LUCENE-9047: Write checksum as big endian in NRT replicator

[iverase]

Left over from #107, we should read / write checksums using big endian in the NRT replicator.

[uschindler]

Hi @iverase
the merge broke tests: We fixed over the weekend the NRT tests. They had a bug in the security policy, so no NRT Test wasn't running at all. So you did not see the test failure.

If you fix it, make sure to run the replicator tests with -Dtests.nightly=true, so it runs all tests (takes a few minutes)

I will backport the test fixes also to 8.x later. It's great that we fixed it yesterday :-) The NRT tests were not running since now YEARS!

[iverase]

Thanks @uschindler, yikes that explain it all :)

This fix seems to work, I run the test with the nightly flag and it seems happy:

 ./gradlew :lucene:replicator:test -Dtests.nightly=true

> Task :randomizationInfo
Running tests with randomization seed: tests.seed=3735F7B6C10CE977

> Task :lucene:replicator:test
:lucene:replicator:test (SUCCESS): 44 test(s), 1 skipped
The slowest tests (exceeding 500 ms) during this run:
  190.02s TestStressNRTReplication.test (:lucene:replicator)
   7.52s TestIndexAndTaxonomyReplicationClient.testConsistencyOnExceptions (:lucene:replicator)
   6.52s TestNRTReplication.testFullClusterCrash (:lucene:replicator)
   5.44s TestIndexReplicationClient.testConsistencyOnExceptions (:lucene:replicator)
   3.49s TestNRTReplication.testCrashPrimary3 (:lucene:replicator)
   3.36s TestNRTReplication.testCrashReplica (:lucene:replicator)
   3.27s TestNRTReplication.testCrashPrimary2 (:lucene:replicator)
   2.65s TestNRTReplication.testReplicaCrashWithCommit (:lucene:replicator)
   2.52s TestNRTReplication.testCrashPrimary1 (:lucene:replicator)
   2.52s TestNRTReplication.testIndexingWhileReplicaIsDown (:lucene:replicator)
The slowest suites (exceeding 1s) during this run:
  190.06s TestStressNRTReplication (:lucene:replicator)
  32.55s TestNRTReplication (:lucene:replicator)
   8.33s TestIndexAndTaxonomyReplicationClient (:lucene:replicator)
   6.28s TestIndexReplicationClient (:lucene:replicator)

BUILD SUCCESSFUL in 3m 27s

[uschindler]

What did you change, or was the test failures on Jenkins only because of the first merge on the directory API and this one fixed it?

This explains why you opened THIS pull request. Sorry for the noise.

Uwe

[uschindler]

It will be funny, when Solr picks the new snapshot builds!

[mikemccand]

The NRT tests were not running since now YEARS!

OH NO!

[uschindler]

I backported the fixes. But I can tell you: It was only a master problm, 8.x was fine. I still backported Dawid's fixes.

[uschindler]

the suppressAccessChecks permission got lost in master, so the NRT tests were not able to crush the test server with SIGSEGV. It's now working in both master and 8.x. I ran the tests and all worked.

[mikemccand]

Awesome, thank you @uschindler!