Re: Module support: No environment in pointer release function

[Philipp Stephani]

Elias Mårtenson <address@hidden> schrieb am Di., 7. Feb. 2017 um 04:16 Uhr:

On 6 February 2017 at 13:33, Paul Eggert <address@hidden> wrote:

Elias Mårtenson wrote:

In
addition to the pointer itself, the ‘emacs_env’ pointer is needed. An
arbitrary ‘void *data’ pointer would be useful too.

Can you package up these pointers into a single data structure, and pass a pointer to that structure? That's what's usually done in situations like this, and is why a single 'void *' should suffice.

I could do this, but as far as I understood, the emacs_env pointer is not guaranteed to be immutable. If it isn't immutable, then I might just as well save it to a global variable and not have to pass it through every function that needs it.

What are the guarantees with regards to the emacs_env value?

The big restriction is that its lifetime is limited to the lifetime (storage duration) of the argument that is used to obtain the pointer. I.e. in a function

emacs_value module_fun(emacs_env* env, ...) { ... }

the environment represented by env is only valud in the body of module_fun; reusing env or *env once the body is finished is undefined behavior.

(I'm aware that this is not documented at all, and I've been meaning to write some documentation for a while.)