Yes: Racket needs a consistent starting point for the GC'ed stack across calls into the Racket.
At Mon, 24 Sep 2012 17:32:26 -0700, Dan Liebgold wrote: > Is there any reason the scheme_main_setup trampoline can't be done later in > the program (i.e. only at the point I want to invoke Racket stuff)? > > On Mon, Sep 24, 2012 at 5:09 PM, Matthew Flatt <mfl...@cs.utah.edu> wrote: > > > At Mon, 24 Sep 2012 15:10:08 -0700, Dan Liebgold wrote: > > > Does anyone have an example of embedding 3m Racket (5.3) in a C++ program > > > without overriding main? I have a legacy C++ app that wants to eval some > > > Racket code but I'd like to be minimally invasive. Might need to shell > > > out... > > > > One strategy is to start a separate OS-level thread for Racket. The > > rest of the C++ program would send expressions to the Racket thread > > evaluation via a pipe or some kind of semaphore-based communication. > > > > On Windows, you could also try using MzCOM, which is sort of half-way > > between managing your own thread and starting a separate process. > > > > -- > Dan Liebgold [dan.liebg...@gmail.com] _________________________ Racket Developers list: http://lists.racket-lang.org/dev
