• TaskListener#getCharset() always returns Nonnull according to the documentation: https://github.com/jenkinsci/jenkins/blob/master/core/src/main/java/hudson/util/StreamTaskListener.java
• StreamTaskListener accepts null Charset in API and then may return null in getCharset()
• Once I added FindBugs annotations for StreamTaskListener in https://github.com/jenkinsci/jenkins/pull/3489 , FindBugs predictably failed.
• TaskListener checks getCharset() for null in the code internally, so probably the annotation should be just changed to Nonnull. But it's formally incompatible change

Current state:

• API users may get NPE if they handle TaskListener according to the requirements

Nonnull annotation has been introduced by jglick in https://github.com/jenkinsci/jenkins/commit/9ed6b01317424f3f441022b3a2f23fbbc5ae1543

I see two options:

• Relax requirements in TaskListener#getCharset() and allow null returns. It is a "breaking change" from API PoV (now we require API users to check for nulls)
• Rework StreamTaskListener to disallow nulls and to default to something (UTF-8). May break some use-cases as well since now it defaults to the system's default charset