Re: Your Jan 15 change broke the x86

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

  In message <19990131134018.A17516@cygnus.com>you write:
  > I've not seen an actual machine with offsettable autoincs.
I'm not aware of one either.  But stranger things have happened :-)

  > Ok, so what?  How does this apply to the fact that the user has
  > little or no control over how and whether autoincs are generated?
  > Or are they supposed to simply write
  > 
  > 	asm("" : "=<"(*++x) : ">"(*y++))
  > 
  > and hope?
Right, the user has little control over whether or not an asm is generated,
and IMHO, an autoinc constraint by itself in an asm is just asking for trouble.

Should we reject it at expansion time or in reload?

I'd say expansion time, but I worry that we'd start balking at asms that have
worked for years.  Then again, it may be the case that nobody uses the <>
constraints in an asm without either an "o" or "m".



jeff