[3007.x] transports: fix and normalize PublishServer method signatures

[Sxderp]

the arguments to publish, specifically io_loop, needs to be consistant across the different transports. This argument was not consistant and caused exceptions.

[welcome]

Hi there! Welcome to the Salt Community! Thank you for making your first contribution. We have a lengthy process for issues and PRs. Someone from the Core Team will follow up as soon as possible. In the meantime, here’s some information that may help as you continue your Salt journey.
Please be sure to review our Code of Conduct. Also, check out some of our community resources including:

• Community Wiki
• Salt’s Contributor Guide
• Join our Community Slack
• IRC on LiberaChat
• Salt Project YouTube channel
• Salt Project Twitch channel

There are lots of ways to get involved in our community. Every month, there are around a dozen opportunities to meet with other contributors and the Salt Core team and collaborate in real time. The best way to keep track is by subscribing to the Salt Community Events Calendar.
If you have additional questions, email us at saltproject@vmware.com. We’re glad you’ve joined our community and look forward to doing awesome things with you!

[Sxderp]

I guess I never gave a full rundown on why this is needed.

The master creates a publish server. The factory method creates a MasterPubServerChannel with the chosen transport using ipc_publish_server, we're going to come back to that function in a moment.

In the pre_fork we call MasterPubServerChannel._publish_domain, which calls the publisher function on the transport. If the arguments are not correct then the function call will fail.

Going back to ipc_publish_server (and ipc_publish_client right next to it). There is an arbitrary restriction on using TCP only for these functions. I'm willing to bet that's because /for some reason/ (cough) zeromq was not working and the solution was to just use TCP instead

After my patch I was able to remove the TCP restriction and everything worked.

I have also added a test for required arguments. Perhaps it may make more sense with abstract classes? I'm not really sure. I'm also not 100% on what folder / file should be modified for the tests.

[Sxderp]

@dwoz I think this is fully ready. The CI is not running "core tests" so my test isn't actually executed. I'm not sure if there's something I need to do for that?

Also perhaps it may make sense to make the test more dynamic. Create and check a "Base" transport and ensure all arguments on that one are present on the child classes. If is that is the best route I can update. Maybe ABC can do that.

[Sxderp]

I have rebased onto 3007.x.

I have also extended the PR to include the other methods of PublishServer.

This will partially fix #66577 (follows are required to completely fix that one).

@dwoz
@dmurphy18

A followup PR should probably be made for PublishClient..

[Sxderp]

I think the write callback in transport.tcp is a holderver from IPCServer. I could not trace any code that actually used it. Local tests seem to agree.

[Sxderp]

@dmurphy18 requesting Workflows get ran again.

[dwoz]

ssl is not a valid option for all transports.

[Sxderp]

Then it should be ignored by that transport. We already have issues with trying to pass the 'ssl' config when a transport does not support it. Do you expect the factory methods to know which transports do and do not support SSL and not pass it along? I think that defeats the purpose of the factory methods.