chore(release): 0.3.20

[randy3k]

Release 0.3.20

Highlights

• Run full-document, range, and multiple-range formatting in a lazy, cancellable background worker (Run explicit formatting in the background #764, Run formatting in background process #251)
• Stabilize diagnostics for Quarto and R Markdown documents, preserve nested .lintr settings, and clear diagnostics when disabled (Stabilize diagnostics #763, .lintr appears to be ignored for diagnostics when languageserver lints open documents #726)
• Enforce strict mutual exclusivity of result and error in JSON-RPC responses per JSON-RPC 2.0 (fix: ensure error and result are mutually exclusive in JSON-RPC responses (#766) #768, [BUG REPORT]: Cancellation responses include both result:null and error, and violate JSON-RPC 2.0 #766)
• Avoid passing null or zero-length buffers to qsort in reference resolution (fix: avoid passing null or zero-length buffers to qsort in reference resolution #767)
• Speed up test-coverage workflow with 4 CPUs, path filtering, and 4-way balanced test sharding (ci: speed up test-coverage workflow with 4 CPUs, sharding, and path filtering #769)

Changes

• Updated DESCRIPTION with Version: 0.3.20 and Date: 2026-09-19
• Updated NEWS.md with release notes for 0.3.20

[randy3k]

R-hub UBSAN Check Results (release/0.3.20)

• Commit: 98705f0 (chore(release): 0.3.20)
• Workflow Run: R-hub run 35455318597 (undestructive-lobster)

Platform	Environment	Duration	Check Status	UBSAN / Sanitizer Diagnostics
clang-ubsan	Ubuntu 22.04.5 LTS / R-devel (r90566) / Clang UBSAN	2m 25s	Status: OK	Clean (0 runtime errors, 0 warnings, 0 notes)
gcc-asan	Fedora Linux 42 / R-devel (r90185) / GCC ASAN+UBSAN	3m 39s	Status: OK	Clean (0 runtime errors, 0 warnings, 0 notes)

Both UBSAN test platforms passed with 0 errors, 0 warnings, and 0 notes. No undefined behavior diagnostics were detected.

[randy3k]

I think I will wait for a few more days to see if we will receive any bug reports from 0.3.19. Will probably make a release on next Tue/Wed.