Dependencies/requirements of a dependency need to be installed/satisfied
before the dependency. The fact that @pour_failed may be false is
irrelevant to that imperative if we weren't pouring to begin with, so
this commit now checks for that case as well.
If a requirement is for a dependent that's already installed and that
dependency is not using a `default_formula` (which would have already
been converted from a `Requirement` to `Dependency` at this stage) then
we want to stop it killing the build.
Read the discussion in https://github.com/Homebrew/homebrew-core/pull/3703. If you
have a better idea, please file a competing PR. I'm sick to death of discussion.
Closes#662.
Signed-off-by: Mike McQuaid <mike@mikemcquaid.com>
Doesn't feel like a good reason to abort installation for users if the
formula author has specified a conflict that doesn't exist. Instead,
behave more like the `TapFormulaUnavailableError` but print a message
telling people to report to the formula authors.
Closes https://github.com/Homebrew/homebrew-versions/issues/1327
When a given dependency appears multiple times in a formula's dependency
tree, the inherited options for that dependency should accumulate rather
than being overwritten each time that dependency is considered by
expand_dependencies. In particular, this impacts "universal" since the
dependency should be built with universal unless all of its instances in
the dependency tree don't have "universal" as opposed to only if the last
instance considered has "universal."
Closes Homebrew/homebrew-core#1604.
Closes#308.
Signed-off-by: ilovezfs <ilovezfs@icloud.com>
When passing formula options with value, e.g. `--with-qt=5`, to the
child process responsible for building a formula, `ARGV.value` would be
invoked with `nil`. Handle this more elegantly (no change in behavior).
For consistency, use a regular expression adapted from `Options.create`
instead of the somewhat bogus one used before.
Also enables sandbox for --interactive and --debug use of install
and test, using automatic retention.
Closes#66.
Signed-off-by: Andrew Janke <andrew@apjanke.net>
- add `HOMEBREW_PRODUCT` global variable
- only differentiate between `/usr/local` and `non-/usr/local` Homebrew
prefixes to avoid sharing sensitive user information
- note if e.g. build errors are occurring under CI
- Add `HOMEBREW_NO_ANALYTICS` variable (this will be how people opt-out
when this is enabled for everyone)
- Add `HOMEBREW_ANALYTICS_DEBUG` variable to output all the analytics
that are sent
- Move Bash analytics code to `Library/Homebrew/utils/analytics.sh`
- Add documentation for our analytics and why/what/when/how and opt-out
- Only official Homebrew commands are reported
- Ruby analytics are now reported in a forked, background process
Calling `compute_dependencies` will make sure both requirements and
dependencies are expanded, so that any referenced taps can be
auto-tapped. Prior to this commit only dependencies were expanded for
the sake of auto-tapping, so dependencies of requirements would cause
installation to fail whenever a tap unavailable exception was
encountered.
ClosesHomebrew/homebrew#50271ClosesHomebrew/homebrew#50281.
Signed-off-by: ilovezfs <ilovezfs@icloud.com>
Previously we got all the options based on the build, any inherited
options, any passed options and the tab. We want to make sure that these
then exclude any options that don't exist as, otherwise, passing an
option that doesn't exist (or has been deleted) is enough to stop a
bottle being poured. This was particularly nasty on upgrades where we
deleted options and one left in the tab would stop the bottle being
poured.
ClosesHomebrew/homebrew#47891.
Signed-off-by: Mike McQuaid <mike@mikemcquaid.com>
Currently `brew install —build-from-source wget` builds all the
dependencies also from source. I can see people wanting to do this when
`HOMEBREW_BUILD_FROM_SOURCE` is set by passing it on the command-line
is mostly just annoying; it means you have to use `—build-bottle` and
deal with the CFLAGS and `post_install` changes if you want to build
from source. Tweak `formula_installer` so this behaviour is more
intuitive.
ClosesHomebrew/homebrew#47889.
Signed-off-by: Mike McQuaid <mike@mikemcquaid.com>
Setting the `FC` environment variable allows the user to override the
default Fortran compiler. The `SharedEnvExtension#fortran` method then
checks whether any of `FCFLAGS` and `FFLAGS` environment variables are
set and, if not, suggests to use `--default-fortran-flags` when running
`brew install`. However, `--default-fortran-flags` never makes it to
`build.rb` (that is executed by `FormulaInstaller`) because the flag is
not whitelisted in `FormulaInstaller#sanitized_ARGV_options`.
This change rectifies this and thereby eliminates the confusing message.
FixesHomebrew/homebrew#47852.
Use FormulaAuditor.audit_prefix_has_contents from audit.rb, which has
better logic.
ClosesHomebrew/homebrew#47678.
Signed-off-by: Mike McQuaid <mike@mikemcquaid.com>
I've set all the variable data to versions that are dependent on the
latest source file date and various modification dates also to the
latest source file date. With this if you rerun `brew bottle`
multiple times in a row you will see the same checksum even if you have
`brew reinstall`ed (as long as upstream does not hardcode e.g. the
build date).
I debugged this with diffoscope and worked on this as part of the
Athens 2015 reproducible builds workshop:
https://reproducible-builds.org/events/athens2015/ClosesHomebrew/homebrew#46587.
Signed-off-by: Mike McQuaid <mike@mikemcquaid.com>
A formula should be built from source by default if the --cc option is
passed to specify a particular compiler.
Added a test to test_formula_installer: test_not_poured_from_bottle_when_compiler_specified
Modified test_formula_installer to assert that the formula was not poured
from a bottle. Similarly modified test_formula_installer_bottle to assert
that the formula *was* installed from a bottle.
Added an install method to the TestballBottle formula (the same as the
Testball formula's install method) so that the TestballBottle formula can
be "built from source".
FixesHomebrew/homebrew#46046 - Build from source should be the default behavior if --cc
option is passed
ClosesHomebrew/homebrew#46162.
Signed-off-by: Mike McQuaid <mike@mikemcquaid.com>
We previously set `ignore_deps?` as true for DependencyInstaller to
avoid duplicated dependencies resolution. (See a9fc82aea30506eeacbddeb8b53fb85de8acb9d4)
However, this will cause problem when pouring bottle of a dependency is
failed. In this case, it will try to build dependency from source but
failed due to uninstalled build deps for this formula.
Another disadvantage for using `ignore_deps?` hack is we cannot
distinguish users passing `--ignore-dependencies` flag from we are in
`DependencyInstaller`.
So, let's differentiate these using `skip_deps_check?`
