On 3/22/10, Mark Cave-Ayland mark.cave-ayland@siriusit.co.uk wrote:
Blue Swirl wrote:
Okay - in that case I'll rip it out. Just as a side question, what
could I
use to test that I haven't broken the ELF code in QEMU? I'm guessing I
could
test using something like the FC12 bootable CD for each platform (x86,
PPC
and SPARC64) to check?
I think SILO and BSD loaders are still using a.out and Solaris uses FCode.
Right - I now have an experimental patch available which moves all of the loaders except linux_load.c into libopenbios/ and refactors everything else accordingly. I suspect if other people were to help out, we could do a similar thing with the Linux loader too as the overall code looks similar.
The good parts of the patch are that it consolidates a lot of code that was split across multiple arch/ directories into one place and so a lot of duplicate code can be ripped out. This also has the advantage of being able to offer several loaders across multiple architectures very easily.
The bad news is that while I can compile across the entire range of platforms, there are still some remaining issues. For example, my SPARC64 Milax ISO test now crashes which will require some digging. However, at the moment it's quite a large patch and part of me feels I should commit what I've got so that other people can get eyeballs on this and help out.
Anyone feel strongly either way? I've attached what I have so far.
Bad commits aren't very nice. Perhaps you should split the patche into more digestable pieces? For example the implicit vs. explicit file descriptor change should be safe.