The suggestion of unlisted special pages confuses users.
This is similar to bug 20868: Search suggest shouldn't offer special pages with limited access to users who don't have such rights.
Version: unspecified
Severity: normal
The suggestion of unlisted special pages confuses users.
This is similar to bug 20868: Search suggest shouldn't offer special pages with limited access to users who don't have such rights.
Version: unspecified
Severity: normal
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Open | Feature | None | T29411 Search Suggest bugs (tracking) | ||
Open | None | T27399 Search suggest shouldn't offer unlisted special pages |
@Raymond Can you give an example of this? We need the search query you entered and the results that were displayed to reproduce this.
@Deskana Blame me that I forgot to write down same examples. Now here from core only:
I'm really not sure what to do with this, honestly. It's not really wrong for search to return these pages, it's just that these pages don't give you a nice form when you navigate to them. This seems like correct behaviour to me.
Maybe we could instantiate the class and check isListed() function for it? I'm not sure though it's the right thing. After all, if you search inside Special: space, you must realize these pages are, well, special :) So you expect to get page-specific result when you go there, including maybe something that doesn't make sense in current context. After all, maybe you do want to find how that blank page is called, and that's what you search for.
OTOH, if there are permission issues, I'm not so sure. The problem is permissions are checked in execute() code of the page, for each page by its own, so search has no way which permissions are associated with specific page. Unless we refactor SpecialPage to have separate callable permission check, I don't think it's possible for the search to know if the user has permission to access the page or not.
This seems like a lot of work for not such a large gain, and as mentioned above I'm not even convinced the behaviour that we're seeing here is incorrect. I can't really prioritise this highly.