[gmpi] Re: ABI's / Nutshell

Hi Chris,
Good, but needs tweak...

> One of the function pointers points to a constructor
> function,

Not exactly. The factory contructs the plugin when you pass it the ID. The factory passes back a pointer to the object. The first 4 bytes of the object always point to the function table. So you have the two things you need:
-the instance pointer
-the fuction table

> This bundle does not necessarily conform to any
> pre-existing platform or compiler or component architecture calling
> conventions.

Well, the function pointers *must* comply with the ANSI C standard calling convention.

> To decommission a plug instance and free its memory, the host calls the
> package file's anti-factory function.

No, the object's function table holds the destruction function (not the factory). (At present the 3rd function pointer.)

Best Regards,
Jeff

Chris Grigg wrote: 
Tim wrote:

On Wed, Feb 23, 2005 at 12:13:45PM -0800, Chris Grigg wrote:

 >The pointer could be
 >called "pointer_to_my_functions", but at present it's called
 >"vtable".

 I like the design, but this naming, I humbly suggest, is a mistake
 because of the possibility of confusion with the C++ vtable.



Ok, let's dump the word vtable from the GMPI vernacular.  How about
"ftable" or "fntable" (function table) or "mtable" (method table?) or ..?

Ideas?


This would be a good time to do some terminology clarification in general. Not quite sure how to contribute to that, as I'm not sure I'm totally following how packaging and lifecycle work, so let me take a crack at a summary, avoiding language-isms and platform-isms:

***
Plug Packaging and Life Cycle

Every GMPI plug product (instantiable plug class) in the world is identified by GUID. Plug products are delivered in a designated package file format allowing one or more plug products per package. To discover package files, the host inspects a designated folder in the file system (path is platform dependent). To discover the plug products contained in a package file, the host reads the package manifest (metadata).

Every package file has exactly one factory function, whose offset into the package file is itself found at a known offset. To instantiate a plug product, the host loads the package file into memory and executes its factory function, passing in the GUID of the desired plug product; factory allocates and initializes any static storage for that plug product; upon return the factory provides the host with a bundle of function pointers representing the GMPI plug API for that instance. The host will make all calls into the plug product instance exclusively via these function pointers. This bundle does not necessarily conform to any pre-existing platform or compiler or component architecture calling conventions. One of the function pointers points to a constructor function, which calls back to the host to allocate storage for the plug product instance and then initializes it. The host is able to create any number of simultaneous instances of a given plug product, though this is subject to any limitations the author of the plug product (or the package's factory function) may choose to impose.

Once a plug product instance has been constructed, host makes all normal operating calls using the bundle of function pointers; beautiful or ugly music ensues according to taste and talent.

To decommission a plug instance and free its memory, the host calls the package file's anti-factory function (again via an offset at a known offset). When the refcount for the corresponding plug product reaches zero, the anti-factory also frees any static memory used by the plug product.

***
a) What did I get wrong in the above?

b) How exactly does the factory, or a plug, call back into the host for services like memory, files, etc.? Doesn't the factory need a bundle of host callback function pointers too?

c) Is the plug prohibited from calling anything outside itself (other than the host-provided services callbacks)? If not, since we aren't using a dynamic linker to load plugs, how exactly are those symbols (system lib calls etc.) resolved by the time the plug needs to call them?

    -- Chris G.



----------------------------------------------------------------------
Generalized Music Plugin Interface (GMPI) public discussion list
Participation in this list is contingent upon your abiding by the
following rules:  Please stay on topic.  You are responsible for your own
words.  Please respect your fellow subscribers.  Please do not
redistribute anyone else's words without their permission.

Archive: http://www.freelists.org/archives/gmpi
Email gmpi-request@xxxxxxxxxxxxx w/ subject "unsubscribe" to unsubscribe




----------------------------------------------------------------------
Generalized Music Plugin Interface (GMPI) public discussion list
Participation in this list is contingent upon your abiding by the
following rules:  Please stay on topic.  You are responsible for your own
words.  Please respect your fellow subscribers.  Please do not
redistribute anyone else's words without their permission.

Archive: http://www.freelists.org/archives/gmpi
Email gmpi-request@xxxxxxxxxxxxx w/ subject "unsubscribe" to unsubscribe

Other related posts: