-Linux NewsLink 1.114

From Newsgroup: comp.lang.c++

Am 02.10.2024 um 05:16 schrieb Andrey Tarasevich:

This is different from situation when regular constructor throws.
In that case individual sub-object destructors are invoked one by one.

Of course, but for me it wasn't obvious that the delegating constructor
gets a completed object since I used delegation for the first time.
--- Synchronet 3.20a-Linux NewsLink 1.114