Re: Remove data race in libgcj interpreter

[David Daney <ddaney at avtrex dot com>]

Andrew Haley wrote:
> David Daney wrote:
> > Andrew Haley wrote:
> > > +
> > > + #ifdef DIRECT_THREADED
> > > +     CHECK_INSN (&&insn_instanceof);
> > > + #endif /* DIRECT_THREADED */
> > > +
> > Just a little style pedantry...
> >
> > Should we re-write the CHECK_INSN macro so that it is a nop ifndef
> > DIRECT_THREADED?
> >
> > That would make the many places you have this construct look much cleaner.
>
> OK, but would it be easier to understand when reading the code?  That's
> why I didn't do that.  Sometimes clarity and cleanliness don't go
> together.  What do you think?

For me there are two issues:

1) Mutex free synchronization scares me.  But That part of the patch 
seems plausible.

2) Of much less importance is the question of littering the code with 
#ifdef.  I stated my preference, but don't feel strongly enough about it 
that I would be offended if your original patch were committed.

David Daney