This was stolen from this page.
Why Windows 95 and Windows 98 would crash after 49.7 days of uptime
There’s a famous bug in Windows 95 and Windows 98 (now patched) that caused these systems to stop functioning after 49.7 days of uptime. I used to wonder how this happened – in addition to wondering how many Windows 95 / 98 systems could achieve such uptimes to begin with!
Finally, I did a little math and figured out what must be behind this bug. It’s actually quite cute.
First, we need to know about a function called GetTickTime which is part of the Windows API. This function returns a DWORD value representing the number of milliseconds which has elapsed since the system has come up.
What’s a DWORD? A Word on an x86 is 16 bits, so a Double Word is 32 bits. The maximum number expressed in 32 bits is 2 to the power of 32.
2^32 = 4,294,967,296
We’re expressing a number of milliseconds here. How many milliseconds are there in one day?
(milliseconds per second) * (seconds per minute) * (minutes per hour) * (hours per day) = 1000 * 60 * 60 * 24 = 86,400,000.
Finally, we can find out how many days’ worth of milliseconds we can fit in the DWORD:
4,294,967,296 / 86,400,000 = 49.7102696
Voila! After 49.71 days, the number of milliseconds since startup exceeded the maximum value of a DWORD, and rolled to 0 (kind of like the odometer on Al Bundy’s Dodge rolling to 0 after the 999,999,999th mile)
Depending on how you process the return value of GetTickCount(), this may or may not be a problem. Clearly some code in Windows didn’t handle this too well, and Windows hung.
Great story, I never knew about that. Mainly because I never had a DOS-based Windows system that managed to accumulate 50 days of uptime.
Windows Server, yes. Win7, yes. Win95/98, not so much.