Follow-up to #2964 / #2974.
Once the Boole fork has activated and stabilized, and the pre-fork code paths are removed from the codebase (alan domain branch in DomainTypeAtSlot, pre-fork proposer formula, alan topic names, discovery next-domain acceptance, etc.), the two-leg fork test matrix stops earning its keep — both legs would exercise identical code.
At that point, in the same PR as the pre-fork code removal (so CI and code never disagree about which fork worlds exist):
Do not do this at fork activation — as long as pre-fork branches physically exist in the code, the pre-fork leg is the only coverage they have.
Follow-up to #2964 / #2974.
Once the Boole fork has activated and stabilized, and the pre-fork code paths are removed from the codebase (alan domain branch in
DomainTypeAtSlot, pre-fork proposer formula, alan topic names, discovery next-domain acceptance, etc.), the two-leg fork test matrix stops earning its keep — both legs would exercise identical code.At that point, in the same PR as the pre-fork code removal (so CI and code never disagree about which fork worlds exist):
SSV_TEST_BOOLE_FORKswitch entirely once it loses its meaning.ssv-boole-postjob from.github/workflows/unit-test.ymlandspec-test-boole-postfromspec-test.yml. Keep the originalssv/spec-testjobs — they own the Codecov upload and the required-status-check slots.TestBuildPeerTrimScores_ComputesScoresForAllCandidates_BooleTopicsbecomes redundant with its non-Boole sibling at that point, so one of the pair goes.ssv-boole-post/spec-test-boole-postwere added to branch-protection required checks in the meantime (see Enable the SSV_TEST_BOOLE_FORK=post unit-test matrix (deferred from #2941) #2964), remove them from that list first so merges don't get stuck waiting on deleted jobs.Do not do this at fork activation — as long as pre-fork branches physically exist in the code, the pre-fork leg is the only coverage they have.