It's mentioned in the makefile, but if you define NDEBUG in Crypto++ you must also define NDEBUG for your app, and vice versa, because it changes the memory layout of some Crypto++ classes. If you are not doing that consistently, then that's probably what's causing the seg fault.
On Fri, Jan 09, 2004 at 04:17:11PM -0800, Michael Hunley wrote: > BTW: I don't know if this is part of the problem or just makes it more > confusing, but I was (accidentally) building the debug Crypto++ with a > release build of my code (NDEBUG defined and -O1), so crypto was > un-optimized, except for templates and inlines. I am trying (so far > unsuccessfully) to build both versions and link in the right version in my > executable (using GNUMakefile and gmake for Crypto++ and boost's bjam > system for my app and libs). There appear to be problems with asserts > still showing up in the linker even with a make clean. > > Would anyone be interested in a bjam make for Crypto++ once I put it together? > > Michael Hunley > Senior Engineer > PocketPurchase, Inc.
