You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
T27 OS: clear the three inherited red checks (duplicate bodies, absurd Verilog widths, t27b ratchet) in the t27os specs #11030
Refs #10782. Depends on #10783 (the t27os stack is not on master yet; bees start from master, so this waits until it lands).
Problem
Three CI checks are red on every PR that touches the T27 OS specs, and the cause is in the specs, not in the PR:
duplicate-bodies: the bodies of chan, clamp01, digits_val, flag_bit, scale, ws_at, ws_back, ws_before, ws_fwd, ws_len_at and ws_len_before are copied into the t27os specs and were in no ledger. Reuse the body that exists; a bless is the last resort and must say why.
Verilog widths (absurd-width ledger): glass_window.t27 emits a width of 1535999 bits and pty_server.t27 one of 9175039 bits. A width at or above 2^20 bits is an arithmetic accident, usually a big byte buffer declared as a packed value.
t27b-native-ratchet: 25 t27os specs are UNLISTED and 4 ledger rows are STALE.
Found while checking 9 agents for T27 OS (PR #11015). The three reds were inherited from the base, not caused by that PR.
Boundary
specs/t27os/glass_window.t27
specs/t27os/pty_server.t27
specs/t27os/pty_logic.t27
specs/t27os/pty_client.t27
specs/t27os/term_pane.t27
specs/t27os/vt.t27
User Scenarios
Given a PR that changes only a t27os spec, when the duplicate-bodies check runs, then it finds no new duplicate body and exits 0.
Given glass_window.t27 and pty_server.t27, when the Verilog widths check scans the corpus, then neither is listed as NEW and the job exits 0.
Given the t27os specs, when the t27b-native-ratchet check runs, then it reports 0 UNLISTED and 0 STALE for t27os and exits 0.
Requirements
FR-001: Each duplicated body MUST be replaced by a call to the one body that already exists, or the copy MUST be justified in the ledger move in the same commit.
FR-002: The two byte buffers behind the absurd widths MUST be declared so that no emitted width is 2^20 bits or more, with the spec tests of both files still passing.
FR-003: The t27b ledger rows for the t27os specs MUST be moved in the same commit, down only where a row went STALE.
FR-004: No width, duplicate or ratchet check MUST be loosened to make the job pass.
Success Criteria
The duplicate-bodies, Verilog widths and t27b-native-ratchet jobs finish green on the PR: 3 checks, exit 0.
git diff --stat shows 0 files outside the Boundary and the ledgers those three checks read.
Activity