Add Python generator support for deprecated interface fields

[Pengkun-ZHU]

Implements the Python generator side of field-level deprecation support. This PR is based on and should be merged as a companion to ros2/rosidl#945, which establishes the IDL annotation pipeline for C and C++ generators.

When a message field is annotated with @deprecated(text='...') in an .idl file, the generated Python message class will:

• Decorate the property getter and setter with @typing_extensions.deprecated(text) (PEP 702), enabling static analysis tools (mypy, pyright, IDEs) to flag deprecated field usage at development time
• Emit a DeprecationWarning at runtime when the field is accessed or set

The C extension _msg_support.c.em wraps internal accesses to deprecated fields with DISABLE_DEPRECATED_PUSH/POP to suppress warnings in generated conversion code.

Related

• Depends on: Allow deprecating fields of interfaces rosidl#945

[Pengkun-ZHU]

@InvincibleRMC @jmachowinski, I did a Python support following the implementation in ros2/rosidl#945. Could you help give it a review?

[christophebedard]

I'll let @InvincibleRMC take a look, but FYI you should target the rolling branch. Please rebase on rolling and then modify the PR to target the rolling branch.

[InvincibleRMC]

@Pengkun-ZHU What happens if you do equality checks? Do those emit deprecation warnings? I don't think they should.

[Pengkun-ZHU]

@Pengkun-ZHU What happens if you do equality checks? Do those emit deprecation warnings? I don't think they should.

Nice catch, I'll make __eq__ use self._<field> instead of getter to avoid deprecation warning. Similarly I'll do it for __repr__. Any more to add?
On the other hand, I do feel __init__ still needs warns if a deprecated field is passed — that seems like the right behavior to alert users who are explicitly constructing messages with deprecated fields.

[Pengkun-ZHU]

I'll let @InvincibleRMC take a look, but FYI you should target the rolling branch. Please rebase on rolling and then modify the PR to target the rolling branch.

Duly noted with many thanks. I'll rebase and modify.

[Pengkun-ZHU]

Closing in favor of a new PR targeting rolling branch.

[Pengkun-ZHU]

New PR targeting rolling branch is filed:
#256

[christophebedard]

FYI, in the future: You can rebase your branch and then change the target branch of your PR to rolling; no need to re-open a new PR.