The Julia developers are pleased to announce that the fourth beta release for Julia v1.7.0 is now available. You can download binaries at https://julialang.org/downloads in the “upcoming release” section for macOS (x86-64, AArch64!), FreeBSD (x86-64), Windows (32-, 64-bit), glibc Linux (i686, x86-64, AArch64, PowerPC), and musl Linux (x86-64). Check out the NEWS file to see what will be new in 1.7.0.
As a beta version, this should not be considered production-ready. Rather, it’s intended to give users, especially package developers, a chance to try our their code with 1.7.0 prior to a full release. Note that 1.7 on Travis, AppVeyor, Cirrus, and GitHub Actions now refers to 1.7.0-beta4.
Let us know in the issue tracker if you run into any issues. Please note that any bugs you may encounter should be posted there rather than being discussed in this thread to ensure visibility to the developers.
Note to ARM-based Mac users: x86 macOS binaries should still work on ARM-based Macs with Rosetta, but the new AArch64 macOS binaries do not require the compatibility layer.
Note to FreeBSD users: Support for FreeBSD 11 has been dropped, with the minimum supported version now 12.0. As such, Julia v1.7.0-beta4 was built on FreeBSD 12 rather than 11, and is unlikely to work on 11.
beta4 was going to be a release candidate possibly, but turned out not to be, i.e. its issue was 1.7-RC1/1.7-beta4.
My guess is we’re going to see 1.6.3, and 1.6.4 first (soon), since the former has been finalized “KristofferC merged commit 11e64d1 into release-1.6 2 days ago” (while not at the download page) and the latter in the works.
Of the open issues, two are regressions (usually easy to find out why, and then fix), thereof one a performance regression, and that seems to be the only holdup, with the two other issues with fixes available already.
EDIT (now fixed): In addition there are other issues that may not have been put on the 1.7 milestone. Since xoshiro256++ is a new major feature in 1.7, this new issue regarding it from 14 hours ago may delay (I made a PR, a partial fix, for that issue): https://github.com/JuliaLang/julia/issues/42150
I want to urge everyone to heavily test the Linear Algebra (esp. BLAS and LAPACK related) capabilities in the run up to 1.7. For those who haven’t followed along, we have a brand new system to pick the BLAS and LAPACK libraries that gives us tremendous flexibility, as discussed in this JuliaCon talk by @giordano and @staticfloat:
The default is just openblas as before and nothing should change compared to 1.6. However, by using MKL, you should be able to replace the use of openblas with MKL. See the notes in the MKL.jl package:
Once 1.7 is released, we will start linking various libraries in BinaryBuilder to libblastrampoline, so that this flexibility (to replace BLAS without needing to build a new system image) can be made available to almost anything that links to BLAS and LAPACK.
