Add support for armv7l.

[jsirois]

No description provided.

[jsirois]

@Lauszus thanks for getting the ball rolling in a-scie/jump#256.
I think this approach using cross is probably a bit cleaner, but I'd like your feedback. In particular, unlike in your change, I'm just rolling with env::consts::ARCH here for armv7l which means the binary suffix for arm 32 is just arm. I'd like to make sure we agree on a suffix, because I'll want ptex, scie-jump and science to all use the same one.

[jsirois]

@Lauszus note that no cargo config is needed for linker flags, just specifying the musl targets is enough to get you a statically linked executable:

:; file dist/ptex-linux-* | grep -v sha256
dist/ptex-linux-aarch64:        ELF 64-bit LSB executable, ARM aarch64, version 1 (SYSV), statically linked, stripped
dist/ptex-linux-arm:            ELF 32-bit LSB executable, ARM, EABI5 version 1 (SYSV), statically linked, stripped
dist/ptex-linux-x86_64:         ELF 64-bit LSB pie executable, x86-64, version 1 (SYSV), static-pie linked, stripped

I built those on my GNU-Linux x86-64 box using cross run -p package --target {x86_64-unknown-linux-musl,aarch64-unknown-linux-musl,armv7-unknown-linux-musleabihf}.

[benjyw]

Looks great, about as streamlined as I can imagine it being. And I concur that cross-rs is likely the way to go (and it looks like they officially bless tonistiigi/binfmt...)

[Lauszus]

@Lauszus thanks for getting the ball rolling in a-scie/jump#256. I think this approach using cross is probably a bit cleaner, but I'd like your feedback. In particular, unlike in your change, I'm just rolling with env::consts::ARCH here for armv7l which means the binary suffix for arm 32 is just arm. I'd like to make sure we agree on a suffix, because I'll want ptex, scie-jump and science to all use the same one.

Yeah that is better. I will update my PR to use cross as well.

I don't think we should use arm for the executable, as it is very ambiguous. Anything else I have ever seen either uses armv7l or aarch64. This is also what Python standalone uses and what is used for Raspberry Pi wheels.

[Lauszus]

Suggested change

docker run --privileged --rm tonistiigi/binfmt --install all

[Lauszus]

Is this line really needed?

[jsirois]

Yes. Without it, lint works (cross clippy), but not test or package (cross test or cross run -p package). Although the cross images contain a cross-compiler toolchain for each target, that's only 1/2 the battle for commands like cargo test and cargo run, the other 1/2 is actually running the binary that is compiled and that requires qemu / binfmt support.

[jsirois]

I don't think we should use arm for the executable, as it is very ambiguous. Anything else I have ever seen either uses armv7l or aarch64. This is also what Python standalone uses and what is used for Raspberry Pi wheels.

Ok, fixed up to use armv7l like you did in the scie jump.

If this goes green I'll merge and test out the release process. If that goes well, I'll swing back to your similar scie jump change and get that released and then work on science and then Pex.

Thanks again @Lauszus.

[Lauszus]

I don't think we should use arm for the executable, as it is very ambiguous. Anything else I have ever seen either uses armv7l or aarch64. This is also what Python standalone uses and what is used for Raspberry Pi wheels.

Ok, fixed up to use armv7l like you did in the scie jump.

If this goes green I'll merge and test out the release process. If that goes well, I'll swing back to your similar scie jump change and get that released and then work on science and then Pex.

Thanks again @Lauszus.

You are welcome. I can send an initial PR for pex and science if you want, as I started on it, but then stopped as I was waiting for a new jump and ptex release.

[jsirois]

I can send an initial PR for pex and science if you want, as I started on it, but then stopped as I was waiting for a new jump and ptex release.

That sounds great.

[Lauszus]

I can send an initial PR for pex and science if you want, as I started on it, but then stopped as I was waiting for a new jump and ptex release.

That sounds great.

I opened two draft PRs: a-scie/lift#106 and pex-tool/pex#2620. I'll just clean them up a bit.