[Zlib-devel] how trusted is contrib

[William A. Rowe, Jr.]

Mark Adler wrote:
> 
> If folk believe that there is enough testing, we could consider  
> moving the asm stuff into the main directory to make a default part  
> of the zlib build.

Would folks entertain a perpetual test, using some arbitrary data stream (the
contents of your swap file or raw partition might be useful) broken up into
pseudorandom sized blocks, with one thread leading the encoding and a second
thread picking up the decoding/compare?  I figure after a few weeks of that,
virtually any quirk would fall out.  I suppose a -b maxblocks or -s maxbytes
flag might be useful to provide a predictable end to the madness.

Just pondering how we could stress zlib more than hello, hello!  hello world

What other edge cases - other than the bytestream and the packet size, are
likely to trigger edge-case bugs in the code?

Bill