Abhishek Padmanabh
2011-03-21 11:17:18 UTC
Consider the below code that I came across recently:
#include <iostream>
int main ()
{
bool a=false;
std::cout << "insert 0 or 1:";
std::cin >> a;
if (a==true)
{
std::cout << "a is true";
}
else
{
std::cout << "a is false";
}
}
If you run the above code and provide input as say, 8, the code still
returns back saying a is false. I get this behavior with Visual Studio
2008. I would have thought that the integral value should get
implicitly converted to a boolean. In fact, when I debugged the code
in VC++, I do see that it is able to get the value of the user input
(i.e. 8) but then there is a certain check which then fails and
results in setting of the failbit of the stream. If I check the status
of the cin stream post that read, I see the failbit set which means
the read had failed.
relevant bit of VC++ code (file: xlocnum, function: do_get)
Code:
{ // get zero or nonzero integer
char _Ac[_MAX_INT_DIG], *_Ep;
int _Errno = 0;
const unsigned long _Ulo = ::_Stoulx(_Ac, &_Ep,
_Getifld(_Ac, _First, _Last, _Iosbase.flags(),
_Iosbase.getloc()), &_Errno);
if (_Ep != _Ac && _Errno == 0 && _Ulo <= 1)
_Ans = _Ulo;
}
The check _Ulo <= 1 is what rejects it. _Ulo is an unsigned long which
does get successfully evaluated to 8 (the input). So, it is basically
only allowing any other inputs than 0 and 1 via the stream. Could this
be a bug in VC++? I am not sure but I would have thought that it would
have considered any non-zero value as true. Because, if I just simply
set the value of boolean as 8 instead of reading from the stream, the
automatic conversion rules kick in and the code works as expected.
Any views on this as to what the standard says because of which the
above is considered a failure? Thanks for your help.
#include <iostream>
int main ()
{
bool a=false;
std::cout << "insert 0 or 1:";
std::cin >> a;
if (a==true)
{
std::cout << "a is true";
}
else
{
std::cout << "a is false";
}
}
If you run the above code and provide input as say, 8, the code still
returns back saying a is false. I get this behavior with Visual Studio
2008. I would have thought that the integral value should get
implicitly converted to a boolean. In fact, when I debugged the code
in VC++, I do see that it is able to get the value of the user input
(i.e. 8) but then there is a certain check which then fails and
results in setting of the failbit of the stream. If I check the status
of the cin stream post that read, I see the failbit set which means
the read had failed.
relevant bit of VC++ code (file: xlocnum, function: do_get)
Code:
{ // get zero or nonzero integer
char _Ac[_MAX_INT_DIG], *_Ep;
int _Errno = 0;
const unsigned long _Ulo = ::_Stoulx(_Ac, &_Ep,
_Getifld(_Ac, _First, _Last, _Iosbase.flags(),
_Iosbase.getloc()), &_Errno);
if (_Ep != _Ac && _Errno == 0 && _Ulo <= 1)
_Ans = _Ulo;
}
The check _Ulo <= 1 is what rejects it. _Ulo is an unsigned long which
does get successfully evaluated to 8 (the input). So, it is basically
only allowing any other inputs than 0 and 1 via the stream. Could this
be a bug in VC++? I am not sure but I would have thought that it would
have considered any non-zero value as true. Because, if I just simply
set the value of boolean as 8 instead of reading from the stream, the
automatic conversion rules kick in and the code works as expected.
Any views on this as to what the standard says because of which the
above is considered a failure? Thanks for your help.