Whoops! The contest site had been down for a little while due to some configuration problem on the server here.

This site will remain up for a while, and I definitely think we'll run another contest in the future. Share your ideas/suggestions!

All of the valid entries should be displayed on the Entries Page. Entries that are not displayed were either nonsense (few people uploaded empty files, for example), a duplicate (version 1 after version 2 was submitted), or one that we couldn't compile/run.

If you don't see your entry listed and think it should be there, please let me know via the contact form.  I'd be happy to flip the entry's status to Valid.

As mentioned in the announcing article, David Maxwell from Coverity ran the bulk of OMGWTF entries through Coverity's Prevent. It would seem that a lot of contestants did not shoot for that defect-to-line ratio or "1", but the static analysis tool did come up with some interesting finds, which may or may not have been intentional.

Here's a summary of the defects in the finalists. The first number is what we considered "actual" defects (such as not freeing memory, etc), where as the second represents all defects (such as warnings about unsafe usage of sprintf, etc):

Entry #100043	Buggy 4 Function Calculator	15 / 77
Entry #100066	FileSystemHashMapNotepadCalculator	45 / 181
Entry #100099	estimator	19 / 21
Entry #100103	TerseCalc	2 / 87
Entry #100123	WTF Web Calc	38 / 86
Entry #100175	OMG!OCRCAL	15 / 78
Entry #100206	Rube Goldberg's Calculator	20 / 132
Entry #100252	Universal Calculator	5 / 62
Entry #100253	Terry's Calculator	5 / 148
Entry #100265	FerronCalc	0 / 35
Entry #100336	VICE, Virtual Integrated Circuit Engine	142 / 222
Entry #100342	ExtensibleCalc	0 / 62

David wrote up some hilights ...

Finalist #11, VICE, Virtual Integrated Circuit Engine topped the list of programs with the highest defect score among the finalists. Capturing the title was a matter of sheer volume in this case. The provided unit tests made room for more defects. Allocating large data structures on the stack, and throwing exceptions that aren't caught anywhere would be bad programming in a real world application, but fortunately this is just for amusement.

 

Here's an example from virtualorgate.cpp:

 

62   	void VirtualOrGate::Set_InputB(const VirtualConnection * input)
63   	{
64   	  if(input == NULL)
65   	  {
66   	    throw std::invalid_argument("input cannot be null");
67   	  }
68   	  InputB = input;
69   	  TransistorB.Set_Base(input);
70   	}

That throw std::invalid_argument() is never caught, right back up through main().

 

 

Finalist #2, FileSystemHashMapNotepad Calculator was king of memory leaks. Each one was small, but is leaked on every calculation, and that means that they, err, 'add up'.

 

381  	float DoAdd(float op1, float op2)
382  	{
383  	
384  		char* op1str = new char[30];
385  		sprintf(op1str, "%.4f", op1 );
386  	    char* op2str = new char[30];

There's no cleanup at the end of the operation functions, and no deletion of those memory allocations.

 

In third place for defect score is Finalist #5, WTF Web Calc, here's a sample from wtfcalc.c:

 

75   	        clientsock = accept(bindsock,(struct sockaddr*)&incoming, &ilen);
76   	        printf("accepted %d\n", clientsock);
77   	        
At conditional (1): "done == 0" taking true path
78   	        while(!done) 
79   	        {   char c[2] = {0,0};;
Event negative_returns: Tracked variable "clientsock" was passed to a negative sink.
80   	            if(read(clientsock,&c,sizeof(c)-1) != 1) break;
81   	            len++;
82   	           
83   	            strcat(line,c);
84   	            if (c[0] == '\n') {
85   	                printf("rcvd %d: %s",nline,line);
86   	                if(!nline++) strcpy(get,line);
87   	                
88   	                memset(line,0,sizeof(line));
89   	                if(len == 2) {
90   	                    printf(">> end of req\n");
91   	                    done = 1;
92   	                }
93   	                len=0;
94   	                
95   	            }
96   	            
97   	        }

If the call to accept fails, it will return an error code, as a negative number. The code doesn't handle that case, and passing a negative value as the first argument to read will always fail (Invalid descriptor). When the read fails, the program will break out of the while loop, but then it will carry on processing the 'input' which was never read...

