Yup, I got an older version (0.9.6) to compile OK with CW Pro 7.1 and GUSI (POSIX compatibility library). Although there is a port that is totally "native" (i.e. not using GUSI) it is completely unmaintained. I managed to get a newer version built later, but I had to manually sync the file list from the Makefile to the CW project file. So it was a bit of a pain, and I wanted something I could build with an older version of CW.
PolarSSL is not as widespread, but it is actively maintained and used widely enough in embedded systems that I feel comfortable using it. Plus, it being small and more manageable/readable is a huge plus. Once I rewrote the PPC asm stuff to work with CW the performance became quite good, too. Surprisingly there is also 68020 assembly code upstream, but I haven't ventured into porting it just yet.
The biggest issue for my port is that of entropy and RNG, but that is also a problem for OpenSSL. PolarSSL might actually have an advantage here because of its use on limited embedded systems without OS or hardware entropy sources. I am not at all an expert in this field, so I have to trust the result that I got from the entropy analyzer "ent," which showed good-enough results on both PPC and 68k.
On a more general note, I've also been writing an OAuth library in C that works with PolarSSL on CodeWarrior. It can successfully authenticate against Twitter over SSL on my Mac. I am cleaning it up by decoupling the HTTP/network code from the OAuth signature stuff, so when that's done I'll put it up on Github.
PolarSSL is not as widespread, but it is actively maintained and used widely enough in embedded systems that I feel comfortable using it. Plus, it being small and more manageable/readable is a huge plus. Once I rewrote the PPC asm stuff to work with CW the performance became quite good, too. Surprisingly there is also 68020 assembly code upstream, but I haven't ventured into porting it just yet.
The biggest issue for my port is that of entropy and RNG, but that is also a problem for OpenSSL. PolarSSL might actually have an advantage here because of its use on limited embedded systems without OS or hardware entropy sources. I am not at all an expert in this field, so I have to trust the result that I got from the entropy analyzer "ent," which showed good-enough results on both PPC and 68k.
On a more general note, I've also been writing an OAuth library in C that works with PolarSSL on CodeWarrior. It can successfully authenticate against Twitter over SSL on my Mac. I am cleaning it up by decoupling the HTTP/network code from the OAuth signature stuff, so when that's done I'll put it up on Github.