kayg/calternal
Watch
1
0
Fork
You've already forked calternal
 0
Code Issues 190 Pull requests 61 Projects Releases Packages Wiki Activity Actions 26

build: rusqlite 0.41 (libsqlite3-sys 0.39) for RUSTSEC-2026-0459 #1043

Closed
kayg wants to merge 4 commits from fix/rusqlite-0.41-rustsec-2026-0459 into dev
pull from: fix/rusqlite-0.41-rustsec-2026-0459
merge into: kayg:dev
Conversation 4 Commits 4 Files changed 16 +47 -31
kayg commented 2026-11-09 18:24:35 +00:00
Owner
Copy link

Fixes RUSTSEC-2026-0459 (bundled SQLite in libsqlite3-sys, fixed in >= 0.39).

Change

  • One workspace dependency: rusqlite = { version = "0.41", features = ["bundled"] } in the root Cargo.toml; the 12 crates use workspace = true (search adds hooks).
  • Lock: rusqlite 0.40.2 -> 0.41.1, libsqlite3-sys 0.38.2 -> 0.39.1 (bundled SQLite 3.53.2 -> 3.54.1).
  • .cargo/config.toml (LIBSQLITE3_FLAGS=-DHAVE_FDATASYNC=1) and every PRAGMA are untouched.

0.41 breaking changes hit

  1. Connection::progress_handler now takes &mut self (used by the search time limit in read::ranked). read::search/candidates/ranked take &mut Connection; the vocabulary closure now receives the connection.
  2. Transaction still only derefs to &Connection, so the search reader's unchecked_transaction() guard can't give the &mut. The read snapshot is opened with BEGIN DEFERRED / ROLLBACK by hand; a connection that can't leave the transaction is dropped, not pooled.

No other call site in the 12 crates needed changes.

Gate results and perf tables follow in comments.

🤖 Generated with Claude Code

Fixes RUSTSEC-2026-0459 (bundled SQLite in libsqlite3-sys, fixed in >= 0.39). ## Change - One workspace dependency: `rusqlite = { version = "0.41", features = ["bundled"] }` in the root `Cargo.toml`; the 12 crates use `workspace = true` (search adds `hooks`). - Lock: rusqlite 0.40.2 -> 0.41.1, libsqlite3-sys 0.38.2 -> 0.39.1 (bundled SQLite 3.53.2 -> 3.54.1). - `.cargo/config.toml` (`LIBSQLITE3_FLAGS=-DHAVE_FDATASYNC=1`) and every PRAGMA are untouched. ## 0.41 breaking changes hit 1. `Connection::progress_handler` now takes `&mut self` (used by the search time limit in `read::ranked`). `read::search`/`candidates`/`ranked` take `&mut Connection`; the vocabulary closure now receives the connection. 2. `Transaction` still only derefs to `&Connection`, so the search reader's `unchecked_transaction()` guard can't give the `&mut`. The read snapshot is opened with `BEGIN DEFERRED` / `ROLLBACK` by hand; a connection that can't leave the transaction is dropped, not pooled. No other call site in the 12 crates needed changes. Gate results and perf tables follow in comments. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
rusqlite 0.41 pulls libsqlite3-sys >= 0.39, which carries the SQLite fix
for RUSTSEC-2026-0459. The bundled feature stays on in the one workspace
entry; search adds hooks on top.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Breaking change hit: Connection::progress_handler now takes &mut self.
The ranked() time limit is the only caller, so read::search and
candidates take &mut Connection, and the typo vocabulary closure gets the
connection as an argument instead of borrowing the snapshot.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
search: open the read snapshot by hand (Transaction has no &mut access)
Some checks are pending
CI / server (pull_request) Waiting to run
Details
CI / rust (pull_request) Waiting to run
Details
CI / web (pull_request) Waiting to run
Details
Markdown extraction contract / markdown-native-contract (pull_request) Waiting to run
Details
Markdown extraction contract / markdown-editor-contract (pull_request) Waiting to run
Details
9cfdaaa1e6 
Second 0.41 break fallout: Transaction only derefs to &Connection, and
progress_handler now needs &mut Connection. BEGIN DEFERRED / ROLLBACK on
the reader keeps the one-snapshot guarantee. A connection that could not
leave the transaction is dropped, not pooled.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
kayg commented 2026-11-09 19:24:50 +00:00
Author
Owner
Copy link

Gates at head 60306ce08 (dev-01 calternal-dev-01-compute02, via ~/bin/slot)

Gate Result
cargo fmt --check OK
cargo clippy -j4 --workspace --all-targets --locked -- -D warnings exit 0
cargo test -j4 --workspace --locked exit 0, 127 test binaries ok, 0 failed
cargo audit (cargo-audit 0.22.2, same advisory DB, dev base vs PR) RUSTSEC-2026-0459 gone. Vulnerabilities 5 -> 4; the 4 left are old: RUSTSEC-2026-0381 webpki-roots, -0382 rustls-webpki, -0414 and -0433 pulldown-cmark. Warnings identical (unsafe-libyaml, smallstr, tokio unsound). No new advisories.

SQLite before / after (in-memory probe, LIBSQLITE3_FLAGS="-DHAVE_FDATASYNC=1 -DSQLITE_ENABLE_MEMORY_MANAGEMENT")

  • sqlite_version(): 3.53.2 -> 3.54.1 (rusqlite 0.40.2 / libsqlite3-sys 0.38.2 -> 0.41.1 / 0.39.1).
  • PRAGMA compile_options diff: only the new MAX_SCHEMA=10000000 line (SQLite's own default) and the version.
  • Finding: libsqlite3-sys 0.39 stopped passing -DSQLITE_ENABLE_MEMORY_MANAGEMENT. The first probe showed it missing (our PRAGMA shrink_memory calls depend on it), so commit 60306ce08 adds it to LIBSQLITE3_FLAGS next to -DHAVE_FDATASYNC=1. The probe above is after that fix; bundled stays on.
  • PRAGMAs in our code are untouched.

Size and build time (release, fat LTO, -p calternal -p calternal-perf -p calternal-mail-bench, cold target/release)

dev 0878cb550 PR 60306ce08
calternal unstripped 31,992,520 B 32,036,216 B (+43.7 KB, +0.14%)
calternal stripped 27,292,144 B 27,328,624 B (+36.5 KB, +0.13%)
cold build wall 687 s 726 s

The build wall times ran on a busy shared host (load avg ~30), so they are noisy; user CPU was 1596 s vs 1717 s (+7.6%), which is within what I would expect from load, not measured as a regression.

Perf ABAB is queued behind other lanes (FIFO) and will be posted next.

