When I have more features implemented in my project, I work on it.

I personally use logfiles, OS and general PC infos, assert- and trace-macro, usually. For extended error statistic I thought about an addition, just an idea: I give all my classes, methods and important sections an unique number. When a crash or an illegal termination occurs the number of the last activated section is retreived and then a permanent disk-stored array (int g_aErrorArray[sections_max]) is increased at the corresponding position (g_aErrorArray[num]++). After (a few) crashs an user could then also report the error array - you get a fine overview which parts of your code cause serious trouble.
Better would be an automatic tool: Don't store the numbers but the stack context (subroutine backjumps). A tool can than trace back from source lines to your methods and full automatically generate the error visulaizations... yes the second approach looks better.

Btw, thx for your help... will have a look if I can trace back my code from windows crash messages (of a 'release version'). Moak