Changes in version 0.4.0-0.0.1.9000 - Add a Win32 file-mapping implementation for vendored minibwa so Windows and Rtools builds no longer require the unavailable POSIX sys/mman.h. - Add a versioned, lossless native FASTQ query-group stream for downstream BAM/CRAM producers. The installed C API exposes bounded one-template views, typed CIGAR/tag data, header facts, cooperative cancellation, explicit ownership/lifetime rules, exact read-group preservation, and reason-coded paired FASTQ validation without a SAM-text or per-record-R-callback path. - Update vendored minibwa to exact upstream commit 84019dc454013134570479c50355741ed8318a8e (MB_VERSION 0.4-r411-dirty; archive SHA256 fe7e5a14bf26882b3f77804bec210c3b0fa3e073555664da1ce6f53aef80565c). Changes in version 0.3.0-0.0.1.9000 - Add a MinGW/Rtools-only kthread.c dead-return guard after pthread_exit() to avoid -Wreturn-type install warnings on Windows. - Make kseq.h's kroundup64() width-safe on 32-bit targets such as wasm32, avoiding Emscripten -Wshift-count-overflow diagnostics. - Add mb_map_batch() for native columnar mapping of many query sequences in one call, plus mb_map_pair_batch() for explicit two-batch paired-end input. - Add native FASTX iterators and mb_map_fastx_batch() so file-backed batches can be mapped without materializing sequence strings in R. - Add a read-level sidecar to alignment batches with read IDs, query lengths, fragment/mate fields, and per-read hit offsets/counts, exposed through ALTREP read columns and the C API. - Add a concise print method for native alignment batches. - Return simd_info() as a classed rminibwa_simd_info object with a concise print method for dispatch diagnostics. - Fold in upstream minibwa v0.3 / MB_VERSION 0.3-r391 at commit a679d17cb822c6cca67ab78c439f9c9826f1b07d. - Switch package versions to the - scheme; development builds now use 0.3.0-0.0.1.9000. - Add a MinGW/Rtools-only kalloc.c format-string guard for size_t diagnostics while leaving the non-MinGW upstream code path unchanged. Changes in version 0.0.0.9000 - Bootstrap Rminibwa as a GPL (>= 2) R package wrapping pinned upstream lh3/minibwa C sources. - Add native bindings for index build/load, raw-vector query mapping, native external-pointer alignment batches, ALTREP column views, and packed CIGAR accessors. - Install inst/include/Rminibwa.h and register C-callable accessors so downstream packages can consume alignment batches directly from C. - Add staged KSW SIMD dispatch following the RsimdDispatch pattern: portable scalar/SIMDe fallback, SSE4, and AVX2 backends are compiled as separate objects and selected at runtime without global ISA compiler flags. - Add AVX2 dual-gap extd2 support from the minibwa binding work and keep it as an optional runtime-dispatched backend. - Vendor minibwa at commit f0fd108456d459e5da3755f29c8700dc7f6fe0df with a patch queue and provenance metadata. - Add Rtinycc inline downstream-C examples in the README, vignette, and tests. - Build and install a package-provided minibwa executable from vendored sources; CLI wrappers now use it by default. - Add developer-only benchmarks comparing internal staged backends and optional external Python/Rust bindings from an explicit local checkout. - Add make asm / tools/check-assembly.R to audit generated SIMD instruction families in staged objects and shared libraries. - Add pkgdown configuration and GitHub Actions workflows for standard R CMD check, Fedora/R-devel CRAN-style check, and GitHub Pages deployment.