Fabian Cenedese writes: > > > > > > I realize that this is a fairly broad question, but I would like to know > > > > > if there is anything in particular I should take into account when > > > > > trying to generate binaries with the same checksum. > > > > > > > >There could be all manner of reasons. To begin with, diff the > > > >assembly language source produced by g++ -save-temps. That'll get us > > > >started. > > > > > > Is this bigger project made of single or multiple source files? If you're > > > talking about the linked object file it could also be that the linker doesn't > > > always put the modules in the same order (no need to). > > > >Why not? It's not as if the linker randomizes its inputs. Different > >linkers might do things differently, sure, but I'm assuming the OP > >isn't doing anything really silly such as using different linkers. > > My case may have nothing to do with the OP's problem. But we use > gcc/as/ld for our embedded boards. Even with the completely same > environment (we only have few anyway) the linked endresult can > differ when compiled on different machines. That's interesting. As far as I'm aware we have never guaranteed that gcc/binutils will generate binary reproducible output, but there must be a reason for any differences. > It may have to do with the fact that we collect some modules in > libraries and feed these to the linker. The order of the modules > inside the libraries may not necessarily be the same as we > delete/add/update the libraries, depending on the changed source > files. But even a complete rebuild doesn't always output the same > file, the changes are obvious in the according map file. Andrew.
