Re: --enable-shared doesn't work on Sun-3

[Jeffrey A Law <law at cygnus dot com>]

  In message <9808101756.AA19115@moene.indiv.nluug.nl>you write:
  > >  And it's accepted?  In that case, NeXT must have hacked
  > >  the assembler.
  > 
  > No, that's why I complained about the (seemingly automagically  
  > generated) --enable-shared build on m68k-next-nextstep3 in May this  
  > year.
  > 
  > So, on my system, this code is generated in spite of the fact that:
  > 
  > a. My assembler doesn't know how to deal with it (because it's gas).
  > b. It's a Sun trick, which should mean it shouldn't be generated
  >    *at all* on a m68k-next-nextstep3 system.
At the time when that code was written, I don't believe any standard
had emerged for how to handle PIC on m68ks (we're talking about 1990/1991
if I remember right).  By the time the dust settled my interest in
m68ks was nil.

Basically, if the target doesn't specify that it wants what became
svr4 style PIC or some other PIC variant, then the m68k port falls
back to the StunOS style PIC.

At one time gas ignored the extra argument, but presumably that went
away over time as the m68k parser became more strict.

  > Then it might be advisable to have egcs *not* generate this code,  
  > except for a m68k-sun-* system, right ? (And generate an error  
  > message in case someone specifies -fpic on the command line, because  
  > it can't be dealt with anyhow ;-).
Feel free to hack something up.   You'd probably want to have one of
the sunos config files (or configure itself) define a macro to
enable/disable tacking on the ",a1" argument.

jeff