* global: add RUBY_TWO global variable.
* test-bot: use RUBY_TWO global variable.
* github: produce better curl error messages.
If we don't know why curl has failed then ensure that the error messages
that it produced are included as part of the user output.
Changes to our style configuration:
- Consolidate all rules related to Ruby 1.8 compatibility in one place.
- Codify our de-facto preference for `alias_method` over `alias` (drops
offense count by 54 after turning this on).
- Drop `Style/SignalException` as `only_raise` has been the new default
for quite a while (since RuboCop 0.37.0).
The vendor Ruby will be put inside `Library/Homebrew/vendor/portable-ruby/<version>`,
with a symlink `Library/Homebrew/vendor/portable-ruby/current` pointed to it.
In addition, a `Library/Homebrew/vendor/portable-ruby-version` will
track the latest version of vendor binaries.
This gives us version control on vendor Ruby and enables us to bump vendor
Ruby whenever needed such as security update.
Fix minor regression in b55250c44eacd49646407b26a73753965d4d36d0. The
bottle is placed in the current working directory, is unaffected by the
the single teardown introduced in #475, and thus needs manual handling.
Closes#483.
Signed-off-by: Martin Afanasjew <martin@afanasjew.de>
Test was added in 61614d35294c8faffaeb5c7f0ec6dea78dbbc30f, but never
had any effect because it was replaced by a method of the same name a
few lines further down.
We might as well be consistent about it, since it's not obvious to people
what the next step is after we've pointed out the directory isn't writable.
Fixes#476.
* 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.
Instead of writing a custom ensure for every test let's just nuke
all the files every time. This may be something we might want to use for
other unit tests too. It leans heavily on the fact that a
`FileUtils.rm_rf` on files that don't exist is very quick and things
like `brew cleanup` are super slow in comparison.
Before:
```
$ brew tests --only=integration_cmds --official-cmd-taps
Finished in 49.764724s, 1.0047 runs/s, 5.2648 assertions/s.
```
After:
```
$ brew tests --only=integration_cmds --official-cmd-taps
Finished in 43.014769s, 1.1624 runs/s, 5.8352 assertions/s.
```
The `build` and `source_modified_time` arguments are always coming from
the matching attributes of the `Formula` instance. Thus query `formula`
for them instead of passing them individually.
