The ‘Sindogs’ Excel Bug

Steven Sinofsky, in a fascinating post on software reliability on his Hardcore Software blog/newsletter:

I created many bugs on my own before Microsoft and learned how to find bugs planted in Microsoft code during my training in Apps Development College (ADC), but I learned about my first commercial bug during my first summer at Microsoft. DanN my lead in ADC shared (JonDe refreshed my memory of the specifics) the story of the infamous “Sindogs” bug in Excel 2.0, which was the first Windows version that shipped with Windows 2.0 about 18 months before I arrived. The bug manifested itself when an important part of Windows, a plain text file with all the system settings, was corrupted — in the file where it was supposed to say “[Windows]” it somehow was changed to “[Sindogs].” Neither the word Sindogs appeared in any code nor did any code write that string, so its appearance was rather mysterious. The bug took days to materialize and was only discovered after Excel testers ran an automated test to create and print charts over and over, for many hours. Eventually, through a significant amount of sleuthing, the team narrowed it down to a bug in drawing code in Windows, which was called when adding arrows to charts then printing them on old-school dot-matrix printers. There was a memory corruption, which changed the contents of the settings file that was in memory before it was saved to the disk. Stories were told about it for years. The Excel team even renamed their file server after the bug, and through Office 97 we connected to the server “\\SINDOGS\REL” (REL was short for release) for release builds of Excel.

Imagine tracking down a crazy bug after the product was in market and trying to figure out what caused it, then multiplying that by all the possible printers, video cards, and programs involved. Looking back, it was an engineering marvel that anything worked at all. In moments of frustration, or desperation, that is what we told ourselves.

Thursday, 6 January 2022