feat: dotnet publish re-entrant

[bruno-garcia]

Use --skip-duplicate to allow the job to be retried

From docs:

When pushing multiple packages to an HTTP(S) server, treats any 409 Conflict response as a warning so that other pushes can continue.

[asottile-sentry]

fwiw the craft targets are intended to not be reentrant to prevent accidental re-release

[bruno-garcia]

Got it. I'm this case we end up with half the packages release and half not. Which is a pain to fix by hand. With this change we'll be able to finish the step at least.

[BYK]

@bruno-garcia when you check off a target in publish, it is skipped. Is this multiple packages in a single target? If so, this may be grounds for granting an exception maybe?

[bruno-garcia]

@bruno-garcia when you check off a target in publish, it is skipped. Is this multiple packages in a single target? If so, this may be grounds for granting an exception maybe?

It's a single target. It's just 1 command dotnet publish and it does the job on all packages.
The command is made re-entrant with the flag I added in this PR.

We don't really gain anything by not having that flag in, right? But definitely would be helpful having the flag in.

fwiw the craft targets are intended to not be reentrant to prevent accidental re-release

In this particular case, it won't re-release anything since dotnet publish either fails if it's already there, or just ignores it (the behavior I need with this PR)

[BYK]

either fails if it's already there, or just ignores it (the behavior I need with this PR)

Just a heads up, I think we should clearly document this as the "just ignores it" is the behavior we explicitly did not want in Craft. Given your circumstances, I think this case deserves the exception but we should clearly document that this is a departure from the norm.

PS: We can also revisit the norm, just saying we should note the odd one out.

[bruno-garcia]

np, I'll add a note to the README