> 
> The existing pmax code (such as locore.s) appears to be written to
> follow the MIPS toolset convention, which is for the C compiler to not
> prepend an underscore.  From looking at GCC configuration files,
> 4.4SBD seems to follow the same convention as the mips toolset. 
> 

We are not bound by the behavior of the mips toolset.

> At least that's what "mipsbd" targets for GCC do.  Is that really
> right for 4.4bsd?? (If so, how does netstat find kernel symbols??)
> 

My impression has been for a while that the 4.4bsd pmax port was
basically built with the mips tools, or GCC stuff configured to look
like them.

> It looks like it'd be less work to just conditionalize the few
> programs that really need to know about the external format of names
> -- ps and netstat are all I've run across so far -- and to follow the
> naming convention used by gcc for "mipsbsd" targets.  My
> rationalization is "if it's good enough for CSRG it's good enough for
> NetBSD".
> 

Frankly, I don't use the "good enough for CSRG" argument anymore.
They did some good work, and some bad hacks.  I evaluate all of their
stuff carefully.

If all the other netbsd targets are configured to do prepend _, I see
no reason why the pmax port should be an exception.  It would have to
be explictly conditionalized against......why bother.

> Are there any strong opinions the other way??
> 
> --Jonathan

Yes...see above.

I'm quite sure there is some option/#define to gcc to make it do the
right thing.