On 11/12/13 22:56, Daniel Kinzler wrote:
The thing is: we are changing what's in the list of relevant options. Before the deployment, there was nothing in it, while with the new code, the user language should be there. I suppose that means we need to purge these "pointers".
Would bumping wgCacheEpoch be sufficient for that? Note that we don't care much about puring the actual parser cache entries, we want to purge the "pointer" entries in the cache.
Yes, bumping $wgCacheEpoch is sufficient. It's not possible to purge the pointer entries without also invalidating the actual parser cache entries, since changing the set of used options will change all of the parser output keys.
If you just deploy the code without bumping $wgCacheEpoch, the pointers will be overwritten each time a parser cache entry is saved, effectively unlinking all the existing variants of that parser cache entry. So it would sort itself out eventually, without any pollution.
It's not necessary to call ParserOutput::recordOption(). ParserOptions::getUserLangObj() will call it for you (via onAccessCallback).
Oh great, magic hidden information flow :)
Yeah, there's a tradeoff between the system being easy to use and being easy to understand. Most people who write parser extensions are pretty naïve about how the parser cache works, which is why I accepted Platonides' idea of using ParserOptions accessor calls to determine cache-varying options, in r70783. It seemed to have the best chance of working with naïvely written extensions.
https://www.mediawiki.org/wiki/Special:Code/MediaWiki/70783
-- Tim Starling