On Fri, 29 Jan 1999 18:57:16 -0800 
 Jonathan Stone <jonathan@DSG.Stanford.EDU> wrote:

 > Longer-term, I'd prefer to tweak genassym.sh to make -D_KERNEL
 > configurable from the command line. Then we can generate
 > machine-generate assembler #defines for both both 32-bit and 64-bit
 > definitions.

nonono... Things in header files that userland includes must be
statically declared in header files checked into CVS.

 > Any objections to generating one machine-generated file and pulling it
 > in, or do we need three (one each for setjmp.h stuff, signal.h stuf,
 > asm.h stuff)?

I thought we discussed this, and there were strong objections based on
the fact that it's too easy to break the ABI that way!

 > Also, do you remember why _JBLEN got bumped from 83 to 87?
 > Did struct sigcontext grow by 16 bytes?

Um, yes!  The new signal masks!

----------------------------
revision 1.2
date: 1998/09/16 23:15:08;  author: thorpej;  state: Exp;  lines: +2 -2
Need 87 longs for a jmp_buf now (we use sigcontext, which grew).
----------------------------

 > Last, is there a list of changes pending on bumping the libc
 > major-version?  It'd make a lot of things a lot cleaner to just always
 > allocate jmp_bufs/sigcontexts large enough for the 64-bit clean
 > userland, even for normal ILP32.

The different reg size (i.e. 64-bit clean userland and all that) should
be handled via a different emul.

        -- Jason R. Thorpe <thorpej@nas.nasa.gov>