# Numba Meeting: 2026-01-20

Attendees: Todd Anderson, Graham Markall, Stuart Archibald, Swapnil Patel, Siu Kwan Lam, Kaustubh Chaudhari,
FPOC (last week): Graham
FPOC (incoming):

NOTE: All communication is subject to the [Numba Code of Conduct](https://github.com/numba/numba-governance/blob/accepted/code-of-conduct.md).

Please refer to [this calendar](https://calendar.google.com/calendar/embed?src=5rqnddm4gjrdfjm31fsv182epk%40group.calendar.google.com) for the next meeting date.

## ~~Office hour signup~~

## Discussion

- (esc) - (https://github.com/numba/llvmlite/pull/1387#issuecomment-3723958449)
- Do we accept vcpkg based solution or wait for conda/miniconda to become available for Windows-on-arm?
- Accept and label as experimental?
- "Tier 1.5" - https://github.com/numba/numba/issues/10406 - could be accepted into this tier. (also could be considered "experimental")
- Security implications of vcpkg-based package management for wheel builds
- Conclusion (TBC): Will not accept as it requires changing build infrastructure, but value in having demonstrated that it works on WoA.
- Tier 1.5:
- GHA and conda support considerations:
- We might build variants of things we already support, but not change the build infrastructure to support it
- (esc) - https://github.com/numba/numba/pull/10384
- Do we list all Tier 2 platforms and attempt to accurately maintain this document?
- The Q regarding `osx-64` (Intel chips, not M* series) has come up at least twice via discourse since that platform was relegated from Tier 1.

For next time:

- (esc) https://github.com/numba/numba/pull/10352
- Are we OK to just remove headers?
- (esc) https://numba.discourse.group/t/type-annotations-in-llvmlite/3193
- Does anyone strongly object to attempting to add type annotations to llvmlite?
- (swap357) Should we reconsider 30 days to stale bot policy for issues ? Especially during winter-break. can we pause the bot then?
- (esc) https://github.com/numba/numba/pull/10386
- Do we need to implement the full NumPy select stuff?
- (esc) https://github.com/numba/llvmlite/issues/1389#issuecomment-3770977823
- Do we want to support `osx-64` for another 18 months?

## New "Ready for Review" PRs

## 1. New Issues
- [numba#10379](https://github.com/numba/numba/issues/10379) - Nested jitted function does not handle nonlocal variable correctly
- [numba#10380](https://github.com/numba/numba/issues/10380) - Support ._replace() on namedtuples
- [numba#10388](https://github.com/numba/numba/issues/10388) - LLVM backend code generation error on ARM64
- [numba#10394](https://github.com/numba/numba/issues/10394) - Numba tanh inconsistent with numpy tanh
- [numba#10399](https://github.com/numba/numba/issues/10399) - Execution of `np.linalg.slogdet` within an `@njit` decorated function hangs
- [numba#10400](https://github.com/numba/numba/issues/10400) - `ValueError` when using `parallel=True` with `np.empty` on structured array's `.dtype`
- [numba#10402](https://github.com/numba/numba/issues/10402) - Confusing error message: "cannot assign slice of shape (47,) from input of shape (48,)" when assigning (47,) array to (48,) slice
- [numba#10403](https://github.com/numba/numba/issues/10403) - The np.bitwise_count() function fails
- [numba#10404](https://github.com/numba/numba/issues/10404) - SVML unavailable in numba v0.63.1
- [llvmlite#1380](https://github.com/numba/llvmlite/issues/1380) - FFI fails to build on FreeBSD against llvm20
- [llvmlite#1388](https://github.com/numba/llvmlite/issues/1388) - Re-enable `pylief` based test for 3.14t on CI
- [llvmlite#1390](https://github.com/numba/llvmlite/issues/1390) - Port SVML patch to LLVM 20 to re-enable SVML

