> Anyone interested in adding a command line option to gcc for this?

Try "-mlong-calls" and see if that is recognized.

Various gcc ports have been introducing this switch, and in at least one
case it took a while before someone remembered to document it.

As for the portability of the interface: we are basically screwed here
because many RISC CPUs don't see fit to provide a full-range direct
call instruction. CISCs generally do, and sparc is one of the few RISC
chips to also do it. (64-bitters avoid the issue by using literal pool
entries to get the addresses out of the instruction stream entirely...)

Every so often the subject of gcc dependence comes up and usually what I
see people saying is that most likely we do depend on gcc already, and
it's not really practical to spend time preventing this for philosophical
purity's sake, as gcc is so darn ubiquitous.

I'd say the question is not how to fix this portably, but rather how to
minimize the part of the fix that is not portable.

Todd Whitesel
toddpw @ best.com