- Remove limitation on OS.
- Add search for Mac OS X developer tools locations by using `xcrun` instead of trying to guess paths.
- Add search for locations based on path of `$GIT_EXE` (set by `lib/theme`).
Short-circuit the loader rather than indenting nearly the whole file.
ALSÖ, assign the `_pip_completion()` handler directly once it's loaded so that we get out of the way once we load it.
Convert from indented if-block to return then unindented code. This should have basically one line change at the top, one line removed at the bottom, and then all whitespace.
* CI: disable Ubuntu 16.04 as it's EOL
https://github.blog/changelog/2021-04-29-github-actions-ubuntu-16-04-lts-virtual-environment-will-be-removed-on-september-20-2021/
* main: lint false positive
* install: lint
* plugins/cmd-returned-notify: don't `export`
* plugins/xterm: lint
* plugins/git: lint
* plugins/goenv: lint
* plugins/alias-completion: lint false positives
* plugins/alias-completion: fix SC2155, SC2154
Declare `locals` at the top of the function
* completion: lint completions using `bash_completion` functions
Match the style of the existing code
* completion/knife: lint false positives
* completion/knife: lint
* completion/sdkman: lint
* completion/composer: lint
* Move `.shellcheckrc` under `themes/`
* lib/theme: fix SC2155, SC2154, SC2034
* lib/colors: don't warn on unused variables
We assign a large number of variables here and they may or may not be used anywhere else, so disable SC2034 for this file (only).
Alsö disable SC2005 as the functions in this file were written before `printf` was invented and have to do some fancy metascripting to get escape sequences interpreted reliably. I’m not smart enough to fix this to use `printf`, so leave it for now.
* themes/agnoster: lint
* themes: disable SC2154 for colors
Each one of these themes will need it’s own fix for SC2154, possibly upstream.
Due to the way themes are, it's entirely normal to have a *lot* of false positives for SC2034. So much so, that I have to admit that it is probably just not worth linting for SC2034 despite my dislike of blanket ignore rules.
* themes: disable SC2154, fix SC2155
Each one of these themes will need it’s own fix for SC2154, possibly upstream.
Due to the way themes are, it's entirely normal to have a *lot* of false positives for SC2034. So much so, that I have to admit that it is probably just not worth linting for SC2034 despite my dislike of blanket ignore rules.
* Delete `.shellcheckrc`
* remove executable bit
Give up and accept defeat that bash-completion can't reasonably be audited for unbound parameters. Wrap invocation with disabling strictness, and restore after if it was enabled.
Use new function `_bash_it_homebrew_check()` in existing plugins and completions which look for Homebrew.
Alsö, use `$OSTYPE` instead of calling external `uname` binary.
* lib/helpers: use `$OSTYPE` instead of `$(uname)`
* plugins/osx: use `$OSTYPE` instead of `$(uname)`
* plugins/boot2docker: use `$OSTYPE` instead of `$(uname)`
* plugins/python: use `$OSTYPE` instead of `$(uname)`
* plugins/base: use `$OSTYPE` instead of `$(uname)`
Alsö, use `[[` instead of `[` as the former has less insane argument handling being shell syntax rather than a builtin command that must emulate being a real binary
* completion/brew: use `$OSTYPE` instead of `$(uname)`
* completion/git: use `$OSTYPE` instead of `$(uname)`
Alsö, use `[[` instead of `[`.
* completion/fabric: use `$OSTYPE` instead of `uname`
* theme/demula: use `$OSTYPE` instead of `$(uname)`
* theme/rana: use `$OSTYPE` instead of `$(uname)`
Need to use grep WITHOUT case-sensitivity. Otherwise results such as com.apple.iChat and com.apple.imagent won't be handled correctly and neither one can be completed.
Fixed another comparison that prevented "defaults read com.apple.iChat <complete>" from correctly returning the available defaults keys in that file.
Bash-completion supports pre-defining $BASH_COMPLETION as the path to the main script, so use that if it's defined.
Alsö, don't load homebrew's completion if we've successfully loaded one already.
Instead of using the profile.d version, just invoke the script. The profile.d script preemptively short-circuits if it thinks that bash-completions has already been loaded, which it does by using the $BASH_COMPLETION variable, which is expressly supported by upstream to specify the location of the script...so it will entirely be never loaded if this is set.
