Ian, Thanks for your quick response. Xcode 2.1 uses gcc 4.0.0 with target powerpc-apple-darwin8. Xcode 2.2, using gcc 4.0.1, was just released and I'm downloading it now. I'll check if this behavior is still present. Thanks again, Yoav -----Original Message----- From: Ian Lance Taylor [mailto:ian@xxxxxxxx] Sent: Thursday, November 17, 2005 11:58 PM To: Yoav Gonen Cc: gcc-help@xxxxxxxxxxx Subject: Re: Difference in __cxa_demangle between gcc 3.3 and 4.0 "Yoav Gonen" <yoavg@xxxxxxxxxxxxx> writes: > I'm not sure this is the right place to ask, but here goes. > I'm compiling the following program: > > #include <iostream> > #include <cxxabi.h> > > int main(int argc, char * const argv[]) { > size_t length; > int status; > char* p = abi::__cxa_demangle("d", 0, &length, &status); > if (p) free(p); > return 0; > } > > I'm checking the values of status and p before freeing it. > On gcc 3.3, status = 0 and p = "double". > On gcc 4.0, status = -2 and p = NULL. > > Is this a bug? > > BTW, I'm using Xcode 2.1 on a Mac OS X 10.4. There was a time that the demangler acted as you describe. However, it was fixed for the 4.0 release (the demangler was originally changed to follow the spec for __cxa_demangle, but the spec was determined to be buggy, and both were changed back). I don't know anything about Xcode 2.1. If you are seeing this behaviour with the FSF version of gcc 4.0, then you have found a bug. Ian
