Beagle SCM

POST-019: be post output noise — drop the be: post … echo + name the bump put

be post on a parent with an advanced submodule emits noise: dispatcher echo lines be: post . / be: post abc (pure progress chatter), and an anonymous put put: + staged 1 put row(s) block. The mid-post put itself is FINE — it stages the bumped sub gitlink so the parent commit records the new tip — but it names nothing, so the user can't tell it staged abc/. Drop the be: echo; make the put SAY what it put. See POST-018, BE-005, Submodules.

Input

Context

Observed be post 'JS: tcp/udp network bindings' from ~/beagle with an advanced abc submodule; a put: block appears in the middle of the post output.

• Verbatim: parent recurses post abc/?00e5d050#… (mod abc/UDP.h), THEN put put: + staged 1 put row(s), THEN parent post ?803b3fae#… whose rows already include mod abc (the same gitlink bump).
• The interleaved put is the gitlink bump (put <sub>#<40-hex>, sniff/POST.c:234 / 485) staging the new abc tip into the parent before the parent commit.
• Its banner is the POST-018 ROWS routing (put_feed_summary_tail, sniff/PUT.c:652), suppressed only under SNIFF.quiet; the parent-scope bump runs non-quiet.

Goals

The internal post-driven gitlink-bump staging must not print a user-facing put: banner inside a post stream.

• The sub-bump shows ONLY as mod <sub> in the parent post rows (already there).
• Suppress the standalone put put: table + staged N put row(s) for the internal gitlink-bump path; leave explicit be put UX unchanged.
• Repro test: be post with an advanced sub asserts NO put: banner between the sub-post and the parent-post (table-driven, test/post/ suite).

Constraints

• Do not change WHAT lands — the gitlink bump must still be committed; only its banner is suppressed.
• Confirm the mechanism by trace before fixing (recursion-debug practice); the SNIFF.quiet guard at PUT.c:657 is the likely lever, not yet trace-confirmed.

WIP

Design decisions

As shipped: the internal bump runs be put -q <sub> with stdout CAPTURED to a pipe (BE.cli.c::bepost_bump_sub, :3028-3098); under -q PUTStage opens the HUNK table with EMPTY state so a no-op sub flushes no banner (sniff/PUT.c:1174). The be: post <sub> stderr echo is gone — a clean/no-op sub instead emits a ULOG post <sub> marker row on stdout (BE.cli.c:3628-3639, bepost_emit_outer).

TODOs

• Trace the put call site + why it prints non-quiet (fixed at the source: quiet empty-state banner + captured-pipe stdout, PUT.c:1174 / BE.cli.c:3034).
• Repro coverage in test/post/: post/09, post/10, post/23 assert the sub descent reports via stdout ROWS/marker rows, no stderr echo, no put banner.
• Gate the put: banner off for the internal gitlink-bump path; ctest green.

Blockers and bummers

(none)

Outcome

• LANDED (audit 2026-07-07): no POST-019-tagged commit in jab log; the fix is verified in-tree at tip b0fb0ce4 (POST-019 comment tags at BE.cli.c:20/ 3342/3628/3755/3843, sniff/PUT.c:1180), carried by the untagged 21-26Jun batches (SUBS-029 3fd64016/c2e1b0d9 reworked bepost_bump_sub; the 25/26Jun 1fd33d9f/0447e4a7 tool-fix batch).
• Both goals shipped: internal gitlink-bump put: banner suppressed (sub shows only as mod <sub> in the parent rows); be: post ./be: post <sub> echo replaced by the stdout post <sub> marker row. Tests post/09, post/10, post/23.