While I doubt that - in practice - a compiler would generate incorrect code in this case, I agree that it's a possibility unless we compile with "-fno-strict-aliasing". But I don't want to do that, because it could make the whole of the code less efficient. (And changing the Makefiles to compile only this one file - "MatroskaFileParser.cpp" - with that flag would be messy.)
However, reading through some of the comments at the bottom of <http://labs.qt.nokia.com/2011/06/10/type-punning-and-strict-aliasing/>, it's noted that one can get around strict aliasing restrictions by using "memcpy()". So, with this in mind, try the following alternative implementation of "MatroskaFileParser::parseEBMLVal_float()", and let us know if it makes your compiler happy: Boolean MatroskaFileParser::parseEBMLVal_float(EBMLDataSize& size, float& result) { unsigned resultAsUnsigned; if (!parseEBMLVal_unsigned(size, resultAsUnsigned)) return False; if (sizeof result != sizeof resultAsUnsigned) return False; memcpy(&result, &resultAsUnsigned, sizeof result); return True; } Ross Finlayson Live Networks, Inc. http://www.live555.com/
_______________________________________________ live-devel mailing list live-devel@lists.live555.com http://lists.live555.com/mailman/listinfo/live-devel
