The assert that fails is: XP_ASSERT(offset < reg->hdr.avail); in nr_FindAtLevel() offset is equal to 303 and reg->hdr.avail = 162 The stack backtrace is: KERNEL32! bff767d7() nr_FindAtLevel(_regfile * 0x00c495c0, long 303, char * 0x00b1faac, _desc * 0x00b1fcb0, long * 0x00000000) line 1443 + 43 bytes nr_RegAddKey(_regfile * 0x00c495c0, long 130, char * 0x0045a670, long * 0x00b1fcf4) line 1881 + 26 bytes nr_InitStdRkeys(_regfile * 0x00c495c0) line 1806 + 25 bytes NR_RegOpen(char * 0x00c49640, void * * 0x00b1fd30) line 2137 + 9 bytes NR_StartupRegistry() line 3512 + 14 bytes nsRepository::Initialize() line 345 nsRepository::checkInitialized() line 323 + 5 bytes nsRepository::RegisterFactory(const nsID & {...}, const char * 0x00412840, int 0, int 0) line 466 NS_SetupRegistry() line 145 + 20 bytes nsViewerApp::SetupRegistry(nsViewerApp * const 0x00c49d30) line 158 nsViewerApp::Initialize(nsViewerApp * const 0x00c49d30, int 1, char * * 0x00c48040) line 169 + 12 bytes main(int 1, char * * 0x00c48040) line 95 mainCRTStartup() line 338 + 17 bytes As you can see, it happens almost as soon as the system is started.

Turned out that deleting the c:\windows\nsreg.dat file solved the problem. I don't know if the file was corrupt or from an earlier (incompatible version) but that's what was causing the problem. This could do with some sanity checking, but since the program died in an assert (the release build worked fine), sanity checking the sanity checking could be regarded as going a bit far :)

closing bug, as it seems you found the problem. Plus scullin no longer works here.