Elliotte, This PR keeps your default and adds an opt-in. So the -10 is not about noise, defaults, or users who want a quiet build. It is about the option existing at all.
- Spark and Kyuubi deliberately deferred the 3.6.1 upgrade to keep this output for human auditing. Flink upgraded and their license checker broke outright: it parses the "Including ..." lines, and 3.6.1 made it report that nothing was shaded. They forced this plugin's logger to DEBUG in CI, then shipped a follow-up PR to clean up the resulting noise. That is the "just use the log level" workaround, and its cost is real, per-consumer, and not settable from the pom. - MSHADE-479 was not a defect report and not a user request. It was filed against a build where this output was in your way while you were looking for an unrelated test failure, and the PR calls the messages “log cruft". That is a preference. It was applied to everyone, in a patch release, with no compatibility path. When a change is a preference, "make it configurable" is the remedy, not an escalation. - "Logs are not API" is true and irrelevant here. The flag promises nothing about text, order, or format, and makes no stability commitment. It only lets the old level be selected again. Hyrum's Law explains why removing the output broke people; it does not justify refusing a way back. Countless systems and scripts on this planet depend on logs to function, and build tooling is no exception: CI summaries, license checkers like Flink’s, release and audit scripts, log shippers. Maven's own INFO/WARN vocabulary is already an interface in practice. - "Look in the jars" answers what is inside, not which dependency put it there or why something was filtered out. A jar cannot tell you that an artifactSet or filter pattern matched nothing; that diagnostic exists only in the log. On complexity: it is one boolean in a mojo that already exposes dozens. If one boolean is too much, the lower-complexity answer is to revert MSHADE-479. Accept the opt-in or revert the change. Rejecting both is a veto, not a compromise. Herve and Romain are in favour on this thread, martinpaljak filed #756, and I opened and Herve approved the PR. On the other side there is one -10. If we cannot reach a consensus in this discussion thread, it seems our only option is to initiate a VOTE for a final decision. Thanks, Cheng Pan --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
