On Mon, Jan 31, 2011 at 9:32 AM, nathan binkert <n...@binkert.org> wrote:
> >> Are you ok with this original change and renaming files later? > > I'm fine with that. > Me too. > > > It looks to me like fault.hh is basically just forward declarations while > > faults.hh has the actual definitions, right? Do we have any other pairs > of > > files that are split this way? Usually we don't bother having a file > with > > just forward declarations since you can just type "class Foo;" directly, > but > > the RefCountingPtr typedefs and (soon) the NoFault declaration make this > a > > case where there is value. > > I'd be in favor of file names that actually try to reflect that > distinction, > > like fault_decls.hh and fault_defs.hh. I totally agree with Nate that > the > > current scheme is not really meaningful at all and thus should be > changed. > > The one precedent I can think of is iosfwd in the STL. I think fwd is > a bit more descriptive if it really is a forward declaration since in > many cases you will want a file with some forward declarations and the > other file doesn't really have definitions necessarily. I also think > there's no reason to distinguish the "regular" file by some special > naming. So, my preference would be foo_fwd.hh and foo.hh. > > It probably wouldn't be a bad idea for us to do more _fwd.hh files for > some of the more complicated templates that we use. > That all sounds reasonable to me. I meant to add in my review that the files should have file-level comments distinguishing them as well. Steve
_______________________________________________ m5-dev mailing list m5-dev@m5sim.org http://m5sim.org/mailman/listinfo/m5-dev
