What can go wrong with this code?

[iago]

I dont' see how that would create a problem, though, since it still correctly identifies everything that the documentation says it will...

[Skywing]

I dont' see how that would create a problem, though, since it still correctly identifies everything that the documentation says it will...

The operation of the function is undefined when given a negative integer that is not EOF.  This means that it could possibly identify your input as whitespace even if it is not whitespace, or even crash.

Consider a simple implementation somewhat like this...:

Code Select Expand


int isspace(int ch)
{
char table[255] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 1, ...};
if(ch == EOF) ...
else return table[ch];
}


If you pass a negative value that isn't EOF, it will be reading from memory that isn't in the table.

Of course, exactly what will happen depends on how your particular isspace implementation works, but it probably won't be what you want.

[iago]

Wouldn't it correctly idenfify the -1 as EOF?  Assuming you're using that as a terminator..

[Skywing]

Wouldn't it correctly idenfify the -1 as EOF?  Assuming you're using that as a terminator..

"If you pass a negative value that isn't EOF, it will be reading from memory that isn't in the table."

[iago]

hm, I didnt' read your post right.  I see what you mean now!

I didn't notice you declared table[255].  You're right, then, that if you try to reference a character >127, it'll reference a negative value.

That was my bad

[Adron]

The definition for the isspace I checked didn't mention not supporting negative numbers. It just said it will return true for those particular numbers and false for all others.

[iago]

I agree with Adron here.  If you had given your more recent code instead of using isspace, it would have been more clear.  As it was, everything acted completely predictably in the manner that the functions involved say they will.