restrict std::cerr to errors; use std::cout for warnings and info #32538

1. 
   furszy commented at 6:42 pm on May 16, 2025: member

   Warning messages were previously sent to stderr, which could cause them to be misinterpreted as actual errors. This change redirects warning and informational messages to stdout instead, making it explicit that they are not errors and preventing unintended stderr handling.

   To verify this behavior, added a test case triggering the simplest init warning I could find (one that was actually not previously covered by tests): the -bind “bad-port” message. So, running the test commit on master (without the fix) causes the test framework to fail during shutdown due to a non-empty stderr buffer (treated as an error). With the fix applied, the warning is correctly handled via stdout, and the test passes as expected.

2. init: restrict std::cerr to errors; use std::cout for warnings and info

   Warning messages were previously sent to std::cerr, which could cause them
   to be treated as errors when they are not. This change pipes warning and
   informational messages through std::cout instead, making it clearer that
   they are not errors.

   4ff27834e3
3. test: add coverage to ensure init warnings aren't treated as errors

   Verifies that the bad-ports warning is no longer misinterpreted as an error.
   Previously, it was sent through std::cerr, causing the test framework to wrongly
   classify it as an error during shutdown. Now that the message is routed through
   std::cout, this test confirms that warnings are handled properly and do not
   trigger false failures.

   9e9b16523f
4. 
   DrahtBot commented at 6:42 pm on May 16, 2025: contributor

   The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

   Code Coverage & Benchmarks

   For details see: https://corecheck.dev/bitcoin/bitcoin/pulls/32538.

   Reviews

   See the guideline for information on the review process.

   Type	Reviewers
   Concept ACK	pablomartin4btc

   If your review is incorrectly listed, please react with 👎 to this comment and the bot will ignore it on the next update.

5. 
   maflcko commented at 7:00 pm on May 16, 2025: member

   Warning messages were previously sent to stderr, which could cause them to be misinterpreted as actual errors.

   Not sure.

   • When something literally starts with Warning: ..., I fail to see how it can be misinterpreted.
   • This change means that warnings may be missed in tests/CI or elsewhere (see the failing CI).
   • Sending warnings to stderr is normal and expected. See for example python -c 'print(b"\p")'
6. 
   pablomartin4btc commented at 7:01 pm on May 16, 2025: member

   Concept ACK

   I think some tests using stop_node with the warning in the args would fail as in is_node_stopped checks for the warning in stderr (eg wallet_multiwallet.py), it needs to be fixed too.

7. 
   pablomartin4btc commented at 7:03 pm on May 16, 2025: member

   This change means that warnings may be missed in tests/CI or elsewhere (see the failing CI).

   That’s my concern with this approach.

8. 
   furszy commented at 7:24 pm on May 16, 2025: member

   When something literally starts with Warning: …, I fail to see how it can be misinterpreted.

   It is actually being misinterpreted by our own test framework. The framework does not look at the stderr content, it only checks whether there is something inside stderr or not during shutdown, failing when finds something there.

   This change means that warnings may be missed in tests/CI or elsewhere (see the failing CI).

   That’s ok. That’s what the PR is proposing. To redirect them to stdout. I just missed to update some test cases that are working around the stderr issue during cleanup by manually skipping warning msgs during shutdown.

   Sending warnings to stderr is normal and expected. See for example python -c ‘print(b"\p")’

   This is the most compelling argument of the three. You got me there. Closing..

9. furszy closed this on May 16, 2025

   ---

10. 
    maflcko commented at 6:19 am on May 17, 2025: member

    When something literally starts with Warning: …, I fail to see how it can be misinterpreted.

    It is actually being misinterpreted by our own test framework. The framework does not look at the stderr content, it only checks whether there is something inside stderr or not during shutdown, failing when finds something there.

    That’s intentional. We want the tests to fail when there is any warning or error (from Bitcoin Core or any other tool/sanitizer/lib) that is not accounted for, so that they are properly documented, investigated and accounted for.

11. 
    fanquake commented at 10:20 am on May 17, 2025: member

    We want the tests to fail when there is any warning or error (from Bitcoin Core or any other tool/sanitizer/lib) that is not accounted for

    Opened #32542 in relation to this.

Contributors
furszy DrahtBot maflcko pablomartin4btc fanquake