Correct the behavior and return value of snprintf()

[DrXiao]

Recently, I tested the following C code in the test suite:

int main() {
    char buffer[20];
    int written = snprintf(buffer, 0, "Number: %d", -37);
    printf("%s", buffer);
    return written;
}

If using the current shecc to test, the exit code is 0. However, the exit code is 11 if using GCC to validate it.

Due to this difference, I read C99 7.19.6.5 to understand the specification of snprintf() and noticed the following description in that section:

• If n is zero, nothing is written.
• Output characters beyond the n-1st are discarded rather than being written to the array, and a null character is written at the end of the characters actually written into the array.
• The snprintf function returns the number of characters that would have been written
  had n been sufficiently large.

After observing the current implementation of snprintf() in the built-in C library and ran the related test cases again, I found that the behavior and return value of snprintf() did not match the above description.

Therefore, the proposed changes improve the implementation of snprintf() to comply with the specification, and the related test cases are also adjusted to verify the correct behavior.

Summary by Bito

This pull request enhances the snprintf function in the C library to comply with the C99 standard, particularly for zero size parameters. Modifications ensure correct return values and behavior, along with the addition of new test cases for validation, improving reliability across compilers.

Unit tests added: True

Estimated effort to review (1-5, lower is better): 2 - The changes are straightforward and well-documented, making the review process relatively easy.

[jserv]

Instead of saying "you should refer to," quote the essential contents from C standards if you want to emphasize.

[DrXiao]

Consider the following code:

int main()
{
    char buffer[20];
    for (int i = 0; i < 19; i++)
        buffer[i] = '0';
    buffer[19] = 0;

    int written = snprintf(buffer, 10, "Number: %06d", -35337);

    for (int i = 0; i < 20; i++)
        printf(" %x", buffer[i]);
    printf("\n");
    return written;
}

When using shecc with the proposed changes and GCC to test, it writes at most 10 bytes.

 4e 75 6d 62 65 72 3a 20 2d 0 30 30 30 30 30 30 30 30 30 0

However, using the original shecc to compile and execute, it writes more than 10 bytes.

 4e 75 6d 62 65 72 3a 20 2d 0 35 33 33 37 30 30 30 30 30 0

This test code also shows that the original implementation of snprintf() is faulty. I will add more test cases to validate and refine the comments in the test suite.

[vacantron]

Maybe store the number of written bytes or the original buf that we can simply return the string length by them instead of doing the calculation stuff of len?

[DrXiao]

I have defined a new member called len in the structure to store the return value for the functions in the printf() family.

[jserv]

Shorten the quotation, making the comments lean and still useful.

[vacantron]

LGTM

[jserv]

Thank @DrXiao for contributing!