Intermittent CI failure in wallet_send.py, in `test_maxfeerate` #36363

1. 
   maflcko commented at 6:33 AM on September 28, 2026: member

   The CI failure in "Windows, test cross-built" looks unrelated to this PR: https://github.com/bitcoin/bitcoin/actions/runs/36297704387/job/108561658571

   The only failing test is wallet_send.py, in test_maxfeerate (added in #29278):

   File "wallet_send.py", line 225, in test_maxfeerate
       self.nodes[0].sendtoaddress(self.nodes[0].getnewaddress(), amount=1, fee_rate=Decimal("1.009"))
   test_framework.util.JSONRPCException: Fee rate exceeds maximum configured by user (maxfeerate) (-6)
   

   Originally posted by @azuchi in #36355 (comment)

2. maflcko added the label Wallet on Sep 28, 2026
3. maflcko added the label CI failed on Sep 28, 2026
4. maflcko added this to the milestone 33.0 on Sep 28, 2026
5. 
   DrahtBot commented at 6:36 AM on September 28, 2026: contributor

   https://github.com/bitcoin/bitcoin/actions/runs/36217897680/job/108337434196?pr=36317#step:9:42951

6. 
   ismaelsadeeq commented at 8:48 AM on September 28, 2026: member

   Thanks for reporting. By reading the logs, it seems a fee rate of 1.009 was passed as the -maxfeerate for the wallet, and then we attempted to submit a tx with the exact 1.009 fee rate. The check is that any fee rate above that should be rejected for creation and submission. It seems somehow the transaction ended up with a fee rate higher than what was intended, and the wallet rightfully rejects it.

   My suspect is #29278 (review), because this test on the boundary, and since the wallet may falsely reject, this occur.

   A simple fix would be to not test at the boundary, but that just hides the issue.

   cc @polespinasa @furszy

7. 
   maflcko commented at 9:07 AM on September 28, 2026: member

   I haven't looked at this, but the option would mostly be useful to serve as a belt-and-suspender, not as an option to somehow set an exact fee. So I think overshooting by a small margin is fine?

8. 
   ismaelsadeeq commented at 9:11 AM on September 28, 2026: member

   I haven't looked at this, but the option would mostly be useful to serve as a belt-and-suspender, not as an option to somehow set an exact fee. So I think overshooting by a small margin is fine?

   Yes, I will await a few more eyes before opening a fix.

9. 
   hebasto commented at 10:22 AM on September 28, 2026: member

   https://github.com/bitcoin/bitcoin/actions/runs/36402698838/job/108871080019

10. 
    achow101 commented at 7:45 PM on September 28, 2026: member

    The failure is due to a slightly smaller ECDSA signature which results in a tx size smaller than estimated, which increases the actual fee rate.

    A very quick fix is to make the test use taproot inputs.

11. 
    fanquake commented at 9:42 AM on September 29, 2026: member

    https://github.com/bitcoin/bitcoin/actions/runs/36549817893/job/109345359793?pr=36379#step:11:35852

12. 
    fanquake commented at 9:44 AM on September 29, 2026: member

    https://github.com/bitcoin/bitcoin/actions/runs/36502104665/job/109195103383?pr=36377#step:11:36538

13. achow101 closed this on Sep 29, 2026

    ---

14. pull[bot] referenced this in commit ba8fdb9717 on Sep 29, 2026

Milestone
33.0