Yeah, that's what I'm looking at now.  For the most part, it seems pretty 
straightforward, except that I wouldn't mind knowing wtf an upcall is (google 
wasn't terribly helpful without some additional context), and the comments in 
the existing implementations seem to alternate between talking in top-down 
terms and "stack as conforms to gravity" terms, so I'm still not quite sure 
where all the structures are, and I'm not sure if this code's running in a 
trap handler (meaning saving context would involve shadow registers) or just 
normal userland code.  I'm reading nathanw's paper on the topic in hopes of 
clarifying parts of this, and then I'll probably just program a bunch of 
stuff incorrectly and break things to answer the other questions ;)

TTFN

Reid

On June 9, 2004 04:36 pm, Christos Zoulas wrote:
> In article <20040609221031.241e996d.jkunz@unixag-kl.fh-kl.de>,
>
> Jochen Kunz <jkunz@unixag-kl.fh-kl.de> wrote:
> >On Wed, 9 Jun 2004 12:50:13 -0600
> >
> >Reid Orsten <rorsten@tartarus.ca> wrote:
> >> In the __sigreturn14 thing, with a bit of reading I could do the FPU
> >> context saving stuff (unless there's a big issue that I missed), but I
> >> don't quite get the problem with the flag checking thing, mostly due
> >> to lack of experience with this level of the system.  Are there any
> >> decent docs that I could check out to fix this?
> >
> >Please let me know when you can find any docs. I am more or less exately
> >as blind as you in that corner of my eye. :-(
> >It looks like you need some pthread related syscalls for fixing the
> >__sigreturn14 issue. So the pthread MD bits may be a prerequirement.
>
> You just need to implement the machine portions of get and set context.
> This should be simple enough. You don't need floating point context
> to at first so you can let that slide by...
>
> christos