Avoid showing the 130 error code after hitting ^C

[steelman]

When a user hits ^C on command line to cancel entering a lengthy command, $! in __lp_set_prompt is set to 130 (128 + SIGINT). Displaying this error on the input line (especially in warning colours) makes little sense. Therefore, liquidprompt should check if the command submitted is a real command or is it part of PROMPT_COMMAND in which case the error does not come from a user's command and should not be presented.

Technical checklist:

• code follows our shell standards:
  • correct use of IFS
  • careful quoting
  • cautious array access
  • portable array indexing with _LP_FIRST_INDEX
  • printing a "%" character is done with _LP_PERCENT
  • functions/variable naming conventions
  • functions have local variables
  • data functions have optimization guards (early exits)
  • subshells are avoided as much as possible
  • string substitutions may be done differently in Bash and Zsh (use _LP_SHELL_*)
• tests and checks have been added, ran, and their warnings fixed:
  • unit tests have been updated (see tests/test_*.sh files)
  • ran tests.sh
  • ran shellcheck.sh (requires shellcheck).
• documentation have been updated accordingly:
  • functions and attributes are documented in alphabetical order
  • new sections are documented in the default theme
  • tag .. versionadded:: X.Y or .. versionchanged:: Y.Z
  • functions signatures have arguments, returned code, and set value(s)
  • attributes have types and defaults
  • ran docs/docs-lint.sh (requires Python 3 and requirements.txt
    installed (cd docs/; python3 -m venv venv; . venv/bin/activate; pip install -r requirements.txt))

[Rycieos]

Interesting idea. I can see why one would want this. And I likely will use this once we merge it. However, it needs to be configurable.

• Add a config option to control this, like LP_HIDE_EMPTY_ERROR. Don't forget to add docs for it as well.
• The tracking variable _LP_NOT_A_REAL_COMMAND is very confusing. Rename it to _LP_REAL_COMMAND, and invert the Boolean logic.

[Rycieos]

Bash also doesn't run __lp_before_command() if all the features that depend on it are disabled.

[Rycieos]

I was going to post a detailed review of how you missed a couple important things, but the more I dug and reread, I realized that you are fighting against a terrible codebase, which is my fault of course.

So I decided to fix things up as I saw them. Take a look at my new branch rework/preexec, specifically the first two commits eb334b2 and 436b1b5. This largely reworks and improves how preexecs are handled, no longer runs them on empty prompts ([Enter]), and also fixes a related issue where runtime display was not cleared after an empty prompt.

After those fixes, your change is super simple, so I did that too in commit 1be449e. Please take a look at these 3 commits, and let me know if everything makes sense. If so, I'll merge my first 2 into master, and you can continue from there.

If you accept my (our) commit 1be449e, then you pretty much are only missing docs. You will want to add a section for your new config option here. Remember to detail both scenarios this will trigger.

[steelman]

LGTM +1 Let me prepare the docs now.

[Rycieos]

Looks great, except for one "spelling error":

config.rst:1182: : Spell check: Ctrl: after the user hits Ctrl-C or submits an empty command (i.e. empty string.

You will need to add Ctrl to the docs/spelling_wordlist.txt file.