If all the passed packages are currently installed then use
`runtime_dependencies` to get dependency data more quickly and
accurately.
As discussed in https://github.com/Homebrew/brew/issues/6769.
- This makes use of the new /api/formula-linux and /api/analytics-linux
endpoints in formulae.brew.sh to give Linux users up to date formula and
analytics info for their installed core formulae.
Before, on Linux, the macOS stats for the `ack` formula:
```
$ brew info ack
[...]
==> Analytics
install: 12,422 (30 days), 32,742 (90 days), 97,788 (365 days)
install_on_request: 10,778 (30 days), 28,339 (90 days), 85,202 (365 days)
build_error: 0 (30 days)
```
Now, on Linux, the Linux stats for the `ack` formula:
```
$ brew info ack
[...]
==> Analytics
install: 95 (30 days), 242 (90 days), 737 (365 days)
install_on_request: 94 (30 days), 241 (90 days), 734 (365 days)
build_error: 0 (30 days)
```
f762033a57 introduced a bug such that `reason` is coerced to a boolean, so when displayed it loses information about the actual return value from `MissingFormula.reason`. This patch re-scopes the assignment of `reason` such that the actual reason is retained, while also retaining the sense of the new boolean check.
- The `--local` option doesn't exist in early versions of git (~1.7),
but its behaviour is the default (saving in the current repo's .git
directory).
- As it's the default across versions, we don't need to specify it
everywhere.
- Reported in Linuxbrew/install issue 78.
Output more without requiring `--verbose` and update the messages to
make clear what we're looking for and what we found/didn't find.
This is a little more noisy but it should help explain any delays to
users and make it more obvious what's failed if something does.
- This also required auto-fixes for Layout/EmptyLinesAroundBlockBody and
Layout/InconsistentIndentation once the auto-fixer had got rid of the
"redundant begin"s.
This variable allows a user to configure how often `brew update` runs.
With the previous implementation, however, it was still pretty slow as
it would run some of `brew update` even within the timeout. This feels
related to complaints about Homebrew "feeling slow" since we enabled
this. Address this by:
- making `HOMEBREW_AUTO_UPDATE_SECS` behave more like
`HOMEBREW_NO_AUTO_UPDATE` and entirely skip the invocation of
`brew update` if the relevant tap (homebrew-core or homebrew-cask
depending on command invoked) has been checked in the last
`HOMEBREW_AUTO_UPDATE_SECS`
- Only require/check the update of a single tap rather than every tap
- Increase the default value of `HOMEBREW_AUTO_UPDATE_SECS` from 1m
to 5m to cut people a bit more slack. We're not updating things
often enough for 1m not to feel a bit overkill and 5m feels
appropriate for a Homebrew "session" to only require a single update.