 

Finalist #7, Rube Goldberg Calculator has some memory leaks as well.

 

From calcfunc.cpp:

 

43   	float DoSub(float op1, float op2)
44   	{
45   		XmlServer xml = XmlServer();
46   		xml.Open(defaultXmlPath);
47   		Fraction fraction1 = Fraction(op1);
48   		Fraction fraction2 = Fraction(op2);
49   		Fraction* result = xml.ReadFraction(fraction1, fraction2, XmlServer::FunctionType::Subtract);
50   		if(result == NULL)
51   		{
52   			xml.WriteFraction(fraction1, fraction2, XmlServer::FunctionType::Subtract, fraction1 - fraction2);
53   			xml.Close();
54   			return DoSub(op1, op2);
55   		}
56   		return result->ToFloat();
57   	}

 ReadFraction allocates storage for the Fraction, and DoSub owns that pointer once it is returned, but returning result->ToFloat() returns the value, not the address, and after that it's too late to clean up. Points should be awarded for consistency though, since the same leak exists in all the other operations implemented too. Rube would be proud.

Good news, everyone! Sometime late Thursday evening, Intern Boyd was able to finish up validating each and every submission on the list. Most of the submissions made it to the next round, but a handful didn't; Boyd did what he could to resolve compile/configure errors, but there's just too many submissions and too little time.

Yesterday (Friday), Jake, Boyd, and I sat around the conference table and discussed half of the valid submissions (second half comes Tuesday). I can see already that this is going to be a very difficult decision to pick out finalists, let alone pick the winner. So many of you did such an incredibly good job at writing an incredibly bad calculator. Seriously, it's impressive.

I look forward to writing up and discussing your horrific work! More to come soon ...

Just a quick post to let you all know that we* are very busy going through all of your entries. All **three hundred and fifty** of them. Seriously, three hundred and fifty. Special thanks to those of you who thought it'd be fun to make sure that compiling took at least ten minutes...

* And of course, by "we", I mean Intern Boyd. But he does report that he just finished all of the Win32 ones. Only another 100 to go!

All entries must be submitted by midnight, EST.  Be sure to get yours in so you don't miss the chance to get your high-res JPG (plus a laptop with which to admire it).  We're madly compiling and testing entries and will be announcing winners soon.

Good news, everyone! The GTK Skeleton Solution is now available at http://omg.worsethanfailure.com/Entries/Skeleton/gtk.zip. It's structured almost exactly like the Win32 solution in that the UI and calculator logic are separate. Enjoy!

If, after submitting your submission, you're not receiving the emails, it's most likely because your email provider thinks it's a spam message. Both Gmail and Hotmail are in this club.  And if you don't see it in your spam folder, just drop me an email via the contact form with your submission number, and I'll set your email as verified.

Important: you email must be verified. Otherwise, we won't even look at the submission. And that would be a waste of a perfectly good submission.

I've updated the FAQ section with some questions we've been seeing.  Here're the changes:

How much leeway can I have on the test cases? If I have it say "ERROR" instead of "Err" for 1 / 0, is that still legal?
No leeway; your calculator must display the exact output we list. You can, however, have your program do other stuff outside of the calculator's display screen (using alert messages, opening other windows, etc).

What kind of system will the entries be run on?
Windows XP Professional SP 2 or NetBSD 3.1, 2 GHz with 512 MB memory.

NetBSD?! Why aren't you guys using Ubuntu Purple Poodle or some other obscure distro that no one other than me has ever heard of? I'm so writing an open letter on Slashdot about this...
No, dear god, please! Not an open letter! We'd be ruined! Just deal with NetBSD.

Is there an execution time requirement?
No operation should take more than several seconds. If you're afraid that your architecture will run too slow, email us a brief summary of what you want to do and we'll let you know if it's acceptable.

I have a question about an idea I have. I know you said to post it in the forum, but this is a really unique idea that I don't want the thieving WTF community to steal.
Oh, crap, we should've thought of that. Email Alex or Jake using the contact form and we'll get back to you.

We're trying to get the GTK solution up as soon as we can; expect it by the end of the week.