[aferaudo]

Summary

• Every L1-KPM indication now carries validSymbolMask (valid_symbol_mask in JSON). Bit s set means symbol s of the slot is uplink. DL and guard symbols of a mixed slot are 0, and FDD sets all bits. The gNB computes the mask; this PR carries it from gNB to codelet to controller to dApp, the same way OAI's L1-KPM SM does.
• gNB side: wineslab/ocudu-e3-private#4. The two shared headers are byte-identical to it, so merge both PRs together.

Type of change

• Bug fix
• New feature / enhancement
• New Service Model
• Refactor (no behavior change)
• Documentation (docs/service-models.md and the codelet README)
• Test / CI / packaging
• Other (explain):

Linked issue

None.

Mandatory test checklist

• Over the air. Setup: Foxconn RU, n78, 100 MHz, 4T4R, TDD 7 DL slots / 6 DL symbols / 2 UL slots / 4 UL symbols, a COTS phone, the adaptive_cpu dApp over JSON. Results:
  • every published slot (8, 9, 18, 19) had valid_symbol_mask = 0x3FFF;
  • iperf: 73 Mbit/s UL, 96 Mbit/s DL;
  • the dApp check passed.
• Round trip. A small test encodes masks 0x3C00, 0x3FFF and 0 with the APER and JSON encoders. All three decode back to the same value, and nRxAnt is still 4.
• Special slot over the air (0x3C00): not possible on our RU. ocudu-e3-private#4 explains why. The gNB's mask for slots 7 and 17 was checked offline instead.
• Same code as the over-the-air run. Rebased onto main after Publish UL slots through a native gNB helper instead of a jbpf copy #8 was merged. The code changes are byte-identical to the commit that ran over the air; only the docs moved, into docs/service-models.md. CI builds it as well.
• VERSION, libe3.pc, new dependencies: N/A, none changed.

CI checklist

• build and measure is green for the head commit (to check once CI has run).
• Commit policy, run locally with libe3's scripts/check_commit_trailers.py --base origin/main --head HEAD: it passes (Assisted-by trailer).

Twin-repo coordination

• This PR does not change the E3 wire protocol or public ABI, OR a paired PR exists in each affected twin repo (link below).

Two contracts change:

• gNB ↔ codelet: jbpf_ran_slot_ctx grows from 47 to 49 B, and the verifier checks 49. e3_slot_desc keeps its size: the mask uses 2 of its reserved bytes. Paired PR: wineslab/ocudu-e3-private#4.
• Controller ↔ dApp: APER already carried the validSymbolMask presence bit since the grammar update in Publish UL slots through a native gNB helper instead of a jbpf copy #8, so there is no new grammar. A dApp that ignores the field keeps working, and a missing mask should be read as 0x3FFF.

Workflow confirmation

• Linear, fast-forward-able descendant of main (fbdf325, after Publish UL slots through a native gNB helper instead of a jbpf copy #8), with no merge commits.
• A single commit with an imperative subject and a what/why body.
• I have read and followed CONTRIBUTING.md (libe3's, applied to this repository).

🤖 Generated with Claude Code