>>We were using Microsoft Cobol for CP/M back then. It had indexed files which would become randomly inaccessible once the number of records exceeded 32767. Our distributor was told to keep silent about that, and only when I proved to my boss (a fellow mathematician, luckily) that it would give different results depending on whether you were trying to read certain records by key, by record number or by just reading them sequentially in key order or in physical order. We were given a spare set of indexing routines to use instead of Microsoft's, and I got that to work (imagine storing actual buffers in Cobol, then pointers to those buffers in some control blocks... pointers in Cobol :) only after a few near all-nighters, unnecessary travels and nerves lost.
>
>I don't know what impresses me more, your great mathematical mind and good memory or your never giving up of the past grudges <g>

Not sure I'm still a mathematician... I've gone to the dogs fox long ago.

As for that one, hey, it doesn't happen to everybody that on your first gig you find a serious bug in the runtimes and have to write some code to prove to your boss that it wasn't you. One remembers that until that German guy comes, who hides things around the house so we can't find them.