Re: milo question [OT: L4-Alpha]
Nikita Schmidt <cetus@cnds.ucd.ie> writes:
> On Tuesday, 6 Nov, Goswin Brederlow wrote:
> > UNZIP: this is not a zipped file (which is true)
> > #######################################...
> > Oops: trying to close a non-open file!! (is that because of FAT?)
> > Expected 1, not 4 program headers
>
> MILO has its own ELF handling code (not using libelf or BFD), which, I
> suppose, is the bare minimum sufficient to load Linux kernels (which have
> just one program header). I don't know why L4's image needs 4 headers;
> you could try stripping them with objstrip and then building a new ELF
> image with objcopy/ld - or something like that.
>
> Also, I remember that MILOs had the habit of ignoring load address in the
> ELF header, always loading at 310000 physical, fffffc0000310000 virtual.
> (I don't know if L4 cares, though.)
Thats a problem too, but one thats easily changed in the milo source.
But I hink it might be another good idea to add compression to L4
using the linux compression stuff. That would produce an elf image
with 1 programm header that reloactes itself.
Is the uncompress stuf position independant?
On i386 I know it cheks for its position and the final position and
moves itself out of the way first when they overlap. Does i do the
same on alpha?
May the Source be with you.
Goswin
Reply to: