Mr. de Vries,
std::variant looks like it should work splendidly. boost::variant might be even better. Thank you.
BTW: this is my first user group interaction; I hope I am responding correctly.
Neil Ferguson
-----Original Message-----
From: Maarten de Vries [mailto:maarten@de-vri.es]
Sent: Wednesday, November 21, 2018 1:06 PM
To: boost-users@lists.boost.org
Cc: Ferguson, Neil D CTR (US)
Subject: [Non-DoD Source] Re: [Boost-users] Interprocess: limited boost::any in shared memory
All active links contained in this email were disabled. Please verify the identity of the sender, and confirm the authenticity of all links contained within the message prior to copying and pasting the address to a Web browser.
________________________________
On Wed, 21 Nov 2018 at 20:58, Ferguson, Neil D CTR (US) via Boost-users > wrote:
Sirs and madams,
I hope to use boost interprocess shared memory to pass complex and dynamic data structures between processes on a host efficiently. The structures will be comprised only of simple primitives (int, float, etc.), interprocess basic_string, interprocess containers (vectors and maps), containers of containers, and (possible deal killer) boost::any. From the interprocess documentation, it _looks_ like the preceding is doable, aside from boost::any. The boost::any objects will themselves contain only primitives, basic_string, or interprocess vectors of unsigned char. So it _looks_ like the data held by the any should be okay. My question: does boost::any use pointers, references, or other constructs that preclude the class from working in shared memory?
Thanks for your advice.
Neil Ferguson
Take a look at /usr/include/boost/any.hpp . You will see that the actual content of the any object is a pointer created with `new` and cleaned up with `delete`. So this will not work properly in shared memory.
You could consider using std::variant instead, which does not put storage on the heap (unless contained objects do it themselves, of course).
Kind regards,
Maarten de Vries
