On 24 Apr 2000, Marcus Sundberg wrote:
> Hi,
>
> as part of the move towards a new beta, and hopefully a real release
> a few weeks after that, the sublib handling in LibGGI has been
> rewritten. The log entry is included in this mail. If there's
> something unclear feel free to ask. Also feel free to ask if you are
> new to GGI and don't understand one bit of this, and I'll try to
> answer your questions and put them in some sort of FAQ.
>
> The single sublib entry points and unique names will also make it
> a pretty easy task to implement complete static linking of LibGGI:
> All you need to do is build a table of the entry points, and change
> _ggiLoadDL() to look them up in this table.
>
> ---- begin log entry ----
> Big rewrite of the sublib handling.
> All sublibs now have a single entry point prototyped like this:
> typedef int (ggifunc_dlentry)(int func, void **funcptr);
>
> This function is used to get pointers to the real sublib functions.
> Currently there are three such functions:
>
> typedef int (ggifunc_open)(struct ggi_visual *vis, struct ggi_dlhandle *dlh,
> const char *arg, void *argptr, uint32 *dlret);
> typedef int (ggifunc_exit)(struct ggi_visual *vis,
> struct ggi_dlhandle *dlh);
> typedef int (ggifunc_close)(struct ggi_visual *vis,
> struct ggi_dlhandle *dlh);
> open replaces GGIdlinit() and close replaces GGIdlcleanup(). exit is a
> special function which will rarely be used. It is used when the normal
> de-initialization order of sublibs is not sufficient, for example when
> an OP_DISPLAY sublib needs to access data in an extension sublib (like
> in the mansync case).
>
> open (and exit/close as well) now returns 0 on success and a GGI error
> code on error. The GGI_DL_* codes are returned in the *dlret argument.
>
> The name of the entry point is <symprefix><sublibname>, where:
> <sublibname> is the name of the library with the last dot and everything
> after it stripped.
> <symprefix> is GGIdl_ for LibGGI and the Misc extension.
> For all other extensions the new symprefix paramater from
> ggiExtensionLoadDL() is used for <symprefix>.
>
> _ggiOpenDL() and _ggiAddDL() has been changed to return 0 on success and
> a GGI error code on failure.
>
> _ggiZapMode() no longer takes a mask of sublib types to zap, instead it
> takes a 'zapall' argument which if non-zero tells LibGGI to zap extension
> sublibs and OP_DISPLAY sublibs. Otherwise only OP_GC, OP_DRAW and OP_COLOR
> sublibs are zapped.
>
> The ggi_visual_op* structs has been cleaned up so we can initialize all of
> them with a single function, so despite the above changes LibGGI is now
> actually slightly smaller.
>
> This does not require any changes to applications, full ABI compability
> is maintained. Extensions are affected as ggiExtensionLoadDL() now takes
> an additional symprefix argument, and all extension sublibs needs to be
> changed for the single entry point system.
> ---- end log entry ----
Are the sublibs changed too? Maybe that John have to rewrite his
directx-target...
When do you commit this changes into the cvs-tree?
I want to test it, _before_ the beta will be released!
I think, the most important thing is, that libggi is getting smaller. The
smaller it is, the less mistakes we can make (= less bugs,
more effecient maintainable)...
At least, it is a good work, Marcus.
CU,
Christoph Egger
E-Mail: [EMAIL PROTECTED]
