Logback should default to JVM's default charset instead of ASCII

[gbaso]

Following #23827, default charset where defined for a base logback configuration in 5e26954, as ${CONSOLE_LOG_CHARSET:-default} and ${FILE_LOG_CHARSET:-default}.

Unfortunately, logback interprets that as Charset.forName("default"), which taking a look at sun.nio.cs.StandardCharsets$Aliases is hardcoded to return ASCII. I feel like Charset.defaultCharset() would be a more natural choice.

[wilkinsona]

Thanks for the report, @gbaso. Can you please provide a bit more information about when this has caused a problem? LoggingSystemProperties should set the FILE_LOG_CHARSET and CONSOLE_LOG_CHARSET system properties so I'd like to understand exactly when you're seeing the fallback to default occurring.

[gbaso]

It seems I was mistaken, and my issue was actually related to openjdk/jdk#4733 (comment).

I do, however, still think that spring boot should take a more opinionated approach than defaulting to the "default" charset (which, as I said, is ASCII), using UTF-8 instead.

[wilkinsona]

@gbaso When the system properties are in effect, it should already default to UTF-8. To make sure that we fully understand the problem I'd like to know more about your situation and why that apparently wasn't happening. Can you please provide a minimal example that reproduces the problem that you're having?

[gbaso]

@wilkinsona I apologize if I was not clear enough, spring boot is working correctly, it was a problem on my end (and in the mechanism the JVM sets the default charset).

However by reading the code, specifically https://github.com/spring-projects/spring-boot/blob/main/spring-boot-project/spring-boot/src/main/resources/org/springframework/boot/logging/logback/defaults.xml, it looks like if for some reasons FILE_LOG_CHARSET and CONSOLE_LOG_CHARSET are unset, they default to the string default:

<property name="FILE_LOG_CHARSET" value="${FILE_LOG_CHARSET:-default}"/>

which is interpreted as Charset.forName("default") and is an alias for ASCII.

Given that even OpenJDK is moving to UTF-8 by default, I think we should replace
value="${FILE_LOG_CHARSET:-default}" with value="${FILE_LOG_CHARSET:-UTF-8}".

[wilkinsona]

it looks like if for some reasons FILE_LOG_CHARSET and CONSOLE_LOG_CHARSET are unset, they default to the string default

Yes, that's right. What I'd like to understand is why, in your case, those system properties were unset. Can you please provide a minimal example that reproduces that behaviour?