Add partial support for VERASE and VWERASE in canonical mode in PTYs

[thecodingwizard]

This PR partially fixes #9949 by porting over some code from the Linux kernel's n_tty_receive_char_canon, specifically the eraser function.

It adds partial support for VERASE (backspace) and VWERASE (word erase with ctrl+w).

[ayushr2]

Thanks for all the PRs! As with the other PRs, a test would be appreciated. For example, recently I checked in a PTY fix along with a syscall test in f82d97c.

You can run the syscall test against Linux with: make test TARGETS=//test/syscalls:pty_test_native. Then to test against runsc, run make test TARGETS=//test/syscalls:pty_test_runsc_systrap_directfs.

[ayushr2]

Could you also squash your commits in each PR? When copybara merges this, it applies all the commits onto master (without squashing them).

[thecodingwizard]

Will do. Thanks for providing instructions on how to run the syscall tests!

[ayushr2]

Thanks for providing instructions on how to run the syscall tests!

NP. Running the syscall test directly with bazel may (annoyingly) fail. We have this entire Makefile business going on, in which all bazel commands are run via a bazel Docker container (implemented in tools/bazel.mk).

[kevinGC]

Thanks again for another PR. More comments this time!

Also, we should probably have a simple test for each of VERASE and VWERASE that check that inputing backspace or CTRL+W will have the desired effect.

[thecodingwizard]

I've added some tests, fixed a bug (break should have been continue), refactored some code, and left more comments. Please note that I'm not confident that my comments are actually correct: I think they are, but I do not have extensive experience with terminals.

Thank you for all your feedback -- I don't actually know Golang so your style suggestions, etc. are very helpful! Please let me know if there's anything else I should change or clarify.

[kevinGC]

Good changes, just a few more things.

[thecodingwizard]

Thanks again for your feedback! I:

• Fixed the isContinuationByte bug -- thanks for catching that :)
• Reverted isCtrl check to use toErase < 0x20 -- it was failing tests with unicode.IsControl; I'm not sure if the two are equivalent. In particular, unicode.IsControl(rune(0x80)) seems to return true.
• Rewrote a comment.
• Used std::string instead of variable length arrays; let me know if you prefer std::vector.

Apologies, I must not have run tests properly last night; they should have caught the first two issues.

[kevinGC]

LGTM and thanks for the PR.