Last modified: 2010-07-26 05:54:56 UTC
Created attachment 7566 [details] Two examples of wrong and right versions of the semantic data. So the problem is that article pages have properties of OTHER articles. I discovered that this has something to do with the job queue. When a article A updates its semantic data and then calls the job queue and updates another page B, some (but not all) of the pages A properties get passed over to page B, also categories get passed over sometimes. This is really weird and makes a lot of trouble in my wiki. Also it doesn’t happen every time the wiki processes a page from the job queue but I can not say under which exact circumstances this happens. When I resave a article or use the SemanticUpdateOnPurge extension to update with purge (this extension is definitely not causing the bug!), the page gets the right semantic attributes but therefore another article gets some of the same attributes. This Bug also happens wen running the SMW_refreshData.php. I also tried to remove and rebuild all SMW-DB tables and then running SMW_refreshData.php, same result. Tested SMW 1.5 and 1.5.1.1. Some weeks ago I updated from the latest 1.4.x to SMW 1.5, So perhaps this is a new bug in 1.5. I am not completely sure if this ever happened with SMW 1.4 or earlier. At least I never discovered something like that in earlier versions. This bugs makes inline queries useless because they display wrong results all the time!
I forgot to mention that Laurent from the SMW devel mailing list confirmed the bug although it looks like the bug is gone in his wiki. Ich habe eine Kopie des entsprechenden Wikis die hier öffentlich verfügbar ist: http://test.rpgdev.stejau.de/wiki/Hauptseite Sehr gut zu sehen sind die Auswirkungen übrigens hier: http://test.rpgdev.stejau.de/wiki/Test und (falls auf der Seite das Diagramm nicht angezeigt wird) dort: http://test.rpgdev.stejau.de/wiki/Kategorie:Story-Quests
Sorry for swifting into german language in my last post. It's saying that the bug can be seen on some pages of my public test wiki: http://test.rpgdev.stejau.de/wiki/Hauptseite
Now I was able to solve the problem. The problem was not a SMW bug, it was provoked by a bug in the Variables Extension, Array Extension and HashTables Extension. Variables defined there were like superglobals in case of some job queue updates, semantic updates and page import. I thought I checked for this bug already but I had some unexpected behavior so my test at the first time was negative untill I found some strange behavior on my last page import. I fixed this bug in all of the three extensions above. Everybody using those extensions (especially together with SMW) should update them.