Stroustrup's _The C++ Programming Language_, twice in a row, because I didn't want to be left out for not having taken Computer Science.
Butelhof's _Programming with Posix Threads_, because another book referenced its explanation of why Win32 threading had a bug because it could not, at the time, signal and wait on a mutex in an atomic way. There was something deep to find in those rowboats and bucket stories.
GoF's _Design Patterns: Elements of Reusable Object-Oriented Software_, because when I complained to a mentor that I was struggling to organize a dozen scientist programmers, he handed it to me, laughing.
John Robbins, _Debugging Applications_, first edition, because it was the first time for me that code didn't have the disclaimer, "of course you should insert error-handling code here." Basic x86 assembly in one chapter!