[Discuss] Debian Updates for aarch64 Lag?

[richard.pieri at gmail.com (Rich Pieri)]

On Tue, 26 May 2026 07:05:43 -0700
Kent Borg <kentborg at borg.org> wrote:

> When the Copy Fail bug (a short Python program to a `#` root prompt)
> hit the scene on April 29 I got a new kernel for x86_64 on May 1, but
> I didn't get a new aarch_64 kernel until May 13.

I can't tell you why. But I can tell you why not: your entire chain of
reasoning.

https://ftp.debian.org/debian/pool/main/l/linux-signed-amd64/
https://ftp.debian.org/debian/pool/main/l/linux-signed-arm64/

There are no packages dated 2026-05-01 so I picked the linux-image
packages with 2026-05-02 dates. And I see that every AMD64 kernel has a
matching ARM64 kernel (or two) with timestamps within at most a few
minutes of each other. Perhaps there are wider discrepancies but I
didn't look.

linux-image-6.12.85+deb12-amd64_6.12.85-1~bpo12+1_amd64.deb
linux-image-6.12.85+deb12-arm64-16k_6.12.85-1~bpo12+1_arm64.deb
linux-image-6.12.85+deb12-arm64_6.12.85-1~bpo12+1_arm64.deb
linux-image-6.12.85+deb12-cloud-amd64_6.12.85-1~bpo12+1_amd64.deb
linux-image-6.12.85+deb12-cloud-arm64_6.12.85-1~bpo12+1_arm64.deb
linux-image-6.12.85+deb12-rt-amd64_6.12.85-1~bpo12+1_amd64.deb
linux-image-6.12.85+deb12-rt-arm64_6.12.85-1~bpo12+1_arm64.deb
linux-image-6.19.13+deb13-amd64_6.19.13-1~bpo13+1_amd64.deb
linux-image-6.19.13+deb13-arm64-16k_6.19.13-1~bpo13+1_arm64.deb
linux-image-6.19.13+deb13-arm64_6.19.13-1~bpo13+1_arm64.deb
linux-image-6.19.13+deb13-cloud-amd64_6.19.13-1~bpo13+1_amd64.deb
linux-image-6.19.13+deb13-cloud-arm64_6.19.13-1~bpo13+1_arm64.deb
linux-image-6.19.13+deb13-rt-amd64_6.19.13-1~bpo13+1_amd64.deb
linux-image-6.19.13+deb13-rt-arm64_6.19.13-1~bpo13+1_arm64.deb


> I can imagine the build system is automated, but doesn't triaging and 
> testing patches that make it into the build involve humans? The Linux 
> kernel is very impressive for being written almost entirely in C, yet 
> run on different CPUs.

This isn't "amazing" or "impressive". It's what the C language was made
to do.

> But within limits, they periodically drop targets for a reason, it is
> work to support different CPUs.

Which has nothing to do with being written in C, and everything to do
with user demand vis-a-vis people to provide support for that
architecture.

> An aarch_64 kernel is much more different from an x86_64 kernel than
> is a user land program compiled for the two architectures. Isn't it?

No.

> Aren't humans involved? Aren't these same humans dealing with other
> patches? Aren't they a lot busier than they were a couple months ago??

Yes. Probably. Maybe. And none of this has any effect on the build
system because it all happens before the build system.

-- 
\m/ (--) \m/