John, thanks for your answer. Unfortunately, that's exactly what I was afraid of: The mechanisms is here (XPCOM and the ServiceManager) but it doesn't seem to be used fully nor consistantly. I've tried the plugin newsgroup, checked and rechecked the docs, searched the code and samples, and debugged/stepped thru the internals of the ServiceManager (comparing the data (instanciated interface pointers) stored in its hashtable against pointers of objects used by the browser). Nothing does it, there does not even appear to be any high level components that I could a hold of and drill down the hierarchy to get access to those of interest to me. I am going to try the xpfe and embedded newsgroup like you suggest, those are my last chances before having to hack in the sources... :-( - Denis "John Bandhauer" <[EMAIL PROTECTED]> wrote in message [EMAIL PROTECTED]">news:[EMAIL PROTECTED]... > I don't know much about scrollableviews and the viewmanager and > all that. I'd think the place to ask about that would be the xpfe > newsgroup. > > The xpcom service manager is just a cache of the services that > were created through the service manager. Any given component > might be instantiated in some other way (even if it is used as a > singleton by other code) and not use the same instance as the one > you might get from the servicemanager. xpcom does not yet have a > good way to declare which components even *should* be singletons > and/or instantiated via the servicemanager. > > I'd expect that views would be independent and that you'd either > have to surf from some interface available to the plugin api > directly or get a very high level interface and drill down until > you find your window. But, the fact is that I don't know. > > I'd suggest digging more in the code and docs (such as they are). > Maybe http://www.mozilla.org/xpfe/ > Maybe ask in the xpfe, plugin, and embedding groups. > Maybe dig deeper into the embedding samples to see how that part > of the architecture works and what you might be able to get at > from a plugin. But, note that the frozen parts are few and if you > use "whatever you can find" then you are using internals that are > subject to change. > > Hope this helps. If I'd known the real answer to your question > I'd have replied quicker :) > > John. > > Denis wrote: > > > > Hi, > > the Moz. plugin that I'm working on needs to know what the current > > scrollbars positions are, and I figured one way of doing this (is there an > > easier way?) would be to get access to an instance of one of the classes in > > Mozilla that I know have some knowledge about the scrollbars. > > In simpler terms, I want my plugin to have access to the instance of the > > ScrollableView interface that the browser is already using. > > My understanding of XPCOM is that making a call to: > > nsIServiceManager::GetService(..nsIScrollableView::GetIID()...) > > should return to me either a new instance of the ScrollableView interface if > > none were already registered, or an already existing one if it was > > registered. It appears however that I am always getting a new instance, as > > opposed to the one that the browser is already using. > > > > I've tried saving the ServiceManager (say sm) that is passed to the > > NSRegisterSelf function (these are needed in a plugin to allow it to self > > register I believe), and later on use it as: > > sm->GetService( ...nsIScrollableView::GetIID()..) > > but I get the same results: an new instance of a ScrollableView, as opposed > > to the one that the browser is already using. > > > > Most of the documentation and samples that I looked at usually talk about > > extending Mozilla in one direction: how to write a component that registers > > its interface and services with the idea that you can later on, from within > > Moz. code, call and access these services. But I am interested in doing > > things in the other direction: write a component that can get access to the > > (already instanciated) services of Mozilla... > > > > I have a feelling I might not be looking at the problem from the right > > angle, and that the solution is pretty simple, so I'll take any suggestion > > and comments! > > > > Denis > > > > -------- > > [EMAIL PROTECTED]
