py-wcmatch: updated to 11.0.1
11.0.1
- **DOCS**: Provide information in documentation that discloses information about what should or should not be
expected from Wildcard Match in regards to security, and provide general information to allow users to make good
decisions.
- **FIX**: Fix an issue where in some cases dot handling was not correct within an extended glob pattern when
`EXTGLOB`/`EXTMATCH` was enabled.
py-zensical: updated to 0.0.55
0.0.5
This version fixes two false positives in the reference extractor and reverts
Zensical's bootstrapped zensical.toml to TOML 1.0, since most editor tooling
does not yet support TOML 1.1. Zensical understands both, TOML 1.0 and TOML
1.1, so this change does not affect the functionality of Zensical itself.
py-numpy: updated to 2.5.2
2.5.2
MAINT: Prepare 2.5.x for further development
TYP: Backport multiple static typing fixes 1.
TST: add tests for stable ABI numpy extensions
BUG: fix ``StringDType`` coerce flag in binary ufunc promotion...
BLD: fix meson deprecation warnings
TYP: Backport multiple typing fixes 2.
MAINT: Update x86-simd-sort subproject (5adb334 → fa944ef)
BUG: fix crash on 32 bit systems using abi3t
MNT: remove some obsolete string to bool workarounds
BUG: centralized helper for output coerce and na_object in stringdtype...
BUG: fix CPU feature env diagnostic buffer overruns
BUG: restore ndarray.conjugate() for legacy user-defined dtypes...
TYP: Avoid shadowed dtype annotations
MAINT: Update verdored-meson/meson to match main.
BUG: fix refcount leak on overlapping copyto with where=False
[13 lines not shown]
py-fastjsonschema: updated to 2.22.2
2.22.2
* Fixed wrong results of anyOf, oneOf, not, if/then/else, contains and propertyNames when `fast_fail` is off
* Fixed collecting errors from subschemas behind `$ref` when `fast_fail` is off
* Fixed escaping property name in boolean `dependencies` validation (prevents schema-controlled code injection)
py-project-api: updated to 1.11.0
1.11.0
Adds support for Python 3.15 and 3.15t, and fixes a race in
SubprocessCmdStatus.done: it keyed off returncode, which communicate() sets
before the thread stores the output, so a poller could reach out_err() while it
was still None (free-threaded builds hit this window). done now reports
finished only once the output is stored
py-prometheus_client: updated to 0.26.0
0.26.0
Expose measured duration on Timer context manager
Use specific exception for duplicate timeseries
Make clear() a no-op for label-less metrics
Use object instead of Any for labels() parameters
Add TLS version parameters for start_wsgi_server
Validate Enum arguments before registering the collector
Do not emit a leading zero in floatToGoString exponents >= 10
fix: make test suite pass on Windows and Python 3.12+
Parse the native histogram sum as a float
Remove nameless collectors from the registry on unregister
Fix operator precedence allowing exemplars on any metric type
sysutils/lla: update to 0.5.12
[0.5.12] - 2026-08-15
Added
Plugin Platform v2 with manifest-based packages, API-range negotiation, plugin-owned response buffers, batch decoration, typed machine-output fields, declared permissions, ordered user/system discovery, package diagnostics, and per-package checksums.
Plugin lifecycle destruction, panic isolation, bounded responses, functional contract probes, and CI/release conformance gates for every bundled plugin.
lla plugin doctor, lla plugin info <name>, and lla plugin permissions <name> for inspecting v2 installations.
security_audit plugin for unsafe permissions, SUID/SGID bits, suspicious symlinks, and exposed secret-like files.
media_inspector plugin for MIME detection, built-in image dimensions, optional EXIF, and ffprobe-backed audio/video metadata.
project_context plugin for Rust, Node, Python, and Go detection with lockfile, artifact, toolchain, and Git health fields.
trash plugin for cross-platform recoverable deletion, listing, conflict-safe restoration, and age-gated permanent emptying.
preview plugin with bat text/Markdown highlighting, chafa image rendering, safe archive listings, and built-in fallbacks.
Changed
All bundled plugins now ship with v2 manifests and are packaged as self-contained directories while legacy flat v1 libraries remain loadable during migration.
Invalid plugins are quarantined for recovery by lla clean instead of being permanently deleted.
file_remover remove now moves selected entries to recoverable trash; irreversible deletion is available only through the explicit purge action and confirmation.
[40 lines not shown]
editors/redox: update to 0.8.1
This is the last tiny set of fixes before a thorough editor rewrite, which will mostly be to clean up the scope of the project and the implementation details of some features.