## Gates at head 60306ce08 (dev-01 `calternal-dev-01-compute02`, via `~/bin/slot`) | Gate | Result | |---|---| | `cargo fmt --check` | OK | | `cargo clippy -j4 --workspace --all-targets --locked -- -D warnings` | exit 0 | | `cargo test -j4 --workspace --locked` | exit 0, 127 test binaries `ok`, 0 failed | | `cargo audit` (cargo-audit 0.22.2, same advisory DB, dev base vs PR) | RUSTSEC-2026-0459 gone. Vulnerabilities 5 -> 4; the 4 left are old: RUSTSEC-2026-0381 webpki-roots, -0382 rustls-webpki, -0414 and -0433 pulldown-cmark. Warnings identical (unsafe-libyaml, smallstr, tokio unsound). No new advisories. | ## SQLite before / after (in-memory probe, `LIBSQLITE3_FLAGS="-DHAVE_FDATASYNC=1 -DSQLITE_ENABLE_MEMORY_MANAGEMENT"`) - `sqlite_version()`: **3.53.2 -> 3.54.1** (rusqlite 0.40.2 / libsqlite3-sys 0.38.2 -> 0.41.1 / 0.39.1). - `PRAGMA compile_options` diff: only the new `MAX_SCHEMA=10000000` line (SQLite's own default) and the version. - Finding: libsqlite3-sys 0.39 **stopped passing `-DSQLITE_ENABLE_MEMORY_MANAGEMENT`**. The first probe showed it missing (our `PRAGMA shrink_memory` calls depend on it), so commit 60306ce08 adds it to `LIBSQLITE3_FLAGS` next to `-DHAVE_FDATASYNC=1`. The probe above is after that fix; bundled stays on. - PRAGMAs in our code are untouched. ## Size and build time (release, fat LTO, `-p calternal -p calternal-perf -p calternal-mail-bench`, cold `target/release`) | | dev 0878cb550 | PR 60306ce08 | |---|---|---| | `calternal` unstripped | 31,992,520 B | 32,036,216 B (+43.7 KB, +0.14%) | | `calternal` stripped | 27,292,144 B | 27,328,624 B (+36.5 KB, +0.13%) | | cold build wall | 687 s | 726 s | The build wall times ran on a busy shared host (load avg ~30), so they are noisy; user CPU was 1596 s vs 1717 s (+7.6%), which is within what I would expect from load, not measured as a regression. Perf ABAB is queued behind other lanes (FIFO) and will be posted next.
kayg commented 2026-11-10 05:18:48 +00:00
Author
Owner
Copy link

Perf: ABAB x5 vs dev, with a dev-copy control (perf-01/perf-02 via perf-turn, PR head 60306ce08 vs dev 0878cb550)

Setup: 5 reps, rotating order (dev pr ctl, pr ctl dev, ctl dev pr, ...), dev-ctl = a second copy of the dev binaries. Per rep and arm: calternal-perf (10k notes, 3 warm runs, 2 cold runs, 3000-edit storm; covers cold index/first page, home/notes open/list/append, search, sync round trip, startup, trash, uploads, memory), tasks-bench (10k), storage-bench history/save, mail-bench (100k). Table = median over 5 reps; verdict flags pr worse than dev by >3% AND more than the control's gap / half the dev spread. 169 metrics compared, 12 flagged (3 are "better").

Where pr is worse (decision needed)

metric dev dev-ctl pr pr vs dev per-rep (dev / pr)
mail list_page p50 @100k 46.1 us 47.1 us 48.2 us +4.5% 45.8-47.3 / 47.9-48.3, every rep worse
mail unified_page p50 45.9 us 46.6 us 48.1 us +5.0%
mail account_page p50 56.8 us 58.2 us 58.9 us +3.6%
tasks describe pass_and_rows p95 70.6 us 70.9 us 97.2 us +38% dev 66-73 / pr 95-102, all 5 reps
tasks describe pass_and_rows p50 38.8 us 38.0 us 39.9 us +2.8% (reps 1-2: +9%)
tasks describe pass p50 13.1 us 12.8 us 13.8 us +5.0%
tasks list untracked p50 1.24 ms 1.25 ms 1.29 ms +3.6%
idle RSS p50 18.8 MB 18.8 MB 19.0 MB (+0.2) +1.2% (p95 +3.9%)
RSS after first page p50 21.3 MB 21.3 MB 22.0 MB +3.4% every rep pr higher
trash restore p95 106.4 ms 106.3 ms 111.6 ms +5.0% spread 7%, borderline

Pattern: microsecond-scale point lookups and the small per-note DB writes, about +2 us on ~46 us (~30 us/lookup is what infraternal measured for SQLite 3.54.1 itself) plus ~0.3-0.7 MB RSS. Not tuned around; I did not find an isolated cause because rusqlite 0.41 and SQLite 3.53.2 -> 3.54.1 change together (a split build would need another perf turn; say if you want it).

Where pr is better or flat

All user-visible paths are flat within the control: notes open p50 0.72 ms, home page p50 -6%, tag page p50 -6.8%, search query p50 1.45 -> 1.44 ms (p95 -4.9%), search first_query, sync round trip p50 0.212 -> 0.207 ms, save ack, typing, storm edits/s, durable write p50 24.9 -> 24.7 ms, cold index p50 1.465 -> 1.422 s, cold first page p50 319 -> 293 ms, cold startup p50 19.9 -> 17.6 ms, healthz startup, storage-bench save p50/p99, mail thread_open and storm. Rest of the table is in the collapsed block.

All 169 metrics (median of 5 reps)

