Add support for illumos target (0.6.x)

[pfmooney]

With support for an illumos target on its way into rust, the cfg guards for solaris should be updated to include illumos as well.

[pfmooney]

This will help untangle illumos from Solaris, as noted in #1263.

[pfmooney]

Updated to require minimum working libc version bump and reflect alphabetical sorting request made in #1295.

[Thomasdezeeuw]

Let's focus on getting #1295 merged first, I don't know if we want to add support for a new OS to v0.6.

[jclulow]

Hi @Thomasdezeeuw!

From our perspective in the illumos community, we're pretty keen to get 0.6 support done before 0.7 support. Here is a small shell script that collects all of the requested versions of mio from https://crates.io:

#!/bin/bash

set -o errexit
set -o pipefail

crate="mio"
url="https://crates.io/api/v1/crates/$crate/reverse_dependencies?page="

n=1
while :; do
	if ! res=$(curl -s -f "${url}${n}"); then
		printf 'ERROR: request page %d failed\n' "$n" >&2
		exit 1
	fi

	if ! len=$(jq '.versions | length' <<< "$res"); then
		exit 1
	fi

	if (( len == 0 )); then
		exit 0
	fi

	for (( i = 0; i < len; i++ )); do
		if ! name=$(jq -r ".versions[$i].crate" <<< "$res") ||
		    ! req=$(jq -r ".dependencies[$i].req" <<< "$res"); then
			exit 1
		fi

		# printf '%-24s %s\n' "$req" "$name"
		printf '%s\n' "$req"
	done

	(( n++ ))
done

Based on the output of this script, it's pretty plain that the 0.6 branch is by far and away the most common in the ecosystem:

$ ./assess.sh > /tmp/mio_versions.txt
$ wc -l /tmp/mio_versions.txt
259 /tmp/mio_versions.txt
$ grep -c '^\^0.6' /tmp/mio_versions.txt
215

215 of the 259 crates currently depend on a 0.6 version of mio, perhaps most prominently the flagship module tokio! While we're keen to help out with 0.7, as evidenced by Patrick's other PR, we'd like to be able to build and run all that existing software on our platform today.

Please let us know how we can help to move this forward!

[Thomasdezeeuw]

Ok we've decided to accept this for v0.6 once rustc is updated (rust-lang/rust#71145).

Please let us know how we can help to move this forward!

For the long term we need two things:

• CI: preferably where we can run cargo test, but being able to run cargo check is sufficient. Seeing how its a new target I don't expect tier 1 support in Rustc so we can't (easily) cross check from another OS at the moment.
• Maintainer(s): as far as I know no-one on the Mio team currently has access to a Illumos machine, so if we can ping someone for Illumos specific bug reports that would be great.

[jclulow]

Ok we've decided to accept this for v0.6 once rustc is updated (rust-lang/rust#71145).

Thank you very much! Deeply appreciated.

For the long term we need two things:

• CI: preferably where we can run cargo test, but being able to run cargo check is sufficient. Seeing how its a new target I don't expect tier 1 support in Rustc so we can't (easily) cross check from another OS at the moment.

We're definitely conscious of not being supported by the big CI/CD providers, and can look into what we can do for mio.

• Maintainer(s): as far as I know no-one on the Mio team currently has access to a Illumos machine, so if we can ping someone for Illumos specific bug reports that would be great.

I have created a team, @illumos/rust, to house a list of people you can ask about issues! I've added @pfmooney and myself to start off.

[jclulow]

Looking closer at this, I'm not sure if people outside an organisation can see a particular team. If you want to create a GitHub team in the tokio-rs organisation to track the list of maintainers you can ping for illumos issues, Patrick and I would be happy to be listed there, too.

[Thomasdezeeuw]

@jclulow @pfmooney I've created @tokio-rs/illumos and invited both of you. I couldn't make it a sub-team of Mio so you might get people pinging that team on other tokio repos.

As for this pr, it LGTM, but the CI is having some problems (I'm trying to fix it in #1297).

[pfmooney]

As for this pr, it LGTM, but the CI is having some problems (I'm trying to fix it in #1297).

Is there something I should do to re-trigger the CI build?

[kleimkuhler]

@pfmooney If you can merge master branch in to this, the FreeBSD check was just updated and should not hang anymore.

Edit: Nevermind this is merging into 0.6 branch. May have to do the same update there. I will get that change in.

[pfmooney]

Edit: Nevermind this is merging into 0.6 branch. May have to do the same update there. I will get that change in.

I'm happy to rebase my bits once the 0.6.x branch is in the desired state.

[kleimkuhler]

@pfmooney You should be okay to rebase now. Thanks for the patience on this!

[pfmooney]

@pfmooney You should be okay to rebase now. Thanks for the patience on this!

Thanks for getting the 0.6.x CI situation squared away!

[kleimkuhler]

@Thomasdezeeuw should take one more look, but looks good to me.

[Thomasdezeeuw]

@pfmooney I assume this is a sufficient up-to-date version, I didn't manually check this.

[Thomasdezeeuw]

@pfmooney @jclulow thanks!