Follow-up to #2754003: Move more cache items from the default to the bootstrap or discovery bin
The theme registry has two different cache entries per theme, the full registry and the runtime registry, which is used on all normal requests and is small.
For a default D8 installation, that's the last cache entry from the default cache bin that's loaded on a normal request with warm cache.
To move this to bootstrap, we need to inject a second cache bin service into the Theme Registry service.
| Comment | File | Size | Author |
|---|---|---|---|
| #4 | theme-registry-cache-2761403-4-interdiff.txt | 670 bytes | berdir |
| #4 | theme-registry-cache-2761403-4.patch | 3.11 KB | berdir |
| #2 | theme-registry-cache-2761403-2.patch | 3.1 KB | berdir |
Comments
Comment #2
berdirFirst patch. Not sure about the optional argument there, made the new cache backend optional for now. Might also turn it around, it is only used for tests anyway.
Comment #4
berdirOops.
Comment #5
dawehnerJust check on a plain D8 installation. The runtime cache has 18K in size, the non runtime one, something like 60k, which is all nothing compared to the 200k entries I saw on some other site.
Comment #6
fabianx commentedRTBC + 1
Comment #7
catchComment #9
catchReverted due to #2749955: Random fails in UpdatePathTestBase tests.
Back to CNR.
Comment #11
hussainwebI am not sure if this was the cause for the failures in UpdatePathTestBase tests. I have the commit that reverts this change and still the test fails for me locally. Verbose message for the failures is always this:
Comment #12
hussainwebPlease disregard #11. While debugging for this error, it seemed to fix itself. I am now able to run it multiple times without failing. I don't know if that's conclusive considering it is a random failure.
Comment #13
dawehnerI would have tried to inject a null backend into the theme registry inside
\Drupal\Core\Update\UpdateServiceProvider...Comment #16
berdir@alexpott: I think I remember you saying that this was not actually responsible for those random fails?
Comment #18
lauriiiAny updates on this?
Comment #19
berdirSetting back to RTBC to get feedback from @alexpott, afaik there is nothing else to do here.
Comment #21
catchI just scanned back through that issue, and pretty sure this issue was a red herring with the random fails.
I've reverted the 8.3.x revert, moving this to fixed again!