* Specify page titles as strings instead of split NS/DBK, as suggested by Roan on CR. It seemed sensible to me.
* Pass a Title object to getContent() instead of a string, to avoid unnecessary object construction overhead
* "*" and "/" are valid title characters. Check module input for JS comment end tokens.
* Fixed inappropriate conversion to boolean, when checking result of getContent(). Presumably the idea was to omit empty sections and errors, so that's what I did. Maybe an informative error message would be better in the error case.
* Use LinkBatch for selecting multiple page rows instead of Database::makeWhereFrom2d().
* Fixed assignment expression.
The CSS for the loading tab had:
background-position: center 60%;
The 'center' apparently worked for other browsers, but in IE7 it was getting a bit off, and threw the image over a few pixels.
'center' keyword here is allegedly equivalent to 50% which I also tried, but with the same results.
Instead, using the same fixed position that we specify for the regular rest-state icon works:
background-position: 5px 60%;
Resulting looks ok for me in:
* IE 6 / XP
* IE 7 / XP (fixes regression)
* IE 8 / XP
* Firefox 4b10 / Linux
Shouldn't hurt anything else.
It's not ideal, but at least it's no worse than it was in 1.16. It's preferable because the content area is more likely to have bidirectional text than the UI area. The UI is split up into short strings, generally in a consistent language, so there is less chance of the UBA moving fragments of text around.
Ideally, the direction of each div would be set according to the language of the text it contains.
Fixes:
- now accepts file titles with or without File: prefix instead of demanding you must remove it
- moved result summary line from bottom to top
- added a line telling you no file was found if it wasn't found
- now pulls the reference file's SHA-1 via FileRepo, so can give it a remote file (eg from Commons)
- now pulls duplicate files via the main RepoGroup instead of querying image table manually, so turns up remote duplicates
- dropped the QueryPage standard paging header/footer; file dupe lists are usually very short and it's not worth copying the infrastructure
To make this work, I switched the special page class from using the standard QueryPage paths to doing the query and list itself; QueryPage is currently very tightly tied in to database queries, and doesn't provide a very clean way to drop in an alternative way of looking stuff up (say an API query or just getting a big array you've gotten from somewhere). If that gets improved, this page should be cleaned up to use more of the QueryPage infrastructure again so it's pretty and doesn't scare people coming in to maintain it.
Localization changes:
- added fileduplicatesearch-noresults message
- changed fileduplicatesearch-summary in English master to remove the supplementary line about taking out the 'File:' prefix
