This reverts commit cbe0f44c4f.
This achieves the same result as the reverted commit wrt. process in the
same manner (writing a keepalive into the socket and checking if it
failed to send). However, it does _not_ allow the process to die on
reaching pipe EOF, since this can cause many spams of subprocesses when
stdin is actually not a tty (e.g. in a Docker container).
this fixes a wide range of blocks with invalid blockstates becoming update! blocks on the client.
The most common occurrence of this was air with nonzero metadata left behind by world editors which set blockIDs but not block metadata. This caused large ghost structures of update! blocks to appear from nowhere.
The performance impact of this is very minimal (20 microseconds per chunk load in timings, compared to average 660 microseconds to load tiles).
this was a problem before the recent clean-up; the only reason it just decided to show now is because 2000+25 is valid PHP code, so PHP saved our asses.
when a block has sole ownership of an ID, the state bitmask can be ignored and we can just claim the whole metadata range for that single block.
This fixes a large number of issues with unknown blocks on older worlds where world editors did not remove the metadata, although update blocks will currently still appear on initial chunk send due to lack of AOT conversion (TODO).
when using dynamic reflection (which is the default), any time static reflection comes into play, bad shit starts to happen because of FileReadTrapStreamWrapper.
I attempted to fix these issues (phpstan/phpstan-src#801) and failed miserably.
So, to save the hassle, it's time to just remove OPcache from the picture (which, unfortunately, also means that PHPStan will not benefit from JIT).
