[mtm]

I was a user of a very early version of Zortech C++ (one of the first native C++ compilers (late '80s); no running through cfront) and ran into a very subtle bug in the math routines: sometimes when doing math with 8 bit values the answers would be corrupted. Took a while for me to track down the bug but I eventually found it in the assembly output. Turns out that the math was actually being done with 16 bit registers and the high byte (AH) was not being cleared and this would sometimes set a carry (I think, my memory is fuzzy on the details).

Walter was quick with a fix. Sometimes it is the compiler, but usually it's not.

[mark-r]

I was using a cross compiler for the 68000 around the same time period, and we ran into many compiler bugs. They were always quick to fix them once we could produce a small example program to demonstrate the problem, but it was a bit frustrating. Thankfully it has been many years since I ran into a true compiler bug.