329230 - Prefix "cache" in RDFServiceImpl::GetResource() can lead to mishandling of resource URIs

Status: RESOLVED INCOMPLETE

(Core Graveyard :: RDF, defect)

Description

[Matthew Gertner]

RDFServiceImpl::GetResource() uses the prefix (i.e. scheme) of the resource URI to identify a factory to instantiate the resource implementation. If no factory is found for that prefix, the default one is used. For optimization purposes, the method remembers the last prefix and factory used.

This goes wrong if a resource is instantiated with the URI "foo" (no prefix so the default factory is used) and then a foo:bar resource is instantiated, where the foo: prefix has its own factory. Because the URI "foo" is remembered erroneously as the last prefix, the default factory is used instead of the foo factory.

Comment 1

[Axel Hecht]

Sadly, we can't just error in onvalid URIs. Localstore and in particular,
XUL persistance, has it all over.

Pfft.

Comment 2

[Matthew Gertner]

Yeah, I guess the right thing to do is not remember the "last prefix" if the URI is invalid. I'll take the bug and propose a patch when I get a chance.

Comment 3

[Matthew Gertner]

Attachment: Patch that can be applied painlessly [Wrong bug]

This is a patch that can actually be applied to a normal Mozilla tree without the need for hand-tweaking (as was the case with the previous version). I also made a number of the changes Daniel asked for, mostly (but not exclusively) to do with formatting. I haven't changed the behavior (e.g. of minimizeAll) or added features (e.g. number of clicks to restore) since I still hold out the hope that someone who actually wants these will pick up the ball and implement them.

A much bigger problem with the current patch is that it still crashes when you right-click a few times on the tray icon. I spent a considerable amount of time investigating this, but I don't know anything about the handling of events and frames inside Mozilla, and the code is too complex to fix this or find a workaround without a significant time investment.

If someone knowledgeable in this area can help out (at least by giving me a nudge in the right direction) this would be greatly appreciated. Otherwise this patch is basically useless.

Comment 4

[Matthew Gertner]

Comment on attachment 226161 [details] [diff] [review]
Patch that can be applied painlessly
[Wrong bug]

Sorry wrong bug :-O

Comment 5

[Matthew Gertner]

Taking this off my todo list since we're no longer using Mozilla's RDF implementation.