Repository navigation
fix(sutando-migrate): use uname-s branch for preflight byte count (closes #1474) - #1476
Conversation
…oses sonichi#1474) Replace the BSD stat -f '%z' + heuristic Linux fallback with an explicit uname -s case branch in preflight_summary(). Bug: on Linux, stat -f '%z' means 'filesystem stat' (GNU), not file size. The xargs pipeline produces non-empty, non-zero output (filesystem info), so the [ -z "$bytes" ] fallback trigger never fires → wrong byte count → misleading ETA estimate. Fix: branch by uname -s — Darwin uses stat -f '%z' (BSD file size), Linux/* uses find -printf '%s\n' (GNU). Same approach as stage-readiness.sh and catchup-after-startup.sh already use for cross-platform stat. Test: added Test 8 asserting the uname-s branch exists and the old heuristic fallback is removed. Cosmetic severity only — data integrity and copy correctness are unaffected.
qingyun-wu
left a comment
qingyun-wu
left a comment
There was a problem hiding this comment.
Cold-review from Studio (qingyun-sutando, the bot that filed the original #1474). LGTM with one tiny style nit (non-blocking).
What's right:
-
Code matches my proposed fix in #1474 byte-for-byte. The
case "$(uname -s)"shape + Darwin/Linux|* arms + drop of the heuristic[ -z "$bytes" ] || [ "$bytes" = "0" ]fallback are exactly what I suggested. Net effect on Linux: deterministic byte-count viafind -printf '%s\n'instead of a guessing-game fallback that doesn't trigger when GNUstat -fproduces non-empty garbage. -
Linux|*)default arm correctly catches FreeBSD-without-Darwin, OpenBSD, NetBSD, WSL, etc. — anywhere with GNU find. The same fall-through-to-GNU pattern is used inscripts/stage-readiness.shandskills/catchup-after-startup/scripts/catchup-after-startup.sh. -
Test 8 has both positive (new branch present) AND negative (old fallback comment + trigger removed) assertions — guards against a future revert that silently re-introduces the BSD-stat-on-Linux bug.
-
No collateral damage to the other ~15
stat -f %z … || stat -c %scalls insutando-migrate.sh: those use the short-circuit||pattern which is portable (BSD stat exits non-zero on Linux → falls through tostat -c). The preflight call was unique in chaining viaxargs -0which broke the short-circuit. PR body's note on this is accurate.
Non-blocking nit: the Linux|*) branch could in principle drop the 2>/dev/null on find -printf since GNU find supports it natively — but matching the pre-existing error-suppression idiom in the same function is consistent (the Darwin arm also has 2>/dev/null on the BSD path), so leave as-is. Purely style.
Net: ship it. Sonichi's approval already covers the architectural call; this comment is for the record from the issue-filer.
— Studio (qingyun-sutando)
…rator checklist + workspace-sync node_modules Per Chi's PR #1472 comments (#design 2026-06-06): **CHANGELOG.md v0.3.0 additions:** - Fixed → Migration subsection: Lucy's Maddy v0.8 5-bug sweep (#1475 #1478 #1482 #1483 #1484) - Fixed → Cross-platform: #1476 (uname -s branch for preflight byte count) - Fixed → Security: #1452 #1462 #1466 (execSync → execFileSync sweep) - Added: #1467 (bridges inject skill instructions into owner task files) - Workspace contract rollup → M1: also reference #1473 (pre-flight scan + per-file progress reporter) **KNOWN_ISSUES.md additions** (from migration-thread digest 1512359988834730014): - Pre-migration operator checklist (7 items) — vault URL, remove shell-rc env var, grep .git/info/exclude, confirm wsId, stagger hosts, restart services, post-migration sanity check - Identity-collision rule — exactly one host at a time can hold a Discord token (also applies to Twilio + other per-identity creds) **docs/workspace-sync.md** pre-flight checklist: - Added `node_modules/` row per Chi's PR #1472 comment (covers vault-sync side; #1482 already handles the backup-side via `_BACKUP_DEFAULT_EXCLUDES`) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Summary — closes #1474
Replace the BSD
stat -f '%z'+ heuristic Linux fallback with an explicituname -scase branch inpreflight_summary().Bug
On Linux,
stat -f '%z'means "filesystem stat" (GNU semantics), not file size. Thexargs -0 stat -f '%z'pipeline produces non-empty, non-zero output (filesystem block info), so the[ -z "$bytes" ] || [ "$bytes" = "0" ]fallback trigger never fires. Result: wrong byte count → misleading ETA estimate.Fix
Branch by
uname -s— exactly as proposed in #1474:stat -f '%z'(BSD file size via xargs pipeline)find -printf '%s\\n'(GNU, no stat needed)This is the same cross-platform pattern already used by
stage-readiness.shandcatchup-after-startup.sh.Why the other
stat -fcalls are fineThe other 15+
stat -fcalls in sutando-migrate.sh all use thestat -f %z ... || stat -c %s ...short-circuit pattern: on Linux the first command exits non-zero (error), so||triggers the GNU fallback correctly. Only the preflight byte count usedxargspiping, which produces wrong-but-non-empty output on Linux.Test
uname -sbranch exists,Darwin)andLinux|*)cases present, old heuristic fallback removedALL TESTS PASSScope
scripts/sutando-migrate.sh(5 lines changed),tests/sutando-migrate-preflight.test.sh(11 lines added)staging-workspace-revamp(same branch as PR feat(sutando-migrate): pre-flight scan + per-file progress reporter [DRAFT] #1473 which introduced preflight_summary)