Ship It!

src/sim/syscallreturn.hh (Diff revision 1)

If you're cleaning this up, why bother to continue storing success?

Show all issues

1. Steve Reinhardt April 22, 2014, 10:59 a.m. (April 22, 2014, 10:59 a.m.)

   On some platforms, the syscall return uses multiple registers, so it's theoretically possible to have a successful syscall that has a return value in the 0 > retval > -4096 range.  Having the separate flag lets us represent that.  The old code had an additional constructor that took the success flag explicitly, so we could actually create such instances (not just represent them), but it was unused so I got rid of it.  I could add that constructor back just to document the possibility, if you think that would be helpful.

2. Nathan Binkert April 22, 2014, 12:57 p.m. (April 22, 2014, 12:57 p.m.)

   I'd buy the argument if the variable wasn't private, but because it is, it's simply an implementation that has the possibility of being extended in a certain way, but is not currently that way.   I'm surprised that you'd even consider adding the constructor back. What happened to Professor Steve "Don't overgeneralize your APIs" Reinhardt? :)
   
   Anyway, too many words typed about this topic in general.  Certainly not worth the length of discussion we've had already.  It's certainly not terrible.

src/sim/syscallreturn.hh (Diff revision 1)

I suppose all these methods could be const...while you're at it :-)

Show all issues

1. Steve Reinhardt April 22, 2014, 11:01 a.m. (April 22, 2014, 11:01 a.m.)

   Good point, Mr. Const Enforcer :-).  I'll do that.