[Cbor] Failure Testcases (was: RE: Simple Value Example)

[Richter, Jörg <[e-mail]>]

> > BTW: Is there a collection of “corrupt” CBOR messages that must produce
> > a decode error?
> 
> We have talked about this:
> 
> https://github.com/cbor-wg/CBORbis/issues/16
> 
> But so far nobody has started collecting instances.
> Do you have some we could start with?
> I was planning to scan some of the open source CBOR packages for tests of
> failing decoding.

Okay. I've been thinking about it for a while.

One set of test cases can be easily constructed from valid test cases. 
Just decode a prefix of a valid test case and the decoder should fail with 
an "end of message" error.  Just generate them during normal testing.

I think a second set of failing test cases can be constructed by just 
considering which CBOR prefixes are invalid. This test cases should of course 
fail with a different error than "end of message".

I wrote and attached a simple Lua script to generate all invalid 1 and 2 byte 
prefixes.

This script outputs 578 test cases.  Also attached.

Another set of test cases can be constructed when using tags.  But I think
this is more involved and not clearly defined.  But I might be wrong.
Consider for example a bigfloat. Is it allowed that the array elements themselves
are using a tag?  

Or what if a tag is followed by an element that is not allowed for this tag.  
Like following a tag for a set (tag=258) with a string instead of an array?  Should this 
cause a decode error?

Or is it allowed that Tag 55799 (0xd9d9f7) is the only content of a CBOR file?

- Jörg