Revise network definitions for HorizonOS

[Meziu]

@ian-h-chamberlain @FenrirWolf @adryzz

Following some discussion in rust3ds/ctru-rs#174.

The changes include:

• Update to the hostent definition for HorizonOS (armv6k-nintendo-3ds) following upstream changes.

• A step back on the decision to remove the trailing zero bytes from sockaddr_in that was made all the way back in Horizon OS Nintendo 3DS - Fixed size of sockaddr_in #2725. This change will require a little work in std, but we came to understand it's only right to keep the definitions true to the upstream versions.

Also, should this PR be flagged for release in 0.2.x as well as the 1.0 milestone? Work on tier-3 targets isn't clearly talked about in the new contribution guidelines (not that this PR is of any relevance to the release process).

[rustbot]

r? @JohnTitor

rustbot has assigned @JohnTitor.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[tgross35]

Thanks for the fix! Looks like this is defined here https://libctru.devkitpro.org/in_8h_source.html. Can you verify that tests pass with ./ci/run.sh relevant-target when checking Horizon?

Also, should this PR be flagged for release in 0.2.x as well as the 1.0 milestone? Work on tier-3 targets isn't clearly talked about in the new contribution guidelines (not that this PR is of any relevance to the release process).

You are welcome to stable nominate anything (the label works via rustbot). I don't mind backporting as long as it's not overly conflicty.

[tgross35]

Also, do you have a link to the relevant headers where hostent is defined? Trying to get these crosslinked in all PRs.

@rustbot author

[FenrirWolf]

hostent is in https://libctru.devkitpro.org/netdb_8h_source.html

[Meziu]

Thanks for the fix! Looks like this is defined here https://libctru.devkitpro.org/in_8h_source.html. Can you verify that tests pass with ./ci/run.sh relevant-target when checking Horizon?

The test (rightfully) panics at this line because our target has no specified tests. I will add that our target isn't a host target (it's basically an embedded device) and that cross-compiling and running tests is a problem we've had for quite a long time due to the very weird setup with the required toolchain libraries.

Is it necessary to implement these tests? I might be able to cook up something that checks out the headers (assuming the host has all the libraries installed), but it would serve little purpose as the toolchain used is (and will always be) fundamentally experimental.

[tgross35]

Thanks for the change, looks good now! Could you please squash?

Regarding the tests, no need to do anything here. At some point in the future you are welcome to add a similar function, but it doesn't need to happen now.