Unofficial wish list status.(Jul 2008)

[Walter Bright]

Bill Baxter wrote:
> Yep.  Thanks for sharing your experience.  I'm a researcher, so while 
> I've been at this for coding thing for 20 years too, I don't tend to 
> play around with a lot of million-line code bases written by 
> million-monkey junior programmers.  I did do a short stint for a big, 
> well-known software company once, and got my taste of million-monkey 
> code, though.  Scary stuff.  They wanted me to modify some crufty parser 
> to do some new tricks when there were absolutely no unit tests or 
> example programs or even documentation to show me what it was supposed 
> to be doing in the first place.  I'm amazed now that they didn't tell me 
> to "figure out this mess and document it first, writes some unit tests, 
> then add feature X" but instead they just wanted me to jump right in and 
> throw more gasoline into the fire.  Yikes.

Microsoft once asked me why I didn't update the MFC library that comes 
with DMC++. I said the problem was that MFC is a large complex system 
with not a single test case for it (at least one that is available to 
me). The existing one was tested heavily by others, and I know it works.

If I recompile a new version, I have NO idea whether it compiled 
correctly or not. It took about a year to squeeze all the bugs out of 
the existing one.

STLport is different, because that comes with a test suite. I can 
recompile it, run the test suite, and I know it works.