Uwe Schindler commented on Improvement LUCENE-4323

You fix just hides the problem which is caused by another randomization in LTC:

step2: apply my patch, which forces this to instead be Double.POSITIVE_INFINITY instead of 8.796093022207999E12. This should really have no effect here, since the directories are tiny (all < 1MB)!

Here is the explanation:
Your fix - (setting LMP's maxMerge) is not changing anything here - the max size is still Long.MAX_VALUE internally! The change is really that you change the order of newLogMergePolicy(),newIndexWriterConfig() [they were the other order before] -> this changes random numbers completely.
What I changed in my patch was missing randomization in new LogMergePolicy for cfsRatio! Before it was always 0.1, preventing creating CFS on addIndexes at all! In the original seed without your patch, the cfsRatio was then randomized to ~0.8, causing CFS files to be created -> causing disk Full.

The cause is simply additional randomization, missing before for newLogMergePolicy().

The correct way to fix this test is to prevent CFS files at all: newLogMergePolicy(false).

Should I commit that fix?

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
--------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]

Reply via email to