Dividing two integers should be round towards zero

[r888800009]

Hello, I am trying to bind some existing c programs and libraries to python, and I have a problem when creating an array object. Then I found that the parser did not round towards zero in the integer division, and the array was multiplied by float cause the struct to fail

The following is the current parser result

// Test int div 9 / 2 should be 4
float x3 = 9.0 / 2.0;
int x4 = 9 / 2;
float x5 = 9 / 2;

(4.5, Type('float'))
(4.5, Type('int'))
(4.5, Type('float'))

If you actually test the c program, it should be the following result

#include <stdio.h>
int main()
{
    float x3 = 9.0 / 2.0;
    int x4 = 9 / 2;
    float x5 = 9 / 2;
    printf("%f\n", x3);
    printf("%d\n", x4);
    printf("%f\n", x5);
    return 0;
}

4.500000
4
4.000000

Then I also made some testcases, if necessary r888800009@d02ee92

Also note that python integer division will take floor, but c is rounded towards zero

python3 -9 // 2 = -5
c -9 / 2 = -4

Update1

I've researched the issue, recombine(tok) causes eval to lose type, and currently I think there are two possible ways to do it.
The first one is to avoid recombine and directly processing on AST,
The second is to wrap a class outside the int type to check when division operation.

Both of these methods will inevitably make a lot of modifications to the paser or testcase