>Well, now you don't have AC :)

Test 14 was added by me recently, which is why~
The only hint i can give for it is that the code prevented TL #13, but did it in not exactly a proper way.

problem description is incorrect there are far more than 2000 lines, and one token is far more than 50 characters....

I had OLE #4, and it turned out to be a minor slip in my code. That is, for commands of type «if smth1 <operator> smth2 goto label, i was searching smth1 and smth2 indices in my variable indices array; label i should have searched in my goto indices array, but i accidentally was searching a label index in variable array too; as a result, it returned 0 (which it returned in case if the needed element was the first one, and if the element wasn't found, which shouldn't be happening). So it jumped to the first label and continued from there.

Example:

label1:
A = 10
print 100
label2: goto label2

Due to a mistake, "label2" has been looked for in variables array, which only had "A" in there. It was not found, it returned 0, it jumped to label1, tons of print commands were executed. After mistake is fixed, "label2" is looked for inside labels array, and correctly returns 1.

I can't say for sure you have the very same mistake, but still it is likely just a minor slip like this one.

Pheeew, finally AC!
After WA #3, which had a test like
print 5
end
, i added 4 lines to fix that, and immediately got WA5. Thanks to Fyodor, i immediately understood what i missed, in case like
a = 1
goto x
b = 2
x: goto x
i initially printed
Program terminated. Variables state:
a: 1
b: 0
So, i fixed that too, and only printed out a: 1, but still was having WA5! Curiously enough, after a fix, which i didn't think would help, i finally got AC. In my program, i had this chunk of code:

        writeln('Program terminated. Variables state:');
{
        for i:=1 to vsz do
            if initialized[i] = 1 then writeln(vnames[i], ': ', vars[i]);
}
        for i:=1 to vsz do begin
            val(vnames[i], j, xcode);
            if xcode <> 0 then begin
                if initialized[i] = 1 then writeln(vnames[i], ': ', vars[i]);
            end;
        end;

The thing is, i threw everything into variable array, both constants and variables, like if i met A = 4 + -3, i throw A, 4, -3 into that array. Then, i go through every variable "name" and use standard val procedure to attempt to convert string into integer. "A" isn't a number, so it remains 0 initially, but "4" and "-3" successfully convert into numbers, so i give them values of 4 and -3 initially. But if i write in my program 4 = 100, then later on something like "A = A + 4" would actually add 100 to A. Soooo...

Using that lower uncommented chunk, i go through variable names again, and convert them into numbers, to see, if they're supposed to be constants. If they are, i ignore them, and if they aren't, i check if they're initialized, and if they are, i print them. That gave me WA5. Test like
    4 = 5
    x: goto x
would give
    Program terminated. Variables state:

Then, i commented that lower chunk, and uncommented the upper chunk, so that no matter if "variable"'s name can be converted to number or not, i print it out anyway if it's initialized. So for test like
    4 = 5
    x: goto x
it now gives
    Program terminated. Variables state:
    4: 5

And that fix got me AC! Either stuff like 4 = 5 is present in 5th test, or i'm missing anything else...
Well, that was fun to discover, haha~

I also implemented support for variable names equal to keywords, so things like "not = not not" or "not: not = not or not" or "print: print print" would be parsed adequately, but seems it wasn't really needed.

P. S. In the end, it turned out that the Val function converts not only decimal, but also hexadecimal numbers, which is why i encountered this problem.

Edited by author 22.10.2015 15:12

My program works perfectly on any of my tests, on any tests from this task's discussion, even if there is no 'end' command and if there is "!=" token, but i got WA#6. Is in this test anything unusual or wrong?

Should I, for example, add an 'end' command, as Ludovic did?
http://acm.timus.ru/forum/thread.aspx?id=13124&upd=632835406669279530

And about limit of commands of 10_000_000 : "The if ... goto command works just as the goto command if the condition after if is true". Should I count an "if..goto" command, if condition was wrong?

I just have no ideas, what can be wrong in my algo. Sorry for my bad English.