reps=5

metric dev dev-ctl pr ctl vs dev % pr vs dev % dev spread % verdict
mail.mail.account_page@1accounts@100000.p50 0.05679 0.05816 0.05883 +2.4 +3.6 3.7 REGRESS?
mail.mail.account_page@1accounts@100000.p95 0.07195 0.07123 0.07423 -1.0 +3.2 8.8
mail.mail.build_s@100000.p50 3.521 3.527 3.568 +0.2 +1.3 1.1
mail.mail.build_s@100000.p95 3.521 3.527 3.568 +0.2 +1.3 1.1
mail.mail.fake_headers_per_s.p50 2.105e+04 2.09e+04 2.105e+04 -0.7 -0.0 0.5
mail.mail.fake_headers_per_s.p95 2.105e+04 2.09e+04 2.105e+04 -0.7 -0.0 0.5
mail.mail.list_first@100000.p50 0.04045 0.04111 0.04164 +1.6 +2.9 3.0
mail.mail.list_first@100000.p95 0.04899 0.05245 0.05035 +7.1 +2.8 8.4
mail.mail.list_page@100000.p50 0.04613 0.04706 0.04821 +2.0 +4.5 3.3 REGRESS?
mail.mail.list_page@100000.p95 0.05731 0.05912 0.05823 +3.2 +1.6 7.6
mail.mail.thread_open@100000.p50 0.04673 0.04774 0.04788 +2.2 +2.5 3.1
mail.mail.thread_open@100000.p95 0.06332 0.06615 0.06392 +4.5 +0.9 3.8
mail.mail.thread_storm_merges@newest@1000000.p50 6.155e+05 6.155e+05 6.155e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@newest@1000000.p95 6.155e+05 6.155e+05 6.155e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@newest@250000.p50 1.535e+05 1.535e+05 1.535e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@newest@250000.p95 1.535e+05 1.535e+05 1.535e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@newest@500000.p50 3.074e+05 3.074e+05 3.074e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@newest@500000.p95 3.074e+05 3.074e+05 3.074e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@oldest@1000000.p50 6.155e+05 6.155e+05 6.155e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@oldest@1000000.p95 6.155e+05 6.155e+05 6.155e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@oldest@250000.p50 1.535e+05 1.535e+05 1.535e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@oldest@250000.p95 1.535e+05 1.535e+05 1.535e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@oldest@500000.p50 3.074e+05 3.074e+05 3.074e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@oldest@500000.p95 3.074e+05 3.074e+05 3.074e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@shuffled@1000000.p50 6.155e+05 6.155e+05 6.155e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@shuffled@1000000.p95 6.155e+05 6.155e+05 6.155e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@shuffled@250000.p50 1.535e+05 1.535e+05 1.535e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@shuffled@250000.p95 1.535e+05 1.535e+05 1.535e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@shuffled@500000.p50 3.074e+05 3.074e+05 3.074e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_merges@shuffled@500000.p95 3.074e+05 3.074e+05 3.074e+05 +0.0 +0.0 0.0
mail.mail.thread_storm_s@newest@1000000.p50 2.585 2.612 2.564 +1.0 -0.8 1.8
mail.mail.thread_storm_s@newest@1000000.p95 2.585 2.612 2.564 +1.0 -0.8 1.8
mail.mail.thread_storm_s@newest@250000.p50 0.6123 0.6168 0.5998 +0.7 -2.0 4.2
mail.mail.thread_storm_s@newest@250000.p95 0.6123 0.6168 0.5998 +0.7 -2.0 4.2
mail.mail.thread_storm_s@newest@500000.p50 1.254 1.264 1.243 +0.8 -0.9 2.8
mail.mail.thread_storm_s@newest@500000.p95 1.254 1.264 1.243 +0.8 -0.9 2.8
mail.mail.thread_storm_s@oldest@1000000.p50 2.608 2.649 2.582 +1.6 -1.0 3.5
mail.mail.thread_storm_s@oldest@1000000.p95 2.608 2.649 2.582 +1.6 -1.0 3.5
mail.mail.thread_storm_s@oldest@250000.p50 0.6079 0.6209 0.6042 +2.1 -0.6 3.8
mail.mail.thread_storm_s@oldest@250000.p95 0.6079 0.6209 0.6042 +2.1 -0.6 3.8
mail.mail.thread_storm_s@oldest@500000.p50 1.259 1.267 1.251 +0.7 -0.6 2.4
mail.mail.thread_storm_s@oldest@500000.p95 1.259 1.267 1.251 +0.7 -0.6 2.4
mail.mail.thread_storm_s@shuffled@1000000.p50 2.887 2.887 2.855 +0.0 -1.1 2.7
mail.mail.thread_storm_s@shuffled@1000000.p95 2.887 2.887 2.855 +0.0 -1.1 2.7
mail.mail.thread_storm_s@shuffled@250000.p50 0.6493 0.6647 0.6414 +2.4 -1.2 5.7
mail.mail.thread_storm_s@shuffled@250000.p95 0.6493 0.6647 0.6414 +2.4 -1.2 5.7
mail.mail.thread_storm_s@shuffled@500000.p50 1.361 1.376 1.356 +1.1 -0.4 3.6
mail.mail.thread_storm_s@shuffled@500000.p95 1.361 1.376 1.356 +1.1 -0.4 3.6
mail.mail.unified_page@1accounts@100000.p50 0.04587 0.04661 0.04815 +1.6 +5.0 3.8 REGRESS?
mail.mail.unified_page@1accounts@100000.p95 0.05913 0.05784 0.05825 -2.2 -1.5 28.2
perf.cold.first_page.p50 319.3 312.5 305.3 -2.1 -4.4 15.3
perf.cold.first_page.p95 335.6 332.6 324.9 -0.9 -3.2 25.8
perf.cold.index.p50 1.465 1.449 1.444 -1.1 -1.4 24.6
perf.cold.index.p95 1.552 1.464 1.468 -5.6 -5.4 38.5
perf.cold.mem_peak.p50 142.9 142.9 143.9 +0.0 +0.7 1.1
perf.cold.mem_peak.p95 143.6 143.8 144.4 +0.2 +0.5 1.5
perf.cold.notes_per_s.p50 6445 6831 6810 +6.0 +5.7 31.1
perf.cold.notes_per_s.p95 6826 6899 6924 +1.1 +1.4 20.4
perf.cold.startup.p50 19.84 20.1 17.69 +1.3 -10.8 23.0
perf.cold.startup.p95 20.16 21.91 17.9 +8.7 -11.2 27.6
perf.home.first_page.p50 2.619 2.381 2.259 -9.1 -13.7 151.7
perf.home.first_page.p95 4.085 2.617 4.092 -35.9 +0.2 103.8
perf.home.page.p50 0.758 0.7539 0.7125 -0.5 -6.0 8.6 better
perf.home.page.p95 2.876 2.896 2.993 +0.7 +4.1 111.3
perf.home.page_big.p50 16.57 16.27 16.6 -1.8 +0.2 3.9
perf.home.page_big.p95 18.14 18.57 19.23 +2.4 +6.0 12.9
perf.home.page_next.p50 0.7625 0.7989 0.8043 +4.8 +5.5 14.7
perf.home.page_next.p95 1.121 1.195 1.168 +6.6 +4.1 24.2
perf.home.page_tag.p50 5.321 5.14 4.958 -3.4 -6.8 5.7 better
perf.home.page_tag.p95 7.249 6.469 6.862 -10.8 -5.3 11.3
perf.home.page_title.p50 0.6474 0.6316 0.6388 -2.4 -1.3 5.9
perf.home.page_title.p95 0.7765 0.7368 0.7403 -5.1 -4.7 13.2
