These were changed to extend to make it easier to determine
where the classes come to in the extended callback but that
means that the file is somewhat inconsistent. On the one
hand we're using class methods and on the other we're extend
self. This cleans that up but now the diff is atrocious and
the blame is even worse. Oh well...
This adds a registry for all modules and classes that
cachable is included in. The registry allows us to
programmatically clear all caches in between tests
so that we don't forget to do that when adding a new
class or refactoring code. The goal here is to reduce
the number of flaky tests in the future.
- keep running the command against all os/arch combinations
as the default
- remove todos and deprecations related to changing the behavior
- create constants for os/arch combinations
- Load paths with no API when needed (e.g. for `brew edit`)
- Use no API mode for `brew log` as it's needed there
- Define sharding format for homebrew-cask and homebrew-core inside
`Tap` methods
- Create new formulae/casks in location defined by these `Tap` methods
- Fix a bug in Formulary that made sharded formulae lookup less
efficient (and possibly broke it for core and some API usage)
- Fix various other hardcoded Formula/Cask directory assumptions
Co-authored-by: Bo Anderson <mail@boanderson.me>
This wasn't working with casks because the bottle tag
would be nil here.
It was refactored to not use the bottle tag because
casks don't have bottles.
I also moved the valid_casks? method to extend/os/mac
because casks only run on macOS and the generic OS
tests were failing before.
- I originally thought this was short for "function", but upon closer
inspection all its usages are to do with filenames. So, use "filename",
it's clearer.
We added the `--all` flag (now renamed to `--eval-all`) for various
commands for this behaviour so let's start deprecating this.
Also, introduce a `HOMEBREW_EVAL_ALL` environment variable to use the
existing, less secure, behaviour by default and avoid passing
`--eval-all` everywhere.
- This also required auto-fixes for Layout/EmptyLinesAroundBlockBody and
Layout/InconsistentIndentation once the auto-fixer had got rid of the
"redundant begin"s.
I benchmarked these both locally and neither use of `Thread.new`
provides any measurable speedup (and in the `readall` case appears to
slow things down) on my 8 core machine.
* readall: move readall logic to new class.
* tap: run readall when tapping.
This will prevent tapping an tap with syntax errors from causing issues
for users.
Fixes#58.