### Closed Issues
- [numba#10376](https://github.com/numba/numba/issues/10376) - Numba 0.63.1 Checklist
- [numba#10389](https://github.com/numba/numba/issues/10389) - About RISC-V
- [numba#10392](https://github.com/numba/numba/issues/10392) - When the following code is run repeatedly, the returned results are not consistent every time.
- [numba#10397](https://github.com/numba/numba/issues/10397) - Large overhead for calling certain numpy functions (e.g., np.linalg.solve, np.linalg.slogdet) in numba
- [llvmlite#1385](https://github.com/numba/llvmlite/issues/1385) - Python 3.13.7
- [llvmlite#1389](https://github.com/numba/llvmlite/issues/1389) - LLVM CMake export states LLVM version is 21, llvmlite only officially supports 20.

## 2. New PRs
- [numba#10381](https://github.com/numba/numba/pull/10381) - Improve CFG algo performance for #5611
- [numba#10382](https://github.com/numba/numba/pull/10382) - Update GitHub Artifact Actions (major)
- [numba#10383](https://github.com/numba/numba/pull/10383) - Support `__del__` in `jitclass` with tests and docs
- [numba#10384](https://github.com/numba/numba/pull/10384) - adding osx-64 to the Tier 2 listing
- [numba#10386](https://github.com/numba/numba/pull/10386) - Fix median to support nan
- [numba#10387](https://github.com/numba/numba/pull/10387) - [perf] precompute number of assignments for singly assigned variables loop
- [numba#10395](https://github.com/numba/numba/pull/10395) - make tanh handling of potential overflow for large real more consistent with cpython cmathmodule.c
- [numba#10396](https://github.com/numba/numba/pull/10396) - Add missing type promotions for pythonapi
- [numba#10405](https://github.com/numba/numba/pull/10405) - Update conda-incubator/setup-miniconda action to v3.3.0
- [llvmlite#1387](https://github.com/numba/llvmlite/pull/1387) - Add support for building llvmlite wheels on Windows ARM64
- [llvmlite#1392](https://github.com/numba/llvmlite/pull/1392) - chore(deps): update conda-incubator/setup-miniconda action to v3.3.0

### Closed PRs
- merged - [numba#10375](https://github.com/numba/numba/pull/10375) - Fix #10373: Move `_reload_init` to `CodeLibrary`
- merged - [numba#10377](https://github.com/numba/numba/pull/10377) - Cherry picks and changelog update for 0.63.1
- merged - [numba#10378](https://github.com/numba/numba/pull/10378) - Cherry pick 0.63.1
- [numba#10385](https://github.com/numba/numba/pull/10385) - Fix median to support nan
- merged - [numba#10390](https://github.com/numba/numba/pull/10390) - Fix #10357 `np.asfortranarray` mutability
- merged - [numba#10391](https://github.com/numba/numba/pull/10391) - update wheel builder workflows to install pre-release llvmlite versions
- merged - [numba#10393](https://github.com/numba/numba/pull/10393) - add Numpy 2.4 support
- merged - [numba#10398](https://github.com/numba/numba/pull/10398) - Fix #9931 integer power precision issue for python3.11
- merged - [numba#10401](https://github.com/numba/numba/pull/10401) - Fix: add main label on upload workflow
- merged - [llvmlite#1381](https://github.com/numba/llvmlite/pull/1381) - fix ci buildscripts for flake8 errors
- merged - [llvmlite#1382](https://github.com/numba/llvmlite/pull/1382) - update release checklist
- merged - [llvmlite#1383](https://github.com/numba/llvmlite/pull/1383) - gha/fix llvmdev build disk space issue
- merged - [llvmlite#1384](https://github.com/numba/llvmlite/pull/1384) - chore(deps): update github artifact actions (major)
- merged - [llvmlite#1386](https://github.com/numba/llvmlite/pull/1386) - enable py-lief test dependency for py314 on llvmlite recipe
- merged - [llvmlite#1391](https://github.com/numba/llvmlite/pull/1391) - fix llvmlite_evaluate.py for scheduled events