[neilbags]

Warning

Vibecoded slop. Unreviewed.

Target: Pred8or/podkit:pr/8-tags-adapters, the top of the stack in jvgomg#76–jvgomg#83. The change touches tools/prebuild/libgpod-nano.patch, which doesn't exist on jvgomg/podkit:main yet. It logically belongs with jvgomg#78 (command set selection) and jvgomg#79 (keep_device_choices), so cherry-pick or squash it wherever it fits best.

Problem

On an iPod nano 7G running firmware 1.0.1 (build 37A10002), every database write fails:

[run_post_process_commands]: Error fetching user version command set
Failed to recreate iPod database: ... SQLite post-process commands: no command set 25

The SysInfoExtended read live over USB lists only UserVersionCommandSets 9. The patch accepts only 25, which firmware 1.0.4 provides.

podkit device reset deletes the audio files before it writes, so this failure leaves the device with no iTunesDB.

Set 9 compared with set 25

	set 9 (fw 1.0.1)	set 25 (fw 1.0.4 capture)
commands	117	127
shared commands	116, same order, same SQL
only in this set	UpdateDeviceUpdateLevelTo1107	CreateIsHomeVideo, FakePragmaITunes26, 8× ItemStats{Audio,Subtitle}{Language,TrackID}, UpdateDeviceUpdateLevelTo1109

Change

• run_post_process_commands: on nano 6G/7G, use set 9 when the device lists no set 25. Set 25 is still preferred, and every other set (for example only 26) is still rejected.
• keep_device_choices: copy the track choice columns only when the new Dynamic.itdb has them (new helper, has_track_choices). Set 9 doesn't create them, so the UPDATE failed whenever the device already had a Dynamic.itdb.
• New test test_nano_firmware_1_0_1_command_set_9. It builds set 9 from the set 25 capture, then checks create and resave, the 1107 update level, and that the track choice columns are absent.
• Updated the tools/nano7/README.md paragraph about command sets, and added a changeset.

Testing

• tools/nano7/test-libgpod.sh, run in an Alpine 3.21 builder against the statically built patched libgpod: 19 tests pass, 1 skipped (no time zone data in the builder). The new test fails against the unfixed patch.
• Real device, a nano 7G 16GB on fw 1.0.1, using the musl binaries in the upstream packages/podkit-docker/Dockerfile:
  • device init worked, then sync of a test MP3 and a 14-track MP3 album with artwork.
  • Each write printed 117 out of 117 post process commands successfully executed.
  • podkit device music lists all 14 tracks with the right metadata.
  • Not verified: whether the firmware shows and plays the library.
• Known difference from fw 1.0.4: Library.itdb's user_version stays 0, because set 9 has no FakePragmaITunes26. I don't know whether firmware 1.0.1 checks it.