The team lead on the C++ project I alluded to in another thread was like that. I found delicious irony in the outcome of the story. When ARCO got bought, BP was not allowed to keep ARCO's North Slope assets. As a consequence ConocoPhillips bought them and got the system I'd worked on. I learned from friends that they kept all my code in C & FORTRAN and pitched the C++.
The part that really blew my mind was there was a 6x differential in productivity. It took them 6 weeks to build the UI for 1 week of my work. I saw why when I was asked to fix a piece of the UI code as I'd run out of things to do and they were behind. It was a single, several thousand line function created by a UI builder and then hacked. I fixed it, but was so disgusted I told the project manager I would not work on any more of it. I signed my name to my code and included my ACM email address in the header comment. I got laid off not long after and ARCO was sold to BP a few months later.
The saddest part of that project was sitting with another scientist while he was testing it. If you wanted to display 6 curves, it drew one, erased the screen, drew two, erased the screen and so forth until all of the traces had been drawn. This took about a minute and he would just giggle. And if you attempted to scroll up or down in the curves it repeated all the redraws. The team lead tried to get me to join LinkedIn so he could reference me. I declined.
My first contract project was in active use for 12-14 years with about 1/2 that unmaintained. We had 12 user bugs the first year and it went to near zero. There were a pair of 15,000 line libraries in that package that *never* had a bug found. The single change was when Sun broke the C stdlib and failed to correctly set errno on a call to getcwd(3c). The fix was to test for a null pointer return on Sun systems. It also ran from the same source on AIX, HP-UX, IRIX, Ultrix, CLIX (Intergraph) in addition to SunOS/Solaris. Except for things like the Sun gaff the only #ifdefs were byte sex and FORTRAN RECL= bytes or words. It was a 500,000 line port from VAX FORTRAN & VMS to Unix. Lots of it written by scientists who were not good programmers.
The coolest part of that project was the build system, YAMS, Yet Another Make System. You could create a directory, cd to it, and check out the source code. You could then open xterms on all the systems in that same directory and after making a change, could simultaneously compile and run the regression test suite on all platforms. This made everyone very sensitive to the language standards and portability. The IRIX and Ultrix ports were done as an afternoon lark by me when I didn't really have anything serious to do. Probably waiting for a staff meeting.
Good programmers never have to modify their code unless someone else breaks theirs.
Have Fun!
Reg
