LuciferYang opened a new pull request, #11242:
URL: https://github.com/apache/paimon/pull/11242

   ### Purpose
   
   close #11240
   
   Before each index build, `LuminaVectorGlobalIndexWriter` wrote 
`LUMINA_EXECUTOR_THREAD_COUNT` into the map behind `System.getenv()` by 
reflection (once per distinct value per JVM). liblumina reads that variable 
with libc `getenv`, so the write never reached the native executor. It only 
changed what `System.getenv()` returns inside the JVM, and the environment of 
child processes started from a `ProcessBuilder` whose `environment()` map was 
read.
   
   This removes `setEnv`, `configureExecutorThreadCount` and 
`CONFIGURED_ENV_VARS`. `lumina.diskann.build.thread_count` is still passed to 
the native builder as before. Its description and the README row now say that 
the build runs on Lumina's native executor, whose size comes from 
`LUMINA_EXECUTOR_THREAD_COUNT` in the process environment (default 32), and 
that the option does not change it.
   
   ### Tests
   
   - `LuminaVectorGlobalIndexTest#testBuildDoesNotChangeExecutorThreadCountEnv` 
builds an index and checks that `System.getenv("LUMINA_EXECUTOR_THREAD_COUNT")` 
is the same before and after. It fails without the change, also when the whole 
class runs in one JVM, and needs the Lumina native library (linux/amd64), so it 
is skipped elsewhere.
   - `LuminaVectorGlobalIndexWriterTest#testThreadCountIsPassedToBuildOptions` 
checks that the default (32) and an explicit value of the option still reach 
the native build options.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to