: 	Is this lossage likely to show up in arbitrary third party
   : 	software? Should -msoft-quad-float be the default on sparc64
   : 	until its fixed in gcc?
   
   There is an issue with missing support functions for some operations with
   -msoft-quad-float (as told to me by eeh), but I'm not sure what yet.  He
   said that otherwise, the ABI is compatible with "native" long doubles.
   
   It could be just some extra libgcc support that would need to be pulled in;
   I am researching it.  If this is possible, I do intend to turn on
   -msoft-quad-float by default on sparc64.



last i tried this something went BOOM big time.  maybe it was ld.elf_so,
but that seems unlikely....  i will note that i have previously suggested
turning this on by default, but eeh says not too.


.mrg.