kernel: Restore prim_tty signals option to pass control bytes through in raw mode

[gilbertwong96]

Fixes #11512.

Expose the fix in the public API instead of the internal prim_tty module:
shell:start_interactive({noshell, #{mode => raw, signals => false}}) now
disables the terminal driver's signal and flow control handling in raw
mode, so control bytes (ctrl+o, ctrl+c, ctrl+s/q) reach the application
as data. The option is passed down through user_drv to prim_tty and can
be switched on and off at runtime. On Windows, ENABLE_PROCESSED_INPUT is
cleared.

The test runs a noshell peer on a run_erl pty and verifies that ctrl+o is
delivered, then swallowed, then delivered again as the option is toggled.

[github-actions]

CT Test Results

    5 files    288 suites   3h 1m 24s ⏱️
5 523 tests 5 000 ✅ 520 💤 3 ❌
6 900 runs  6 294 ✅ 603 💤 3 ❌

For more details on these failures, see this check.

Results for commit f6d030c.

♻️ This comment has been updated with latest results.

To speed up review, make sure that you have read Contributing to Erlang/OTP and that all checks pass.

See the TESTING and DEVELOPMENT HowTo guides for details about how to run test locally.

Artifacts

• Complete CT logs (Download Logs)
• HTML Documentation (Download HTML Docs)
• Windows Installer

// Erlang/OTP Github Action Bot

[garazdawi]

Hello! Thanks for the pr!

prim_tty is an undocumented internal module, so there are no guarantees that it will retain functionality or be backwards compatible in any way. So this API needs to be added in a public module somewhere. I propose that we expand shell:start_interactive/1 to include it. Something like:

-spec start_interactive(noshell | {noshell, raw | cooked | #{ mode => raw | cooked, signals => boolean()}} | {module(), atom(), [term()]}) ->
          ok | {error, already_started};

and then pass that option all the way down to prim_tty. The code should also be able to handle that we switch the option on and off again, so the tests should check that that is possible.

Also windows needs to be handled as well and there it is the ENABLE_PROCESSED_INPUT flag that needs to be cleared.

As this is a new API it will not be patched into older versions, but released in the next feature release.

In the future, if you find that you need to use some internal module, please create an issue and let us know so that we can create a public API for it so that it does not get lost.

[gilbertwong96]

Thanks for the review! I've reworked the PR along these lines:

• The option is now exposed via shell:start_interactive/1 as {noshell, #{mode => raw, signals => false}}, with signals => boolean() defaulting to true; the old {noshell, raw | cooked} forms still work.
• It is passed down through user_drv to prim_tty, and the test verifies it can be switched on and off at runtime (ctrl+o delivered -> swallowed -> delivered again).
• Windows is handled: ENABLE_PROCESSED_INPUT is cleared when signals => false.
• The test runs in the to_erl group (run_erl pty) since prim_tty is not part of the public API anymore.

The PR is now based on master since this is a new API.