perf.journal.catch_up.p50 28.15 28.18 28.61 +0.1 +1.6 7.1
perf.journal.catch_up.p95 29.01 29.15 29.73 +0.5 +2.5 7.6
perf.journal.entries.p50 1.018e+04 1.018e+04 1.018e+04 +0.0 +0.0 0.0
perf.journal.entries.p95 1.036e+04 1.036e+04 1.036e+04 +0.0 +0.0 0.0
perf.journal.page.p50 2.4 2.382 2.398 -0.8 -0.1 4.7
perf.journal.page.p95 2.685 2.769 2.718 +3.1 +1.2 15.9
perf.mem.after_first_page.p50 21.33 21.3 22.05 -0.1 +3.4 3.0 REGRESS?
perf.mem.after_first_page.p95 22 21.73 22.44 -1.3 +2.0 3.4
perf.mem.after_opens.p50 41.73 41.14 41.74 -1.4 +0.0 6.1
perf.mem.after_opens.p95 41.84 42.12 42.19 +0.7 +0.8 6.2
perf.mem.after_search.p50 47.79 47.85 47.25 +0.1 -1.1 8.3
perf.mem.after_search.p95 75.29 75.34 75.65 +0.1 +0.5 0.9
perf.mem.idle.p50 18.78 18.77 19 -0.0 +1.2 1.2
perf.mem.idle.p95 19.05 19.32 19.79 +1.4 +3.9 4.7 REGRESS?
perf.mem.peak.p50 47.32 47.59 46.99 +0.6 -0.7 8.6
perf.mem.peak.p95 74.89 74.91 75.32 +0.0 +0.6 0.8
perf.open.api_get.p50 3.924 3.976 3.877 +1.3 -1.2 10.9
perf.open.api_get.p95 9.743 8.961 9.396 -8.0 -3.6 56.2
perf.open.long_note.p50 15.84 15.82 16.05 -0.1 +1.3 211.7
perf.open.long_note.p95 53.39 52.11 53.11 -2.4 -0.5 13.9
perf.open.note.p50 0.7156 0.7009 0.7073 -2.1 -1.2 8.7
perf.open.note.p95 1.316 1.299 1.377 -1.3 +4.6 22.5
perf.search.catch_up.p50 2.005 2.005 2.005 -0.0 -0.0 0.0
perf.search.catch_up.p95 4.013 4.009 4.006 -0.1 -0.2 0.2
perf.search.first_query.p50 4.942 5.437 4.577 +10.0 -7.4 56.3
perf.search.first_query.p95 26.52 29.22 26.28 +10.2 -0.9 5.5
perf.search.query.p50 1.446 1.446 1.435 -0.0 -0.8 6.2
perf.search.query.p95 3.974 4.011 3.942 +0.9 -0.8 15.0
perf.shutdown.p50 64.02 64.08 64.01 +0.1 -0.0 33.1
perf.shutdown.p95 64.24 85.28 84.18 +32.8 +31.0 36.1
perf.startup.healthz.p50 4.997 4.964 5.066 -0.6 +1.4 44.7
perf.startup.healthz.p95 5.415 6.009 5.693 +11.0 +5.1 55.8
perf.storm.converge.p50 159.5 159.7 159.2 +0.2 -0.2 6.6
perf.storm.converge.p95 176.5 175.7 174.9 -0.4 -0.9 10.3
perf.storm.edits_per_s.p50 3845 3854 3903 +0.2 +1.5 4.5
perf.storm.edits_per_s.p95 3948 3976 3973 +0.7 +0.6 5.1
perf.storm.outside_edits.p50 2 2 2 +0.0 +0.0 0.0
perf.storm.outside_edits.p95 2 2 2 +0.0 +0.0 0.0
perf.storm.save_ack.p50 7.296 7.247 7.289 -0.7 -0.1 2.0
perf.storm.save_ack.p95 11.86 11.34 11.62 -4.4 -2.0 12.0
perf.storm.typing.p50 0.7802 0.7784 0.7686 -0.2 -1.5 4.6
perf.storm.typing.p95 0.786 0.8078 0.7855 +2.8 -0.1 6.1
perf.sync.durable_write.p50 24.88 24.7 24.67 -0.7 -0.8 1.2
perf.sync.durable_write.p95 37.04 32.69 33.32 -11.8 -10.0 16.9
perf.sync.round_trip.p50 0.2138 0.2133 0.2074 -0.3 -3.0 8.8
perf.sync.round_trip.p95 0.3494 0.3382 0.3474 -3.2 -0.6 12.3
perf.sync.save_ack.p50 3.815 3.723 3.691 -2.4 -3.3 6.8
perf.sync.save_ack.p95 7.401 5.537 5.517 -25.2 -25.5 38.9 better
perf.trash.delete.p50 64.08 72.78 67.79 +13.6 +5.8 24.9
perf.trash.delete.p95 98.77 96.5 98.94 -2.3 +0.2 5.8
perf.trash.forget.p50 11.49 11.64 11.54 +1.3 +0.4 3.1
perf.trash.forget.p95 15.45 15.21 15.91 -1.6 +3.0 44.0
perf.trash.restore.p50 63.41 63.53 64.52 +0.2 +1.7 6.9
perf.trash.restore.p95 106.4 106.3 111.6 -0.0 +5.0 7.0 REGRESS?
perf.uploads.response.p50 142.8 142 142.9 -0.6 +0.0 4.4
perf.uploads.response.p95 165 163.8 163.9 -0.7 -0.6 12.8
perf.uploads.round_trip.p50 0.5014 0.5016 0.4939 +0.0 -1.5 7.0
perf.uploads.round_trip.p95 0.864 0.7296 0.7589 -15.6 -12.2 18.0
perf.uploads.save_ack.p50 4.23 4.196 4.139 -0.8 -2.2 2.9
perf.uploads.save_ack.p95 8.41 8.5 8.537 +1.1 +1.5 17.2
sb-save.max 60.56 63.67 59.41 +5.1 -1.9 42.4
sb-save.p50 20.19 20.6 20.73 +2.0 +2.7 3.5
sb-save.p90 27.73 28.04 28.66 +1.1 +3.4 8.1
sb-save.p99 37.46 36.66 37.77 -2.1 +0.8 10.4
tasks.cold_plain_s 1.796 1.777 1.8 -1.0 +0.2 3.6
tasks.cold_tasks_s 2.14 2.104 2.148 -1.7 +0.4 3.2
tasks/cold_index_ms/growth_pct 19.27 18.09 20.37 -6.1 +5.7 13.1
tasks/cold_index_ms/plain 1796 1777 1800 -1.0 +0.2 3.6
tasks/cold_index_ms/tasks 2140 2104 2148 -1.7 +0.4 3.2
tasks/describe_ms/pass.p50 0.0131 0.01281 0.01376 -2.1 +5.0 6.0 REGRESS?
tasks/describe_ms/pass.p95 0.03557 0.0351 0.03607 -1.3 +1.4 8.6
tasks/describe_ms/pass_and_rows.p50 0.03899 0.03799 0.03994 -2.6 +2.4 4.2
tasks/describe_ms/pass_and_rows.p95 0.07055 0.07091 0.09723 +0.5 +37.8 9.8 REGRESS?
tasks/list_ms/anytime.p50 1.824 1.819 1.843 -0.3 +1.0 9.6
tasks/list_ms/anytime.p95 2.171 2.032 2.064 -6.4 -4.9 24.8
tasks/list_ms/anytime_by_child_tag.p50 2.232 2.316 2.288 +3.7 +2.5 1.7
tasks/list_ms/anytime_by_child_tag.p95 2.61 2.769 2.756 +6.1 +5.6 7.4
tasks/list_ms/counts.p50 0.2496 0.2605 0.2544 +4.4 +1.9 17.5
tasks/list_ms/counts.p95 0.2741 0.2942 0.2923 +7.3 +6.6 30.7
tasks/list_ms/someday.p50 1.795 1.82 1.833 +1.4 +2.1 5.9
tasks/list_ms/someday.p95 2.089 2.13 2.159 +1.9 +3.3 11.0
tasks/list_ms/tag.p50 2.19 2.221 2.246 +1.4 +2.6 8.8
tasks/list_ms/tag.p95 2.528 2.549 2.654 +0.8 +5.0 20.2
tasks/list_ms/today.p50 2.317 2.28 2.301 -1.6 -0.7 6.8
tasks/list_ms/today.p95 2.664 2.539 2.685 -4.7 +0.8 22.6
tasks/list_ms/today_by_priority.p50 2.152 2.207 2.157 +2.6 +0.2 5.4
tasks/list_ms/today_by_priority.p95 2.461 2.605 2.456 +5.8 -0.2 13.6
tasks/list_ms/untracked.p50 1.242 1.248 1.286 +0.5 +3.6 6.4 REGRESS?
tasks/list_ms/untracked.p95 1.359 1.457 1.428 +7.2 +5.0 9.6
tasks/list_ms/upcoming.p50 2.29 2.262 2.333 -1.2 +1.9 4.3
tasks/list_ms/upcoming.p95 2.614 2.582 2.639 -1.2 +1.0 10.8
tasks/notes 1e+04 1e+04 1e+04 +0.0 +0.0 0.0
tasks/tasks 1e+04 1e+04 1e+04 +0.0 +0.0 0.0
tasks/write_ms/first_touch.p50 15.53 15.53 15.81 -0.0 +1.8 4.1
tasks/write_ms/first_touch.p95 64.7 66.9 68.01 +3.4 +5.1 16.9
tasks/write_ms/warm.p50 4.858 4.889 4.97 +0.6 +2.3 1.1
tasks/write_ms/warm.p95 14.3 15.16 14.5 +6.0 +1.4 8.0

