BeOS port questions

[alanlb at access.rrinc.com]

On Sun, 17 Jan 1999, Nicholas Clarey wrote:

| Howdy all,
| 
| I'm currently working away on a BeOS port of kaffe. It's now building, but
| I fear that the hacks I used are going to cause problems. In fact, I would
| say they'll definitely cause problems.

Hello, Nicholas!

I'm also working on a BeOS port myself, and have made substantial progress
on a native thread subsystem.  I hope to release it for incorporation into
the CVS tree soon.

|  - Beos has no MSG_PEEK defined for the recvfrom command in
| PlainDatagramSocketImpl.c - how bad is this?

For now, I return an error from the relevant function, as my goal is to
stabilize the underlying layers first, then start building from that.  I
don't have an easy answer to this question yet.

|  - Beos has no register entries in it's struct sigcontext. This is causing
| the exception.c nullException and floatingException to die (no ebp/eip
| lines 301, 319). Is there a workaround for this?

There is, but it appears to be undocumented.  EXCEPTIONPROTO, at least for
R4, works out to:

	int sig, int code, struct vregs* ctx

I don't know if this can be relied upon past R4.

With minimal changes, the unix-jthreads system can be made to pass all of
the regression tests, except for the ones that require a working SIGVTALRM
and SIGIO (both of which are also absent on BeOS).  Without async
notification, it would take major surgery to make unix-jthreads work,
hence my work on a beos-native subsystem.

cheers,
Alan