# Watermelon In Easter Hay: what actually ships Songsterr `s35881`. Built 2026-09-07 from the exhaustive handoff, which named a ship file and carried three internal contradictions about which one it was. ## The upload path HARD GATE, corrected by Brandon 2026-09-07. We never submit a revision to an original tab, our own damage included. 1. Open the original tab `s35881`. 2. Click "Create a copy to edit". 3. Rename the copy with **"Brandon edit"** on the end. 4. Import `RESTORED-s35881-TEXTSAFE-v2.gp` and submit the revision on that copy. ### Preflight through /songsterr-upload | run | reference | verdict | meaning | |---|---|---|---| | A | `--ai-export r8768414-LIVE-EXPORT.gp` | REFUSED, NAMEDROP | fires on r8768414's 7 names; the copy starts from r8908034, which holds zero | | B | `--live-tracks 9` | REFUSED, GRID | count-only mode cannot see inheritance; run A printed GRID-INHERITED | Neither fault is caused by the file. Master bars 105 vs 105 identical, tempo automations 1 vs 1 identical, MasterBars sha `457cf4db697a0341` on both sides. ### The NAMEDROP rule is too absolute | revision | source | names present | |---|---|---| | s35881 `r8768414` | 136 CDATA | 7 of 9 | | s35881 `r8908034` | ElementTree, 0 CDATA | 0 of 9 | | s5820647 `r8906157` | ElementTree, 0 CDATA | 0 of 5 | | s5820647 `r8932311` | CDATA-carrying | 5 of 5 | CDATA presence in the uploaded file is the variable. TEXTSAFE-v2 carries 137. ## Full transcription comparison Four witnesses exist on this Mac. Two are machine-readable, two are printed below the 15 px staff-space floor. | witness | resolution | note-level comparison | |---|---|---| | Songsterr `s35881 r7715683`, Ben Dibden1 | native GPIF, 104 bars | yes, the reference | | `Unknown Artist-206 Watermelon in Easter Hay-...-09-06-2026(1).gp` | native GPIF, 279 bars | lane density only | | Drumnet, BartoRomeo, 5 pages | 595x842, 5.0 px/staff space | no | | Kasper Sloots page | 785x1124, 10 px spacing | no | | lane | midi | tab | machine | reading | |---|---|---|---|---| | Ride (middle) | 51 | **1,724** | **596** | tab 2.9x denser | | Electric Snare | 40 | 193 | 0 | tab only | | Snare | 38 | 0 | 11 | machine only, different snare lane | | Kick | 36 | 107 | 475 | machine 4.4x denser | | Crash high | 49 | 53 | 212 | machine 4.0x denser | | Hi-Hat closed | 42 | 0 | 125 | machine only | | Pedal Hi-Hat | 44 | 6 | 0 | tab only | | TOTAL | | **2,091** | **1,428** | | The two disagree on the snare lane itself, so neither corroborates the other. The machine file runs 230 bars of 4/4 plus 48 of 2/4 against the tab's 53 of 4/4 and 52 of 5/4, so a bar-keyed diff is meaningless. It cannot support a claim that the sweep removed 1,128 ride notes. ## Status change, 2026-09-07 afternoon `r8908034` cleared moderation at 14:49 UTC and is the LIVE published revision. It emptied all nine track names, removing seven authored musician credits from the public tab: Frank Zappa, Warren Cuccurullo, Arthur Barrow, Ed Mann, Warren Cuccurullo, Wurlitzer (Peter Wolf), Vinnie Colaiuta. | revision | empty names | note | |---|---|---| | `r7715683` Ben Dibden1 | 2 of 9 | the author's, all seven credits present | | `r8768414` our paren sweep | 2 of 9 | our own .gp upload, 136 CDATA, names preserved | | `r8908034` our restore, LIVE | **9 of 9** | **every name empty** | The repair mechanism is proven here: a CDATA-carrying `.gp` upload of ours already kept the names on this exact song. TEXTSAFE-v2 carries 137 CDATA blocks. The same repair on `s5820647` went live as `r8932311` and restored its names. ## Wider sweep, 16 songs Only two songs carry live credit