9 flagged of 169

Run dir on perf-01: ~/perf/out-rusq-1 (reps 1-3 ran before an API stop; rep 4 was rerun from scratch). Same control gap shows run-to-run noise of ~1-3% on the small metrics.

## Perf: ABAB x5 vs dev, with a dev-copy control (perf-01/perf-02 via `perf-turn`, PR head 60306ce08 vs dev 0878cb550) Setup: 5 reps, rotating order (`dev pr ctl`, `pr ctl dev`, `ctl dev pr`, ...), `dev-ctl` = a second copy of the dev binaries. Per rep and arm: `calternal-perf` (10k notes, 3 warm runs, 2 cold runs, 3000-edit storm; covers cold index/first page, home/notes open/list/append, search, sync round trip, startup, trash, uploads, memory), `tasks-bench` (10k), `storage-bench history/save`, `mail-bench` (100k). Table = median over 5 reps; verdict flags pr worse than dev by >3% AND more than the control's gap / half the dev spread. 169 metrics compared, 12 flagged (3 are "better"). ### Where pr is worse (decision needed) | metric | dev | dev-ctl | pr | pr vs dev | per-rep (dev / pr) | |---|---|---|---|---|---| | mail list_page p50 @100k | 46.1 us | 47.1 us | 48.2 us | +4.5% | 45.8-47.3 / 47.9-48.3, every rep worse | | mail unified_page p50 | 45.9 us | 46.6 us | 48.1 us | +5.0% | | | mail account_page p50 | 56.8 us | 58.2 us | 58.9 us | +3.6% | | | tasks describe pass_and_rows p95 | 70.6 us | 70.9 us | 97.2 us | **+38%** | dev 66-73 / pr 95-102, all 5 reps | | tasks describe pass_and_rows p50 | 38.8 us | 38.0 us | 39.9 us | +2.8% (reps 1-2: +9%) | | | tasks describe pass p50 | 13.1 us | 12.8 us | 13.8 us | +5.0% | | | tasks list untracked p50 | 1.24 ms | 1.25 ms | 1.29 ms | +3.6% | | | idle RSS p50 | 18.8 MB | 18.8 MB | 19.0 MB (+0.2) | +1.2% (p95 +3.9%) | | | RSS after first page p50 | 21.3 MB | 21.3 MB | 22.0 MB | +3.4% | every rep pr higher | | trash restore p95 | 106.4 ms | 106.3 ms | 111.6 ms | +5.0% | spread 7%, borderline | Pattern: microsecond-scale point lookups and the small per-note DB writes, about +2 us on ~46 us (~30 us/lookup is what infraternal measured for SQLite 3.54.1 itself) plus ~0.3-0.7 MB RSS. Not tuned around; I did not find an isolated cause because rusqlite 0.41 and SQLite 3.53.2 -> 3.54.1 change together (a split build would need another perf turn; say if you want it). ### Where pr is better or flat All user-visible paths are flat within the control: notes open p50 0.72 ms, home page p50 -6%, tag page p50 -6.8%, search query p50 1.45 -> 1.44 ms (p95 -4.9%), search first_query, sync round trip p50 0.212 -> 0.207 ms, save ack, typing, storm edits/s, durable write p50 24.9 -> 24.7 ms, cold index p50 1.465 -> 1.422 s, cold first page p50 319 -> 293 ms, cold startup p50 19.9 -> 17.6 ms, healthz startup, storage-bench save p50/p99, mail thread_open and storm. Rest of the table is in the collapsed block. <details><summary>All 169 metrics (median of 5 reps)</summary> reps=5 | metric | dev | dev-ctl | pr | ctl vs dev % | pr vs dev % | dev spread % | verdict | |---|---|---|---|---|---|---|---| | mail.mail.account_page@1accounts@100000.p50 | 0.05679 | 0.05816 | 0.05883 | +2.4 | +3.6 | 3.7 | REGRESS? | | mail.mail.account_page@1accounts@100000.p95 | 0.07195 | 0.07123 | 0.07423 | -1.0 | +3.2 | 8.8 | | | mail.mail.build_s@100000.p50 | 3.521 | 3.527 | 3.568 | +0.2 | +1.3 | 1.1 | | | mail.mail.build_s@100000.p95 | 3.521 | 3.527 | 3.568 | +0.2 | +1.3 | 1.1 | | | mail.mail.fake_headers_per_s.p50 | 2.105e+04 | 2.09e+04 | 2.105e+04 | -0.7 | -0.0 | 0.5 | | | mail.mail.fake_headers_per_s.p95 | 2.105e+04 | 2.09e+04 | 2.105e+04 | -0.7 | -0.0 | 0.5 | | | mail.mail.list_first@100000.p50 | 0.04045 | 0.04111 | 0.04164 | +1.6 | +2.9 | 3.0 | | | mail.mail.list_first@100000.p95 | 0.04899 | 0.05245 | 0.05035 | +7.1 | +2.8 | 8.4 | | | mail.mail.list_page@100000.p50 | 0.04613 | 0.04706 | 0.04821 | +2.0 | +4.5 | 3.3 | REGRESS? | | mail.mail.list_page@100000.p95 | 0.05731 | 0.05912 | 0.05823 | +3.2 | +1.6 | 7.6 | | | mail.mail.thread_open@100000.p50 | 0.04673 | 0.04774 | 0.04788 | +2.2 | +2.5 | 3.1 | | | mail.mail.thread_open@100000.p95 | 0.06332 | 0.06615 | 0.06392 | +4.5 | +0.9 | 3.8 | | | mail.mail.thread_storm_merges@newest@1000000.p50 | 6.155e+05 | 6.155e+05 | 6.155e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@newest@1000000.p95 | 6.155e+05 | 6.155e+05 | 6.155e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@newest@250000.p50 | 1.535e+05 | 1.535e+05 | 1.535e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@newest@250000.p95 | 1.535e+05 | 1.535e+05 | 1.535e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@newest@500000.p50 | 3.074e+05 | 3.074e+05 | 3.074e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@newest@500000.p95 | 3.074e+05 | 3.074e+05 | 3.074e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@oldest@1000000.p50 | 6.155e+05 | 6.155e+05 | 6.155e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@oldest@1000000.p95 | 6.155e+05 | 6.155e+05 | 6.155e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@oldest@250000.p50 | 1.535e+05 | 1.535e+05 | 1.535e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@oldest@250000.p95 | 1.535e+05 | 1.535e+05 | 1.535e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@oldest@500000.p50 | 3.074e+05 | 3.074e+05 | 3.074e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@oldest@500000.p95 | 3.074e+05 | 3.074e+05 | 3.074e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@shuffled@1000000.p50 | 6.155e+05 | 6.155e+05 | 6.155e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@shuffled@1000000.p95 | 6.155e+05 | 6.155e+05 | 6.155e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@shuffled@250000.p50 | 1.535e+05 | 1.535e+05 | 1.535e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@shuffled@250000.p95 | 1.535e+05 | 1.535e+05 | 