Skip to content

Conversation

tianon
Copy link
Member

@tianon tianon commented Feb 27, 2024

This changes the table to be a line per ISA instead of per variant; adds x86-64, POWER8+, and RISC-V; and notes the appropriate "Go analog" in each case (because these values mirror Go's choices, like our use of GOOS and GOARCH).

Of particular note, GOARM64 and GORISCV64 are not fully implemented or released in Go yet, but they are accepted proposals with in-progress implementations or already merged to main:

Closes #852

This changes the table to be a line per ISA instead of per variant; adds x86-64, POWER8+, and RISC-V; and notes the appropriate "Go analog" in each case (because these values mirror Go's choices, like our use of `GOOS` and `GOARCH`).

Of particular note, `GOARM64` and `GORISCV64` are not fully implemented or released in Go yet, but they are accepted proposals with in-progress implementations or already merged to main:

- `GOARM64`: https://golang.org/issue/60905
- `GORISCV64`: https://golang.org/issue/61476#issuecomment-1791156741

Signed-off-by: Tianon Gravi <admwiggin@gmail.com>
| ARM 32-bit, v7 | `arm` | `v7` |
| ARM 32-bit, v8 | `arm` | `v8` |
| ARM 64-bit, v8 | `arm64` | `v8` |
These values SHOULD match (or be similar to) their analog listed in [the Go Language document][go-environment2].
Copy link
Member Author

@tianon tianon Feb 27, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

(or be similar to)

This part makes me nuts, but I felt like I had to somehow address that GOARM does not use a v prefix, GOARM64 will not (at least initially) support v8 (without a minor, like v8.0), we don't do extra things like softfloat or sse2 in this field, etc SOMEHOW in this wording.

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hm, i would be more explicit in tags that are supported. One can reference go constants as valid ones, but specify them explicitly.

There will always be variants that are not yet specified or relevant for go, but do make sense to be explicit for container images that contain code that was written in other languages. For example, whilst GOARM64 is already defined, it doesn't actually do anything different in go - but it does for gcc for example.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The table below does include explicit examples, though 😅

@tianon
Copy link
Member Author

tianon commented Jul 11, 2024

@tianon
Copy link
Member Author

tianon commented Sep 3, 2025

In case it's helpful to someone else, I'll note here that I've copied the current debian:trixie-slim to tianon/test:variants, filtered the list to those with an appropriate variant value that could be set, and added the appropriate and explicit variant values in both the config and index so folks can test runtime behavior:

Each architecture also exists as a single-architecture tagged image manifest so runtime edge case behavior can be more easily fully tested:

  • tianon/test:variant-amd64
  • tianon/test:variant-arm64v8
  • tianon/test:variant-arm32v7
  • tianon/test:variant-arm32v5
  • tianon/test:variant-ppc64le
  • tianon/test:variant-riscv64

Edit: updated arm64v8 image to use v8.0 explicitly as a better testing target / more likely to break things

Edit 2: containerd/platforms#8 (comment) 👀

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Track amd64 variants

4 participants