loss: **s35881** (7 credits, our `r8908034`) and **s68248** Carolina Hard-Core Ecstasy (3 credits, our `r8769058`, logged on 2026-09-06 and never repaired). Eight of our live revisions show no reduction. Five songs now have a moderator's revision live instead of ours: s35886, s620961, s35870, s35887, s749523. Nine revisions remain on moderation across seven songs. Table: `out/NAMEDROP-LIVE-SWEEP-2026-09-07.md` ## Ship target `RESTORED-s35881-TEXTSAFE-v2.gp`, 51,451 bytes, sha256 `87020f8c4243a7441db8187b4d7356b300e62b4d013787343de5ea4a0cfe27fd`, in `~/Projects/_outputs/songsterr-zappa-paren-fix/s35881-Watermelon/`. It supersedes `RESTORED-s35881-TEXTSAFE.gp` and every ElementTree build. ## The defect this pass found GPIF stores a chord in two places: an `` defining the diagram, and a `` reference on the beat that displays it. The paren sweep dropped both halves of E7Msus2 at bar 3. `build_restore_textsafe.py` restored the Item alone, so the definition sat orphaned and no bar printed the symbol. Three checks passed on the broken file: - `'E7Msus2' in text`, which an orphan satisfies - the printed list of chord `Items`, which never says whether a beat references one - the lane-aware set difference, which keys on `(bar, voice, position, midi)` and cannot see a symbol that carries no notes `Beat id 2` is shared by seven voices covering track 0 bars 3 to 9, so the repair clones it copy-on-write and repoints the bar 3 voice alone. Tool: `tools/restore_bar3_chord_ref.py`. ## Corrections to the handoff **NAMEDROP was wrong about deletion.** ElementTree strips the `` wrapper and leaves the text. `RESTORED-s35881-FINAL.gp` carries all 28 track names and "Frank Zappa" as ordinary character data. TEXTSAFE stays the right lane because a reserialized file rewrites 1,038,175 bytes and reformats whitespace throughout. **124 rest cells are not damage.** They differ from the pre-sweep source across six tracks, all with zero notes on both sides, and all in 5/4 bars where a Whole plus a Quarter fills the bar correctly. Note loss across all nine tracks is zero. **Sections 5 and 6 of the handoff are stale.** They name `RESTORED-s35881-FINAL.gp` and list six `.gp` files; disk holds eight. Section 3 is the controlling text. ## Verification against `PRESWEEP-r7715683.gp` | check | result | |---|---| | beat-level chord symbols | 4 of 4, identical locations, bar 3 alone | | CDATA blocks | 137, equal to source | | CDATA `` entries | 28 | | note instances, 9 tracks | 4,856, identical per track | | AntiAccent / Staccato | 1,317 / 0 of 2,091 drum events | | ghost missing / spurious | 0 / 0 | | positions added / removed | 0 / 0 | | lanes with a wrong ghost count | 0 | Transcript: `~/Projects/_outputs/zappa-watermelon-ride-analysis/out/VERIFY-TEXTSAFE-v2-2026-09-07.txt` ## Queue Nine closed with individual proof, eight blocked with a named unblocking condition, one upload decision queued as `q-2026-09-07-a74a94`. Record: `~/Projects/_outputs/zappa-watermelon-ride-analysis/out/QUEUE-CLOSURE-s35881-2026-09-07.md` ## Owed - **Brandon**: one upload decision. The path is the COPY lane, corrected 2026-09-07. We never submit a revision to an original tab, our own damage included. Open `s35881`, click "Create a copy to edit", rename the copy with "Brandon edit" on the end, import `RESTORED-s35881-TEXTSAFE-v2.gp`, and submit the revision on that copy. The original tab and `r8908034` stay untouched. - **A scan**: five Drumnet pages at print DPI unblock every chart-based item at once. - **Another session**: the onset-based DP alignment `q-2026-09-06-57b3e8` unblocks pedal hi-hat.