Page MenuHomePhabricator

Maintenance categories (e.g. Pages_with_missing_files) continuously grow outdated
Open, MediumPublic



Since several days on nl-wiki some categories are filled with pages which should not be there. So far I have seen it concerns only maintenance categories which are filled by templates which add a category to a page if in some way is matched with particular criteria.

It concerns:

  • [[w:nl:Categorie:Wikipedia:Pagina's met onjuiste bestandsverwijzingen]]
  • [[w:nl:Categorie:Wikipedia:Pagina's met ontbrekende references]]
  • [[w:nl:Categorie:Wikipedia:Hulpcategorie voor tijdelijk gebruik]]
  • [[w:nl:Categorie:Wikipedia:Beginnetje nog niet onderverdeeld]]

On the pages in the category the category isn't shown. When I save the concerning pages without changing anything, the page leaves the category.

Version: unspecified
Severity: normal



Event Timeline

bzimport raised the priority of this task from to Medium.Nov 21 2014, 11:49 PM
bzimport set Reference to bz31577.
bzimport added a subscriber: Unknown Object (MLST).

seems to be some sort of caching problem? Do you have an example of pages that were in these categories when they shouldn't have been? wrote:

Yes it seems some sort of caching. I made hundreds of null-edits to make it workable again to do normal maintenance. Only this category I skipped:

All those inside shouldn't be there, example:

Romaine wrote:

Resolved -> Reopened

Today again 26 new pages that shouldn't be in that category came in. wrote:

This is every going on since 1.18, perhaps even every day.

Today 395 pages in category which should not be there:

Who can help to solve this bug?

I bet this is somehow related to bug 31576.

So I'm guessing (and this is very much a guess and probably totally wrong) sometimes [[nl:template:Beginnetje]] (the stub template) gets parsed very very wrongly and some of the magic words aren't interpreted as magic words, and then it causes it to include the wrong category (?) (Which doesn't help with the wtf is wrong with the magic word parsing question) wrote:

The idea Bawolff mentions seems to be correct if I choose random some pages with the category-problem and check for magic words that aren't interpreted as magic words. I see {{PAGENAME}} and {{NAMESPACE}} be on the bottom of the page in red with editting a page telling that these "templates" do not exist.

(In reply to comment #5)

I bet this is somehow related to bug 31576.

This seems to be the case, yes. I've investigated these issues a bit today, see my comment on bug 31576 for details. Also, from now on I'll comment on bug 31576, not on this bug. wrote:

Maybe there is something else going on too. I checked each page and each included template on the pages in the categorie but nowhere is this category inserted. I can remember that someone used it some while ago, but currently it is in no place in use. Still for some reason this category is filled with some pages.

If it is just a bug that the magic words do not function proparly, then still these pages should not have been in this category as the category is nowhere inserted.

Greetings - Romaine

If it was used some time ago, you should ask the category links to be refreshed first (or repurpose this as a maintenance script run blocking bug 16660): it's very common for categories to have ghost entries.

Pages with broken file links may be a bit broken since uploading a file doesnt trigger a refreshlinks update so pages don't get removed from that category (bug 46749). Are there examples for that cat of pages that should have never been in it.

For cite cat see bug 46978. It sounds kind of like the cite clear state hook isnt getting run from the job queue somehow, but I have no idea how that could happen

I think these two issues are separate.