Collation check for Oracle is looking incorrectly at NLS_CHARACTERSET instead of NLS_SORT

Summary

The collation check introduced by JRA-39970 is checking the wrong attribute for Oracle. It's looking at NLS_CHARACTERSET, which defines the character set, and not the NLS_SORT which defines the collation.

Environment

Verified on:

• Oracle 12c
• JIRA 6.4.5

Steps to Reproduce

1. Ensure your Oracle database uses BINARY for NLS_SORT
2. Install JIRA to connect to Oracle as per the directions Connecting JIRA to Oracle
3. Start JIRA
4. Look at the log file

Expected Results

No warnings regarding collation are logged, as the correct collation is in use.

Actual Results

The warning is logged in the atlassian-jira.log file:

2015-06-09 08:01:04,432 http-bio-8080-exec-15 WARN anonymous 480x10x1 1llnki2 172.17.42.1 /secure/SetupDatabase.jspa [jira.appconsistency.db.CollationCheck] 

****************************************************************************************************
You are using an unsupported oracle10g collation: AL32UTF8. This may cause some functionality to not work.
Please use BINARY as the collation instead.
****************************************************************************************************

Workaround

Ignore the warning, since it's not actually valid in this case.