Update `vote_generator_index` dedup only when the queue accepts a replacement (#5131) When a replacement push hit a full bucket, dedup was updated even though the entry was never enqueued, orphaning the root: size()/empty() report on dedup while next_batch() drains the queue, so the verifier wait predicate stayed satisfied with nothing to pop, busy-spinning the voting threads and permanently silencing the root. Harden against any future divergence by backing empty() with the queue and asserting the dedup/queue invariant in next_batch().
__________
Type: commit
Hash: 38c7eda56025713d983094efec5360cd1bd34b9e
Created: yesterday
__________
gr0v1ty/nano-node:38c7eda56025713d983094efec5360cd1bd34b9e
| Test Case | Status | Duration | Median Deviation | Log | Graph |
|---|---|---|---|---|---|
| 5n4pr_conf_10k_bintree | PASS | 141 s | 29.0 s | - | Show |
| 5n4pr_conf_10k_change | PASS | 157 s | -1.0 s | - | Show |
| 5n4pr_conf_change_dependant | PASS | 120 s | 6.0 s | - | - |
| 5n4pr_conf_change_independant | PASS | 113 s | -1.0 s | - | - |
| 5n4pr_conf_send_dependant | PASS | 121 s | 5.0 s | - | - |
| 5n4pr_conf_send_independant | PASS | 123 s | 10.0 s | - | - |
| 5n4pr_rocks_10k_bintree | PASS | 125 s | 2.0 s | - | - |
| 5n4pr_rocks_10k_change | PASS | 230 s | 69.0 s | - | - |