Re: [PATCH] libkdeinit symbol pruning
David Faure <[email protected]>
| Newsgroups | gmane.comp.kde.devel.optimize |
|---|---|
| Organization | KDE |
| Message-ID | <[email protected]> |
On Monday 19 July 2004 13:04, Waldo Bastian wrote: > On Monday 19 July 2004 02:05, David Faure wrote: > > On Monday 19 July 2004 01:06, Karl Vogel wrote: > > > It will break apps that link to the libkdeinit_*.so's or dlopen them and > > > try to resolve symbols other than kdemain... but I guess if any app does > > > that, then it's already broken by violating the kde abi, no?! > > > > Thinking about it, I'm not sure anyone links to a kdeinit module. > > What people do, which is unportable, is to link to a kparts part, but > > that's unrelated. > > What if an application defines a plugin API that plugins can call? (Or worse, > if the application doesn't define an API and plugins just call stuff all over > the place) You are right, this happens: with kded :/ kded modules find symbols in the kded binary itself (so now that I have the kdeinit symbol pruning patch applied, kwallet and kpasswdserver don't load anymore). This always seemed dirty to me - shouldn't we have a libkded instead? -- David Faure, [email protected], sponsored by Trolltech to work on KDE, Konqueror (http://www.konqueror.org), and KOffice (http://www.koffice.org).