JuliaGNSS/Tracking.jl: v5.0.0

Zenodo (CERN European Organization for Nuclear Research)

Abstract

5.0.0 (2026-08-03) feat(bit_buffer)!: drop get_bits, soft bits are the only bit store (4f0c508) feat(bit_buffer)!: store decoded bits as soft bits only, unbounded (a86a561), closes GNSSReceiver.jl#107 fix(cn0)!: divide by the record's real integration time, not one code period (dc81782), closes JuliaGNSS/GNSSReceiver.jl#107 BREAKING CHANGES the exported get_bits accessor is removed — on BitBuffer, TrackedSignal, TrackedSat and TrackState. Read get_soft_bits and take the sign instead: soft_bit > 0 is the hard bit, and unlike get_bits it is not capped at 128 bits. get_num_bits and get_soft_bits are unchanged. The internal convenience constructor drops its two hard-bit arguments, becoming BitBuffer(code_block_buffer, code_block_buffer_length, found, prompt_accumulator, prompt_accumulator_integrated_code_blocks[, soft_bits]). Co-Authored-By: Claude Opus 5 (1M context) noreply@anthropic.com BitBuffer loses its buffer::UInt128 and length::Int fields, so its internal positional constructor goes from eleven arguments to nine. The type is not exported, but code that constructs it positionally or reads those two fields must be updated — this repo's own benchmark/benchmarks.jl did, and it carries per-layout shims precisely because it is run against every revision. The exported surface is unaffected: get_bits is bit-identical to the old packed buffer for every history that previously did not throw (and now returns a value where it used to throw past 128 bits), get_num_bits / length can no longer disagree with soft_bits, get_soft_bits is unchanged, and the legacy 7-argument BitBuffer constructor still works by synthesizing ±1 soft bits. Co-Authored-By: Claude Fable 5 noreply@anthropic.com TrackedSignal, which is exported, gains a last_fully_integrated_num_code_blocks::Int field, so constructing one positionally takes one more argument. Code using the keyword constructor is unaffected; code building TrackedSignals positionally must be updated, as this repo's own test/sat_state.jl was. estimate_cn0 also returns different numbers: any signal whose records span N primary-code blocks now reads 10*log10(N) dB lower, which is the correction, but downstream thresholds calibrated against the inflated figure need retuning. Records of one code block — everything before bit sync, and the default configuration — are unchanged.

Authors 10

1. Soeren Schoenbrod Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

2. Semantic Release Bot
   Affiliation as printed

   @semantic-release

3. siebc
4. giove-a
5. dominformant
6. erikdeinzer
7. Can Özmaden Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

8. Julia TagBot
9. Kristoffer Carlsson
   Affiliation as printed

   @JuliaComputing

10. Tim Besard
    Affiliation as printed

    @JuliaComputing