Could somebody with an Alpha close at hand do me a big favor and test the attached patch? It moves the load address of the dynamic linker from the bogus fixed address "2L*MAXDSIZ" to whereever a userland mmap(0, ...) call would have put it. It should have no noticeable affect on your system, except maybe if you look at "/proc/<pid>/map".

The patch solves a problem for large-memory i386 configurations. I'd like to commit it ASAP, but it would be nice to confirm first that I didn't break Alphas.

Thanks in advance, John