Tweaked the yellow shades in the base colour scheme
Removed the perf corners widget because it's fundamentally not useful for assessing editor performance, which I somehow only just realized :)
devel/sem: update to 0.22.1
[0.22.1] - 2026-08-16
Added
sem find / sem callers / sem refs: new query verbs that answer directly from the mmap index.sem — entity definitions, direct callers (reverse edges), and direct refs (forward edges) — without touching cache.db.
sem grep <pattern>: trigram-accelerated text search over the mmap index. A required-trigram query against the index's TRIGRAM section narrows the candidate file set, each candidate is then verified with the real regex matcher against its current bytes (never stale stored content), and output is rg-compatible file:line:text. Beats rg 11-26x on giant repos (measured on the TypeScript monster and home-assistant-core corpora: 25-53ms vs. rg's 283-980ms for the same pattern). Falls back to a full scan for patterns no trigram query can be derived from.
sem review listen <diff-id-or-url> [--dry-run]: one-command agent attach to a hosted sem-cloud review. Resolves credentials, validates the diff exists, locates the review-listener plugin, and execs claude with the documented flags/env. --dry-run prints the assembled command with secrets masked, without launching or requiring claude to be installed.
Three new sem-cloud MCP tools (join_review, wait_for_branch, reply_to_branch) let an agent join a hosted code review as a live listener: long-poll for reviewer questions anchored to lines of a diff, investigate them in the repo, and stream answers back. Ships with a Claude Code plugin (integrations/claude-review-listener/) that wires the tools up and adds a read-only Stop-hook backstop for headless sessions.
sem diff's hosted upload no longer blocks on the local caller/callee relations pass. With cloud consent on, the diff snapshot uploads immediately with empty relations; the server queues enrichment and replies "enrichmentQueued" (or, against an older server, the CLI runs the existing local pass and PUTs the result afterward). SEM_RELATIONS_LOCAL=1 restores the old blocking single-upload behavior; the local relations pass's own budget is now adaptive to repo size.
Performance
Cold graph builds are 11-30% faster and peak RSS is down 17-40% on giant corpora, versus 0.21. Measured end-to-end on the shipped release binary: home-assistant-core 5.9s, TypeScript monster 11.2s, dotnet-runtime 46.7s, llvm-project 34.7s, linux 35.2s cold full-CLI (warm rebuilds: 0.2-1.3s). Every corpus improved on both the engine-only and full-CLI metrics; full methodology and per-corpus numbers are in RESOLUTION-PROFILE.md's FINALE section.
C#/C++ builds now skip re-parsing files whose facts are already known, closing the last gap in precomputed-facts reuse (JS/TS/Python/Go/Java/Rust already had it). A per-file gate proves a corpus-wide invariant — no entity's parent lives in a different file — before trusting precomputed facts wholesale, so this needed no facts-schema change. Measured on dotnet-runtime: reparse time drops from 10.6s to 65ms.
Parsed file facts now persist to disk as a content-addressed corpus, so a build that has seen a file's exact content before warm-starts it instead of re-parsing from scratch, even in a fresh process. Fixed a regression where checking a shared corpus against a large number of prior contributors got slower as the corpus grew (one repo's known-content rebuild was measured 332% slower against a 7.9GB shared corpus than a 556MB one); it now costs the same regardless of corpus size.
sem context regained a fast tier it had lost, by reading each entity's body from its own file at an indexed byte span instead of walking and hydrating the whole corpus. A prior cascade of cache removals deleted the old fast path along with a correctness bug it had, but left sem context always doing a full corpus load — measured on the TypeScript monster 1.11s down to 48ms, on a mid-size repo 0.16-0.28s down to 4.6ms. Verified byte-identical against the always-correct full-load path across both entity- and file-scoped lookups; still declines (never approximates) on a stale cache, an ambiguous name, or any entity missing a span.
Changed
[7 lines not shown]
devel/gitpane: update to 0.11.0
[0.11.0] - 2026-08-15
Added
gitpane now sleeps when nobody is looking at it. Under tmux, a lightweight probe (one tmux display-message every 3 seconds) tracks whether the pane is visible and whether the session has seen input recently, and drives three states: awake (visible and recently touched: full polling), doze (visible but input-idle for watch.doze_after_secs, default 120: periodic local polls and fetches stop, while filesystem-watcher refreshes still land and a once-per-second heartbeat renders their results, so the display keeps tracking real changes), and deep sleep (detached session, background window, or zoomed away: everything stops, and watcher events are dropped or deferred). Waking refreshes immediately: the local poll fires at once, a root-directory change that happened while hidden is replayed, and the frame repaints; the fetch timer instead resumes on its normal cadence so switching to a pane never triggers a burst fetch of every repo. With many long-lived gitpane instances only the one on screen does any work, which cuts idle CPU (and battery) from a constant 15-20% per hidden instance to zero. Outside tmux behavior is unchanged; opt out with watch.sleep_when_hidden = false.
Changed
The tick, local-poll, and fetch timers now skip missed ticks instead of replaying them in a burst after the event loop stalls.
makedepend: update to 1.0.10.
Alan Coopersmith (5):
Improve man page formatting
gitlab CI: drop the ci-fairy check-mr job
meson: Add option to build with meson
meson: include headers when checking for functions
makedepend 1.0.10
