Understanding compiler options differences in visual studio 2010 and strict C++ compliance

Question

I am grading some homework C++ code and a student used a non-standard constructor for a vector of vectors:

vector<vector<double> > A(rows, cols);

where rows and cols are unsigned integers. The way we taught it in class is

vector<vector<double> > A(rows, vector<double>(cols));

following the fill constructor (2 in http://www.cplusplus.com/reference/vector/vector/vector/)

I am using a batch file to compile all students codes with the command line

cl /O2 /EHsc /Tp <filename>

and this command threw this error at the student line mentioned above:

error C2664: 'std::vector<std::vector<double,std::allocator<_Ty>>,std::allocator<std::vector<_Ty,std::allocator<_Ty>>>>::vector(std::initializer_list<std::vector<_Ty,std::allocator<_Ty>>>,const std::allocator<std::vector<_Ty,std::allocator<_Ty>>> &)' : cannot convert argument 2 from 'unsigned int' to 'const std::vector<double,std::allocator<_Ty>> &'
    with
    [
        _Ty=double
    ]
    Reason: cannot convert from 'unsigned int' to 'const std::vector<double,std::allocator<_Ty>>'
    with
    [
        _Ty=double
    ]
    Constructor for class 'std::vector<double,std::allocator<_Ty>>' is declared 'explicit'
    with
    [
        _Ty=double
    ]

But when I create a project and build it with the default parameters of MSVC 2010 it does not throw neither a warning nor an error there.

I am trying to learn what compiler option is responsible for allowing it go through without a warning in the IDE, and what I would switch it off.

I tried finding an answer in http://msdn.microsoft.com/en-us/library/2tb15w2z(v=vs.100).aspx, but I couldn't an answer.

EDIT: I think this might be helpful to others: thanks to the comments I understand now the constructor called in the IDE is the range constructor (#3 in the link above).

Here's my particular question: both methods use the same compiler with different options (one the default from the IDE, the other one is stated above). The batch file throws an error, the IDE doesn't. I need help identifying what to change in the IDE's command line arguments so that it throws the error.

UPDATE: I included the error message.

UPDATE 2: It turns out the script was being run in a computer with MSVC 2013 and that was the difference

Answer 1

When you're compiling at the command line you are not using VS 2010.

Here's what I get from a VS 2010 command line compile:

C:\so-test>cl /O2 /EHsc /Tp test.cpp
Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 16.00.40219.01 for 80x86
Copyright (C) Microsoft Corporation.  All rights reserved.

test.cpp
Microsoft (R) Incremental Linker Version 10.00.40219.01
Copyright (C) Microsoft Corporation.  All rights reserved.

/out:test.exe
test.obj

I get your error message when I compile from the VS 2013 command line (I get a similar but slightly different message from the VS 2012 command line).

I suggest that you look at the version number displayed to verify what version of the MSVC compiler you're using at the command line.

• VS 2010 is 16.00.xxxx (SP1 is 16.00.40219.01)
• VS 2012 is 17.00.xxxx (Update 4 is 17.00.61030)
• VS 2013 is 18.00.xxxx (Update 1 is 18.00.21005.1)

Answer 2

Actually:

std::vector<std::vector<double> a( rows, columns );

is legal, or at least it was legal until C++11. It wasn't intentionally legal, but the way the standard was worded, a compiler was required to accept it. (I think C++11 fixed this. Although it did so a bit too late, since doing so breaks previously legal code.)

The rule is (or was) simple, if overload resolution on the constructor results in the constructor <template Iter>std::vector::vector( Iter begin, Iter end ) being chosen, and Iter deduces to an integral type, the code must behave as if std::vector::vector( static_cast<size_type>( begin ), static_cast<value_type>( end ) ) were called.