Bug #99815 Insist on using System.nanoTime() to measure elapsed time

• Submitted: 9 Jun 2020 3:01
• Modified: 9 Jun 2020 4:21
• Reporter: Yanming Zhou (OCA)
• Status: Verified
• Category: Connector / J
• Severity: S3 (Non-critical)
• Version: 8.0.x
• OS: Any
• CPU Architecture: Any

[9 Jun 2020 3:01] Yanming Zhou

Description:
System.nanoTime() is introduced since JDK 1.5, It's safe enough to call it directly not by reflection:

https://github.com/mysql/mysql-connector-j/blob/dbd865c9fc68701efc70197ea1e65002520bed91/s...

TimeUtil.getCurrentTimeNanosOrMillis() returns uncertain TimeUnit, It cause Query.getExecuteTime() meaningless, in nanos or ms?

How to repeat:
It's not a bug, It's an improvement.

Suggested fix:
1. Use System.nanoTime() instead of System.currentTimeMillis().
2. Remove property "useNanosForElapsedTime" since it is always true.
3. Rename method names indicate that return nanos, for example Query.getExecuteTime() to Query.getExecuteTimeInNanos()

[9 Jun 2020 3:22] Yanming Zhou

Then there is a bug here, executeTime is in nanos if JDK is 1.5 or later, slowQueryThresholdMillis is always in ms, then all queries will be treat as slow query.

https://github.com/mysql/mysql-connector-j/blob/dbd865c9fc68701efc70197ea1e65002520bed91/s...

[9 Jun 2020 4:21] MySQL Verification Team

Hello Yanming Zhou,

Thank you for the report and feedback.

regards,
Umesh

[16 Jun 2020 23:52] Filipe Silva

Note that `useNanosForElapsedTime` is "false" by default. All seems to work fine if you don't set it to "true". You are right, though. Using nanos really breaks some of the slow query logging features.

Regarding `Query.getExecuteTime()` returning nanos or millis, I'd say this is debatable question. It makes sense to me that you get different units depending on the value you set in `useNanosForElapsedTime`.

Thanks for your interest in Connector/J and MySQL.