Bash completion install doc improvements

[scop]

See individual commits for details.

[rbtcollins]

I've never looked particularly closely at completions. What does bash do to avoid running that command on every completion request?

[scop]

What does bash do to avoid running that command on every completion request?

https://www.gnu.org/software/bash/manual/html_node/Programmable-Completion.html has the background details. There's no anchor to link to on that page, but search for "There is some support for dynamically modifying completions" towards the end of it.

More specifically for this particular case, the https://github.com/scop/bash-completion project that I co-maintain is "in charge" of loading completions dynamically from the mentioned .../bash-completion/completions dirs (system and user wide), contains a completion loader that implements the above: https://github.com/scop/bash-completion/blob/fd7eadc541ef15fdb897911613d6939bcc1bb742/bash_completion#L2713-L2724

[scop]

Rebased/resolved conflicts.

[rami3l]

@scop In order to make the lights green again, you still have to update the UI test to match the new output. Please have a look at files in tests/suite/cli-ui/ and change them accordingly (like what was done in e.g. #3542).

[scop]

Updated, but no success yet: something seems to be changing backslash escapes to slashes in the outputs, don't know what should be done about that.

[rami3l]

Updated, but no success yet: something seems to be changing backslash escapes to slashes in the outputs, don't know what should be done about that.

@scop Sorry for the late reply!

I'll paste the original error below for the sake of convenience:

---- expected: stdout
++++ actual:   stdout
   1    1 | ...
   2    2 | Generate tab-completion scripts for your shell
   3    3 | 
   4    4 | Usage: rustup[EXE] completions [shell] [command]
   5    5 | 
          ⋮
  27   27 |     `/usr/share/bash-completion/completions`, and user-specific ones
  28   28 |     can be stored in `~/.local/share/bash-completion/completions`.
  29   29 |     Run the command:
  30   30 | 
  31   31 |         $ mkdir -p ~/.local/share/bash-completion/completions
  32      -         $ printf 'eval -- "$("$1" completions bash)"\n' \
       32 +         $ printf 'eval -- "$("$1" completions bash)"/n' /
  33   33 |               > ~/.local/share/bash-completion/completions/rustup
  34   34 | 
  35   35 |     This installs the completion script. You may have to log out and
  36   36 |     log back in to your shell session for the changes to take effect.
  37   37 | 
          ⋮
  39   39 | 
  40   40 |     Homebrew stores bash completion files within the Homebrew directory.
  41   41 |     With the `bash-completion` brew formula installed, run the command:
  42   42 | 
  43   43 |         $ mkdir -p $(brew --prefix)/etc/bash_completion.d
  44      -         $ printf 'eval -- "$(rustup completions bash)"\n' \
       44 +         $ printf 'eval -- "$(rustup completions bash)"/n' /
  45   45 |               > $(brew --prefix)/etc/bash_completion.d/rustup.bash-completion
  46   46 | 
  47   47 |     Fish:
  48   48 | 
  49   49 |     Fish completion files are commonly stored in
          ⋮
 125  125 |     toolchain. Not all shells are currently supported. Here are examples for
 126  126 |     the currently supported shells.
 127  127 | 
 128  128 |     Bash:
 129  129 | 
 130      -         $ printf 'eval -- "$(rustup completions bash cargo)"\n' \
      130 +         $ printf 'eval -- "$(rustup completions bash cargo)"/n' /
 131  131 |               > ~/.local/share/bash-completion/completions/cargo
 132  132 | 
 133  133 |     Zsh:
 134  134 | 
 135  135 |         $ rustup completions zsh cargo > ~/.zfunc/_cargo

IIRC this has something to do with after_help's markup support.
I'll look into it and hopefully find a solution real soon.

(cc @epage)

Update: As it turns out, the replacement of \s is made by trycmd, as explained in #3177 (comment).

[epage]

trycmd changes backslashes to forward slashes so that paths can be represented in a more neutral way across platforms. This was modeled off of cargo-test-support.

[rami3l]

@epage Thanks for your quick response!

This seems to be assert-rs/snapbox#91 then.
@djc Before resolving that issue, maybe we should just accept the hack and use / in the snapshots?

[djc]

This seems to be assert-rs/trycmd#91 then. @djc Before resolving that issue, maybe we should just accept the hack and use / in the snapshots?

Seems fine?

[djc]

What is the value of using $1 here? It seems rather obfuscated to me.