1.535e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@shuffled@500000.p50 | 3.074e+05 | 3.074e+05 | 3.074e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_merges@shuffled@500000.p95 | 3.074e+05 | 3.074e+05 | 3.074e+05 | +0.0 | +0.0 | 0.0 | | | mail.mail.thread_storm_s@newest@1000000.p50 | 2.585 | 2.612 | 2.564 | +1.0 | -0.8 | 1.8 | | | mail.mail.thread_storm_s@newest@1000000.p95 | 2.585 | 2.612 | 2.564 | +1.0 | -0.8 | 1.8 | | | mail.mail.thread_storm_s@newest@250000.p50 | 0.6123 | 0.6168 | 0.5998 | +0.7 | -2.0 | 4.2 | | | mail.mail.thread_storm_s@newest@250000.p95 | 0.6123 | 0.6168 | 0.5998 | +0.7 | -2.0 | 4.2 | | | mail.mail.thread_storm_s@newest@500000.p50 | 1.254 | 1.264 | 1.243 | +0.8 | -0.9 | 2.8 | | | mail.mail.thread_storm_s@newest@500000.p95 | 1.254 | 1.264 | 1.243 | +0.8 | -0.9 | 2.8 | | | mail.mail.thread_storm_s@oldest@1000000.p50 | 2.608 | 2.649 | 2.582 | +1.6 | -1.0 | 3.5 | | | mail.mail.thread_storm_s@oldest@1000000.p95 | 2.608 | 2.649 | 2.582 | +1.6 | -1.0 | 3.5 | | | mail.mail.thread_storm_s@oldest@250000.p50 | 0.6079 | 0.6209 | 0.6042 | +2.1 | -0.6 | 3.8 | | | mail.mail.thread_storm_s@oldest@250000.p95 | 0.6079 | 0.6209 | 0.6042 | +2.1 | -0.6 | 3.8 | | | mail.mail.thread_storm_s@oldest@500000.p50 | 1.259 | 1.267 | 1.251 | +0.7 | -0.6 | 2.4 | | | mail.mail.thread_storm_s@oldest@500000.p95 | 1.259 | 1.267 | 1.251 | +0.7 | -0.6 | 2.4 | | | mail.mail.thread_storm_s@shuffled@1000000.p50 | 2.887 | 2.887 | 2.855 | +0.0 | -1.1 | 2.7 | | | mail.mail.thread_storm_s@shuffled@1000000.p95 | 2.887 | 2.887 | 2.855 | +0.0 | -1.1 | 2.7 | | | mail.mail.thread_storm_s@shuffled@250000.p50 | 0.6493 | 0.6647 | 0.6414 | +2.4 | -1.2 | 5.7 | | | mail.mail.thread_storm_s@shuffled@250000.p95 | 0.6493 | 0.6647 | 0.6414 | +2.4 | -1.2 | 5.7 | | | mail.mail.thread_storm_s@shuffled@500000.p50 | 1.361 | 1.376 | 1.356 | +1.1 | -0.4 | 3.6 | | | mail.mail.thread_storm_s@shuffled@500000.p95 | 1.361 | 1.376 | 1.356 | +1.1 | -0.4 | 3.6 | | | mail.mail.unified_page@1accounts@100000.p50 | 0.04587 | 0.04661 | 0.04815 | +1.6 | +5.0 | 3.8 | REGRESS? | | mail.mail.unified_page@1accounts@100000.p95 | 0.05913 | 0.05784 | 0.05825 | -2.2 | -1.5 | 28.2 | | | perf.cold.first_page.p50 | 319.3 | 312.5 | 305.3 | -2.1 | -4.4 | 15.3 | | | perf.cold.first_page.p95 | 335.6 | 332.6 | 324.9 | -0.9 | -3.2 | 25.8 | | | perf.cold.index.p50 | 1.465 | 1.449 | 1.444 | -1.1 | -1.4 | 24.6 | | | perf.cold.index.p95 | 1.552 | 1.464 | 1.468 | -5.6 | -5.4 | 38.5 | | | perf.cold.mem_peak.p50 | 142.9 | 142.9 | 143.9 | +0.0 | +0.7 | 1.1 | | | perf.cold.mem_peak.p95 | 143.6 | 143.8 | 144.4 | +0.2 | +0.5 | 1.5 | | | perf.cold.notes_per_s.p50 | 6445 | 6831 | 6810 | +6.0 | +5.7 | 31.1 | | | perf.cold.notes_per_s.p95 | 6826 | 6899 | 6924 | +1.1 | +1.4 | 20.4 | | | perf.cold.startup.p50 | 19.84 | 20.1 | 17.69 | +1.3 | -10.8 | 23.0 | | | perf.cold.startup.p95 | 20.16 | 21.91 | 17.9 | +8.7 | -11.2 | 27.6 | | | perf.home.first_page.p50 | 2.619 | 2.381 | 2.259 | -9.1 | -13.7 | 151.7 | | | perf.home.first_page.p95 | 4.085 | 2.617 | 4.092 | -35.9 | +0.2 | 103.8 | | | perf.home.page.p50 | 0.758 | 0.7539 | 0.7125 | -0.5 | -6.0 | 8.6 | better | | perf.home.page.p95 | 2.876 | 2.896 | 2.993 | +0.7 | +4.1 | 111.3 | | | perf.home.page_big.p50 | 16.57 | 16.27 | 16.6 | -1.8 | +0.2 | 3.9 | | | perf.home.page_big.p95 | 18.14 | 18.57 | 19.23 | +2.4 | +6.0 | 12.9 | | | perf.home.page_next.p50 | 0.7625 | 0.7989 | 0.8043 | +4.8 | +5.5 | 14.7 | | | perf.home.page_next.p95 | 1.121 | 1.195 | 1.168 | +6.6 | +4.1 | 24.2 | | | perf.home.page_tag.p50 | 5.321 | 5.14 | 4.958 | -3.4 | -6.8 | 5.7 | better | | perf.home.page_tag.p95 | 7.249 | 6.469 | 6.862 | -10.8 | -5.3 | 11.3 | | | perf.home.page_title.p50 | 0.6474 | 0.6316 | 0.6388 | -2.4 | -1.3 | 5.9 | | | perf.home.page_title.p95 | 0.7765 | 0.7368 | 0.7403 | -5.1 | -4.7 | 13.2 | | | perf.journal.catch_up.p50 | 28.15 | 28.18 | 28.61 | +0.1 | +1.6 | 7.1 | | | perf.journal.catch_up.p95 | 29.01 | 29.15 | 29.73 | +0.5 | +2.5 | 7.6 | | | perf.journal.entries.p50 | 1.018e+04 | 1.018e+04 | 1.018e+04 | +0.0 | +0.0 | 0.0 | | | perf.journal.entries.p95 | 1.036e+04 | 1.036e+04 | 1.036e+04 | +0.0 | +0.0 | 0.0 | | | perf.journal.page.p50 | 2.4 | 2.382 | 2.398 | -0.8 | -0.1 | 4.7 | | | perf.journal.page.p95 | 2.685 | 2.769 | 2.718 | +3.1 | +1.2 | 15.9 | | | perf.mem.after_first_page.p50 | 21.33 | 21.3 | 22.05 | -0.1 | +3.4 | 3.0 | REGRESS? | | perf.mem.after_first_page.p95 | 22 | 21.73 | 22.44 | -1.3 | +2.0 | 3.4 | | | perf.mem.after_opens.p50 | 41.73 | 41.14 | 41.74 | -1.4 | +0.0 | 6.1 | | | perf.mem.after_opens.p95 | 41.84 | 42.12 | 42.19 | +0.7 | +0.8 | 6.2 | | | perf.mem.after_search.p50 | 47.79 | 47.85 | 47.25 | +0.1 | -1.1 | 8.3 | | | perf.mem.after_search.p95 | 75.29 | 75.34 | 75.65 | +0.1 | +0.5 | 0.9 | | | perf.mem.idle.p50 | 18.78 | 18.77 | 19 | -0.0 | +1.2 | 1.2 | | | perf.mem.idle.p95 | 19.05 | 19.32 | 19.79 | +1.4 | +3.9 | 4.7 | REGRESS? | | perf.mem.peak.p50 | 47.32 | 47.59 | 46.99 | +0.6 | -0.7 | 8.6 | | | perf.mem.peak.p95 | 74.89 | 74.91 | 75.32 | +0.0 | +0.6 | 0.8 | | | perf.open.api_get.p50 | 3.924 | 3.976 | 3.877 | +1.3 | -1.2 | 10.9 | | | perf.open.api_get.p95 | 9.743 | 8.961 | 9.396 | -8.0 | -3.6 | 56.2 | | | perf.open.long_note.p50 | 15.84 | 15.82 | 16.05 | -0.1 | +1.3 | 211.7 | | | perf.open.long_note.p95 | 53.39 | 52.11 | 53.11 | -2.4 | -0.5 | 13.9 | | | perf.open.note.p50 | 0.7156 | 0.7009 | 0.7073 | -2.1 | -1.2 | 8.7 | | | perf.open.note.p95 | 1.316 | 1.299 | 1.377 | -1.3 | +4.6 | 22.5 | | | perf.search.catch_up.p50 | 2.005 | 2.005 | 2.005 | -0.0 | -0.0 | 0.0 | | | perf.search.catch_up.p95 | 4.013 | 4.009 | 4.006 | -0.1 | -0.2 | 0.2 | | | perf.search.first_query.p50 | 4.942 | 5.437 | 4.577 | +10.0 | -7.4 | 56.3 | | | perf.search.first_query.p95 | 26.52 | 29.22 | 26.28 | +10.2 | -0.9 | 5.5 | | | perf.search.query.p50 | 1.446 | 1.446 | 1.435 | -0.0 | -0.8 | 6.2 | | | perf.search.query.p95 | 3.974 | 4.011 | 3.942 | +0.9 | -0.8 | 15.0 | | | perf.shutdown.p50 | 64.02 | 64.08 | 64.01 | +0.1 | -0.0 | 33.1 | | | perf.shutdown.p95 | 64.24 | 85.28 | 84.18 | +32.8 | +31.0 | 36.1 | | | perf.startup.healthz.p50 | 4.997 | 4.964 | 5.066 | -0.6 | +1.4 | 44.7 | | | perf.startup.healthz.p95 | 5.415 | 6.009 | 5.693 | +11.0 | +5.1 | 55.8 | | | perf.storm.converge.p50 | 159.5 | 159.7 | 159.2 | +0.2 | -0.2 | 6.6 | | | perf.storm.converge.p95 | 176.5 | 175.7 | 174.9 | -0.4 | -0.9 | 10.3 | | | perf.storm.edits_per_s.p50 | 3845 | 3854 | 3903 | +0.2 | +1.5 | 4.5 | | | perf.storm.edits_per_s.p95 | 3948 | 3976 | 3973 | +0.7 | +0.6 | 5.1 | | | perf.storm.outside_edits.p50 | 2 | 2 | 2 | +0.0 | +0.0 | 0.0 | | | perf.storm.outside_edits.p95 | 2 | 2 | 2 | +0.0 | +0.0 | 0.0 | | | perf.storm.save_ack.p50 | 7.296 | 7.247 | 7.289 | -0.7 | -0.1 | 2.0 | | | perf.storm.save_ack.p95 | 11.86 | 11.34 | 11.62 | -4.4 | -2.0 | 12.0 | | | perf.storm.typing.p50 | 0.7802 | 0.7784 | 0.7686 | -0.2 | -1.5 | 4.6 | | | perf.storm.typing.p95 | 0.786 | 0.8078 | 0.7855 | +2.8 | -0.1 | 6.1 | | | perf.sync.durable_write.p50 | 24.88 | 24.7 | 24.67 | -0.7 | -0.8 | 1.2 | | | perf.sync.durable_write.p95 | 37.04 | 32.69 | 33.32 | -11.8 | -10.0 | 16.9 | | | perf.sync.round_trip.p50 | 0.2138 | 0.2133 | 0.2074 | -0.3 | -3.0 | 8.8 | | | perf.sync.round_trip.p95 | 0.3494 | 0.3382 | 0.3474 | -3.2 | -0.6 | 12.3 | | | perf.sync.save_ack.p50 | 3.815 | 3.723 | 3.691 | -2.4 | -3.3 | 6.8 | | | perf.sync.save_ack.p95 | 7.401 | 5.537 | 5.517 | -25.2 | -25.5 | 38.9 | better | | perf.trash.delete.p50 | 64.08 | 72.78 | 67.79 | +13.6 | +5.8 | 24.9 | | | perf.trash.delete.p95 | 98.77 | 96.5 | 98.94 | -2.3 | +0.2 | 5.8 | | | perf.trash.forget.p50 | 11.49 | 11.64 | 11.54 | +1.3 | +0.4 | 3.1 | | | perf.trash.forget.p95 | 15.45 | 15.21 | 15.91 | -1.6 | +3.0 | 44.0 | | | perf.trash.restore.p50 | 63.41 | 63.53 | 64.52 | +0.2 | +1.7 | 6.9 | | | perf.trash.restore.p95 | 106.4 | 106.3 | 111.6 | -0.0 | +5.0 | 7.0 | REGRESS? | | perf.uploads.response.p50 | 142.8 | 142 | 142.9 | -0.6 | +0.0 | 4.4 | | | perf.uploads.response.p95 | 165 | 163.8 | 163.9 | -0.7 | -0.6 | 12.8 | | | perf.uploads.round_trip.p50 | 0.5014 | 0.5016 | 0.4939 | +0.0 | -1.5 | 7.0 | | | perf.uploads.round_trip.p95 | 0.864 | 0.7296 | 0.7589 | -15.6 | -12.2 | 18.0 | | | perf.uploads.save_ack.p50 | 4.23 | 4.196 | 4.139 | -0.8 | -2.2 | 2.9 | | | perf.uploads.save_ack.p95 | 8.41 | 8.5 | 8.537 | +1.1 | +1.5 | 17.2 | | | sb-save.max | 60.56 | 63.67 | 59.41 | +5.1 | -1.9 | 42.4 | | | sb-save.p50 | 20.19 | 20.6 | 20.73 | +2.0 | +2.7 | 3.5 | | | sb-save.p90 | 27.73 | 28.04 | 28.66 | +1.1 | +3.4 | 8.1 | | | sb-save.p99 | 37.46 | 36.66 | 37.77 | -2.1 | +0.8 | 10.4 | | | tasks.cold_plain_s | 1.796 | 1.777 | 1.8 | -1.0 | +0.2 | 3.6 | | | tasks.cold_tasks_s | 2.14 | 2.104 | 2.148 | -1.7 | +0.4 | 3.2 | | | tasks/cold_index_ms/growth_pct | 19.27 | 18.09 | 20.37 | -6.1 | +5.7 | 13.1 | | | tasks/cold_index_ms/plain | 1796 | 1777 | 1800 | -1.0 | +0.2 | 3.6 | | | tasks/cold_index_ms/tasks | 2140 | 2104 | 2148 | -1.7 | +0.4 | 3.2 | | | tasks/describe_ms/pass.p50 | 0.0131 | 0.01281 | 0.01376 | -2.1 | +5.0 | 6.0 | REGRESS? | | tasks/describe_ms/pass.p95 | 0.03557 | 0.0351 | 0.03607 | -1.3 | +1.4 | 8.6 | | | tasks/describe_ms/pass_and_rows.p50 | 0.03899 | 0.03799 | 0.03994 | -2.6 | +2.4 | 4.2 | | | tasks/describe_ms/pass_and_rows.p95 | 0.07055 | 0.07091 | 0.09723 | +0.5 | +37.8 | 9.8 | REGRESS? | | tasks/list_ms/anytime.p50 | 1.824 | 1.819 | 1.843 | -0.3 | +1.0 | 9.6 | | | tasks/list_ms/anytime.p95 | 2.171 | 2.032 | 2.064 | -6.4 | -4.9 | 24.8 | | | tasks/list_ms/anytime_by_child_tag.p50 | 2.232 | 2.316 | 2.288 | +3.7 | +2.5 | 1.7 | | | tasks/list_ms/anytime_by_child_tag.p95 | 2.61 | 2.769 | 2.756 | +6.1 | +5.6 | 7.4 | | | tasks/list_ms/counts.p50 | 0.2496 | 0.2605 | 0.2544 | +4.4 | +1.9 | 17.5 | | | tasks/list_ms/counts.p95 | 0.2741 | 0.2942 | 0.2923 | +7.3 | +6.6 | 30.7 | | | tasks/list_ms/someday.p50 | 1.795 | 1.82 | 1.833 | +1.4 | +2.1 | 5.9 | | | tasks/list_ms/someday.p95 | 2.089 | 2.13 | 2.159 | +1.9 | +3.3 | 11.0 | | | tasks/list_ms/tag.p50 | 2.19 | 2.221 | 2.246 | +1.4 | +2.6 | 8.8 | | | tasks/list_ms/tag.p95 | 2.528 | 2.549 | 2.654 | +0.8 | +5.0 | 20.2 | | | tasks/list_ms/today.p50 | 2.317 | 2.28 | 2.301 | -1.6 | -0.7 | 6.8 | | | tasks/list_ms/today.p95 | 2.664 | 2.539 | 2.685 | -4.7 | +0.8 | 22.6 | | | tasks/list_ms/today_by_priority.p50 | 2.152 | 2.207 | 2.157 | +2.6 | +0.2 | 5.4 | | | tasks/list_ms/today_by_priority.p95 | 2.461 | 2.605 | 2.456 | +5.8 | -0.2 | 13.6 | | | tasks/list_ms/untracked.p50 | 1.242 | 1.248 | 1.286 | +0.5 | +3.6 | 6.4 | REGRESS? | | tasks/list_ms/untracked.p95 | 1.359 | 1.457 | 1.428 | +7.2 | +5.0 | 9.6 | | | tasks/list_ms/upcoming.p50 | 2.29 | 2.262 | 2.333 | -1.2 | +1.9 | 4.3 | | | tasks/list_ms/upcoming.p95 | 2.614 | 2.582 | 2.639 | -1.2 | +1.0 | 10.8 | | | tasks/notes | 1e+04 | 1e+04 | 1e+04 | +0.0 | +0.0 | 0.0 | | | tasks/tasks | 1e+04 | 1e+04 | 1e+04 | +0.0 | +0.0 | 0.0 | | | tasks/write_ms/first_touch.p50 | 15.53 | 15.53 | 15.81 | -0.0 | +1.8 | 4.1 | | | tasks/write_ms/first_touch.p95 | 64.7 | 66.9 | 68.01 | +3.4 | +5.1 | 16.9 | | | tasks/write_ms/warm.p50 | 4.858 | 4.889 | 4.97 | +0.6 | +2.3 | 1.1 | | | tasks/write_ms/warm.p95 | 14.3 | 15.16 | 14.5 | +6.0 | +1.4 | 8.0 | | 9 flagged of 169 </details> Run dir on perf-01: `~/perf/out-rusq-1` (reps 1-3 ran before an API stop; rep 4 was rerun from scratch). Same control gap shows run-to-run noise of ~1-3% on the small metrics.
kayg commented 2026-11-10 09:41:28 +00:00
Author
Owner
Copy link

