Sounds good. I'll start experimenting with DYNINSTuserMessage to see
how it works for me. And I'd still appreciate suggestions on the best
way to send such messages in reverse, mutator->mutatee.
If you just need to push data, just write memory. Either allocate (BPatch_addressSpace::malloc) or look up an existing variable, and use writeValue on the resulting BPatch_variableExpr.
If you need to run code, use a oneTimeCode. It'll be slower than writeValue, though.
Or see if the sharedMem component in the valueAdded subdirectory works for you.
I'm somewhat interested in the dyninst_lock_t too, so I can avoid
drawing in pthreads, but I'm not thrilled that tc_lock_lock() uses a
busy loop. Would you be open to an implementation on futexes for Linux?
Sure thing. Though it's likely you'll already have pthreads anyway; as I recall, gcc loves pulling that in to everything.
![[Computer Systems Lab]](http://www.cs.wisc.edu/pics/csl_logo.gif){kind=logo}