OK, the first problem is the difference between #if defined(USE_DOC_2000_TSOP) #if (USE_DOC_2000_TSOP == 1)
The SiS ipl.h files expect the former, but NLBConfig.py is now generating makefiles that define USE_DOC_2000_TSOP and set it to 0. So it's another one of those 'used to use an #ifdef but should now use an #if' issues. The result is that the ipl.s code as generated for a DOC MIL is looking for linuxbios starting at block 4, when it should be at block 2. The bigger problem is that while the code varies where it expects to see linuxbios, the scripts always start it at block 2. This is really bad. I think for future-proofing linuxbios we should always start linuxbios at block 16 (i.e. at 8192 bytes). That should cover us for a very long time. I would like to hear what Ollie thinks first however. Ollie, can we just always start linuxbios at block 16? or is there some DoC in the works that will cause trouble? hardwaremain.c is my next thing to check out. I can't even figure out where a post of 0x30 comes from. ron
