Benefit from pre-warmed cache in console/worker context #41
Conversation
|
|
||
| return parent::handle($request, $type, $catch); // TODO: Change the autogenerated stub | ||
| parent::boot(); |
There was a problem hiding this comment.
Isn't boot() called in the case of HTTP too? Do we risk the case of calling prepareCacheDir() twice?
Good idea though, I'm just curious about that specific point.
There was a problem hiding this comment.
It would be called twice indeed but the second time it would be skipped because of the first if statement in prepareCacheDir.
is_dir($writeDir) would be true so it would exit directly, as it does when hitting a warm Lambda.
There was a problem hiding this comment.
Right, thanks. Then would it make sense to remove the call from handle?
There was a problem hiding this comment.
Basically Kernel::boot was previously used by the bridge but a feature added in Symfony 5.2 introduced Kernel::preBoot (obviously executed before Kernel::boot) and preBoot was dumping the cache and container before we had a chance to copy it to /tmp.
The call in Kernel::handle was introduced in #37 to fix this issue with Symfony 5.2 and make the bridge work on all versions of Symfony.
That fixed it for HTTP apps but broke it for Console apps/workers/scheduled commands because Kernel::handle is never called in a console context but kernel::boot is (and it's ok to put it here because Kernel::preBoot is never called either in console).
So I don't see any other way of making it work for both HTTP and Console 🙂
There was a problem hiding this comment.
Oh right 😅 thank you for the history, that makes sense then!
Thank you!
No description provided.