Tragically, this function is stubbed. http://lxr.mozilla.org/seamonkey/source/rdf/chrome/src/nsChromeRegistry.cpp#2393 I have code to un-stub it rotting in my tree. I'd sort of like to land bug 70966 first, but it's not essential.

It's possible that this doesn't do all the things it needs to do, but it's sufficient to clean chrome.rdf and any overlays.

Targeting aggressively -- I'd really like to get this in 0.9.7 because of its use to 3rd-party developers.

who was I kidding? -> 0.9.8

This bug is targetted for 0.9.8, and should be moved. Any chance of getting r= / sr= on the proposed patch ?

retargeting.

I have no experince with this code, but the patch looks fine to me :) Should we add helpwanted and nsbeta1 to this ? I'm sure NS wants this functionality. Anyone care to review the patch ?

I have no idea if this patch still applies cleanly (I'll check). My previous efforts to find reviews for this code were completely unsuccessful. Once I check the patch, I'll make another attempt. I'm not overly optimistic.

Perhaps brendan can do a catch-all review on this, if waterson (or some RDF expert) sr=s. Let's get some traction on this patch.

Comment on attachment 61092 [details] [diff] [review] basic implementation >Index: nsChromeRegistry.cpp >+ nsCOMPtr<nsIRDFDataSource> dataSource; >+ rv = nsComponentManager::CreateInstance(kRDFXMLDataSourceCID, >+ nsnull, >+ NS_GET_IID(nsIRDFDataSource), >+ (void**) getter_AddRefs(dataSource)); Use do_CreateInstance instead? >+ if (NS_FAILED(rv)) return rv; Verify dataSource != 0 instead? >+ nsCOMPtr<nsIRDFRemoteDataSource> remote = do_QueryInterface(dataSource, &rv); >+ if (NS_FAILED(rv)) return rv; Verify remove != 0 instead? >+ // Synchronously load contents.rdf. >+ // XXX Ignore old-style manifest.rdf. Packages that are old enough to still >+ // be using it shouldn't be calling this anyway (or at least shouldn't be >+ // expecting it to work). >+ nsCAutoString pkgName; pkgName.AssignWithConversion(aPackageName); >+ nsCAutoString key(NS_LITERAL_CSTRING("chrome://") + pkgName + >+ NS_LITERAL_CSTRING("/content/contents.rdf")); >+ remote->Init(key.get()); >+ remote->Refresh(PR_TRUE); >+ >+ // Uninstall dynamic overlays >+ rv = UpdateDynamicDataSources(dataSource, PR_TRUE, aUseProfile, PR_TRUE); >+ if (NS_FAILED(rv)) return rv; >+ >+ // Remove from all-packages.rdf >+ rv = UninstallProvider(nsCAutoString("package"), aPackageName, >+ aUseProfile); >+ if (NS_FAILED(rv)) return rv; >+ >+ return NS_OK; I'm assuming that we write the datasource back out at this point because there are no live references left to it? Might be worth a comment if that's the case...

Joy, this couldn't be coming at a better time.

mozdev is acting freaky so I have not tested this yet to verify that it still works, but I'm posting this anyway to strike a blow against my own horrid laziness. The CreateInstance() call was to match the local dialect -- shall I go ahead and convert the file over to do_CreateInstance? (There's a lot of stuff in the chrome code that's not quite cutting edge stylistically.)

tever is not RDF QA anymore

Apparently this code has ceased to exist.