From: Richard M Kreuter <kreuter@pr...> - 2007-12-11 16:53:59
Juho Snellman writes:
> rjs@... writes:
> > jsnell@... wrote:
> > > If does not, the first OS dependency there that I can think of
> > > would be signal contexts. So you could check that the contents of
> > > registers and the stack are being pinned correctly.
> > There doesn't seem to be a lot of OS dependent code to break, but I
> > will poke around a bit.
> There are some hidden dependencies. E.g. there's an assumption that
> the signal context will be wholly allocated on the stack (not true on
> OS X, so there's a separate hack for that).
(The following might be a red herring.)
A friend of mine started a port to NetBSD/amd64 last month. He told me
he found that NetBSD/amd64 stores the signal context immediately past
the faulting routine's frame on the stack, and so our code that writes a
new frame to call call_into_lisp is scribbling on the context. (If this
is happening, it might only affect the path through the handler taken
for stray segfaults not related to memory management or stack
exhaustion, and so might not affect the build process, however.)
Get latest updates about Open Source Projects, Conferences and News.