Rebased onto dev fe2efbf11 (picks up #1039, #1048, #1050); force-pushed, head 8cea6f026. No conflicts, no new rusqlite manifest entries needed. Gates on dev-01 at that head: cargo fmt --check OK, cargo clippy -j4 --workspace --all-targets --locked -- -D warnings exit 0, cargo test -j4 --workspace --locked exit 0 (127 test binaries ok). cargo audit: RUSTSEC-2026-0459 still gone; the 4 remaining vulnerabilities (webpki-roots, rustls-webpki, pulldown-cmark x2) and the 3 warnings are unchanged from dev. Perf not redone (code delta is only the rebase; kayg accepted the earlier numbers).

Rebased onto dev fe2efbf11 (picks up #1039, #1048, #1050); force-pushed, head 8cea6f026. No conflicts, no new `rusqlite` manifest entries needed. Gates on dev-01 at that head: `cargo fmt --check` OK, `cargo clippy -j4 --workspace --all-targets --locked -- -D warnings` exit 0, `cargo test -j4 --workspace --locked` exit 0 (127 test binaries ok). `cargo audit`: RUSTSEC-2026-0459 still gone; the 4 remaining vulnerabilities (webpki-roots, rustls-webpki, pulldown-cmark x2) and the 3 warnings are unchanged from dev. Perf not redone (code delta is only the rebase; kayg accepted the earlier numbers).
kayg commented 2026-11-10 10:09:22 +00:00
Author
Owner
Copy link

Merged into dev (4274dafb4, integration merge with #1043/#928/#917/#934; clippy -D warnings + workspace tests green on dev-01). Closing.

Merged into dev (4274dafb4, integration merge with #1043/#928/#917/#934; clippy -D warnings + workspace tests green on dev-01). Closing.
kayg closed this pull request 2026-11-10 10:09:22 +00:00
Some checks are pending
CI / server (pull_request) Waiting to run
Details
CI / rust (pull_request) Waiting to run
Details
CI / web (pull_request) Waiting to run
Details
Markdown extraction contract / markdown-native-contract (pull_request) Waiting to run
Details
Markdown extraction contract / markdown-editor-contract (pull_request) Waiting to run
Details

Pull request closed

Please reopen this pull request to perform a merge.
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels   
backend
Rust server

  
design
Mockups, UI and TASTE

  
docs
ADRs, CONTEXT, TASTE

  
idea
Parked idea, never worked until promoted

  
infra
Deploy, CI, VMs

  
perf
Performance and budgets

  
reliability
Data integrity, corruption traps

  
web
Web client

No labels
backend
design
docs
idea
infra
perf
reliability
web
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
kayg/calternal!1043
No description provided.