Last modified: 2014-09-22 10:33:26 UTC
I was used to search for files placed manually in license cats to spot errors in various bots or image uploads, now the new search is absolutely useless for this as it pumps out all images from these cats with 99.9% of them having the cat transcluded from the license template. I used the search linked below which worked well with the previous search engine: https://commons.wikimedia.org/w/index.php?title=Special%3ASearch&profile=advanced&search=incategory%3A%22GFDL%22&fulltext=Search&ns6=1&redirs=1&profile=advanced
The behavior you are describing is exactly as designed, unfortunately. While the query is much more complicated, you can still get what you were doing: incategory:"GFDL" insource:"Category:GFDL" insource:/\[\[Category:GFDL\]\]/ The <<incategory:"GFDL">> clause is exactly what you were using. The <<insource:"Category:GFDL">> clause limits the results to those that include category:GFDL in the source but has a high false positive rate because it'll find stuff like {{category|GFDL}} as it is using standard full text analysis. The <<insource:/\[\[Category:GFDL\]\]/>> clause limits the results to just those that contain exactly [[Category:GFDL]] in the source. Its a slow check but when you combine it with the other two clauses its always executed last - so the entire thing comes back reasonably quickly.