Summary

Newlines that are part of Unified Logging messages can result in parsing ambiguities. We resolve the parsing ambiguities by annotating newlines in log messages with a special decorator.

Problem

Consider the Unified Logging output resulting from log_info(gc)("A\n[gc] B");:

[gc] A
[gc] B

We cannot recover the fact that this is a single log message "A\n[gc] B". A tool parsing this output will likely incorrectly identify two separate log messages "A" and "B". We should ensure that such parsing ambiguities can never occur.

Solution

We propose to add a special decorator at newlines in log messages to allow unambiguous parsing.

Review and mailing list threads:

• https://github.com/openjdk/jdk/pull/21712
• https://mail.openjdk.org/pipermail/hotspot-dev/2024-October/095207.html

Specification

We propose to add the special decorator [ ] followed by a space at newlines in log messages, where we choose the number of spaces between the brackets to align with the log message decorator. For example, the output resulting from log_info(gc)("A\n[gc] B"); is now

[gc] A
[  ] [gc] B

and we can parse it unambiguously. As an added bonus, the log message also aligns properly, generally improving readability.