There are now a few possible C++ standard libraries a given build could
be using, with subtle incompatibilities and possibility of breakage
when mixed. This makes sure that the dependency chain was compiled in
a compatible manner.
Fortunately all of the Apple compilers use the same libstdc++, and we
don't yet support building with libc++, so this will primarily only
nag users trying to use GNU gcc who already have software installed
with Apple compilers.
Future TODOs:
* Add general support for building with libc++ (compatibility checking
already handled here)
* Possibly track formulae which actually build C++ bindings, so that
users aren't bothered by spurious nagging re: interpreted languages,
pure-C software, etc.
open-uri on Ruby pre-1.8.6 simply will not open https URLs at all,
making it impossible to check the Github API using the same method
as Homebrew.
This may only be disabled temporarily until the issue search is
rewritten, for example to use the curl helper.
The initializer for Formula does a number of validations, but it does
them in a weird order, and some attributes aren't validated under
certain circumstances. This became even more of a mess when most
software package attributes were moved into the SoftwareSpec class.
This commit removes the last vestiges of storing these attributes as
instance variables. In particular, it eliminates #set_instance_variable
and #validate_variable, replacing them with methods that operate on
SoftwareSpec instances, and generate more useful errors.
Doing these validations unconditionally in the initializer means we bail
out much earlier if the formula has invalid attributes or is not fully
specified, and no longer need to validate in #prefix.
Technically we don't need to validate in #brew either, but we continue
to do so anyway as a safety measure, and because we cannot enforce calls
to super in subclasses.
The form "attr :name, true" is deprecated and causes Ruby to emit a
warning in verbose mode. Using attr_{reader,writer,accessor} is more
clear anyway, so do so.
FormulaInstaller now attempts to take a lock on a "foo.brewing" file for
the formula and all of its dependencies before attempting installation.
The lock is an advisory lock implemented using flock(), and as such it
only locks out other processes that attempt to take the lock. It also
means that it is never necessary to manually remove the lock file,
because the lock is not enforced by I/O.
The uninstall, link, and unlink commands all learn to respect this lock
as well, so that the installation cannot be corrupted by a concurrent
Homebrew process, and keg operations cannot occur simultaneously.
This has to be a last resort right? If all we show is that link (plus preceding un-deletable output) and they don't click it then maybe we should just close the ticket and hope they'll use MacPorts instead.
All logs are now stored from each command executed in Formula.install.
Error output is truncated to five lines in an attempt to not overwhelm the user and to encourage users to read the error output and report the bug properly. Maybe we can get that figure up from 70% to 90%.
If you google for "Cowardly refusing to sudo brew" you get a lot of confused users who didn't read any of the Homebrew documentation and then had a hissy-fit.
Rationale: it breaks stuff. You should not be able to install over something already installed *and* linked. brew now prompts you to unlink the existing linked keg first.
I came to fix this because I discovered that `brew install foo` works without --force provided foo's formula version is different to that which is installed and linked. I'm not sure if this was intentionally broken, but it led to at least two tickets I found where people were getting crazy issues trying to install over the top of already installed previous versions.
So I also fixed a whole category of other issues, mostly by moving error handling into FormulaInstaller. Now error's can be caught in Homebrew.install and handled on a formula-by-formula basis. I will next port this behaviour to upgrade.
Also adjust output text slightly for prettiness.
A possibly useful side effect here is safe_system has a defined Exception (subclassing RuntimeError) now.
`DownloadError` is an exception that download stratigies can throw to indicate
that a fetch was incomplete due to a failure in communication.
The `curl` method in `utils.rb` has been upgraded to throw a `DownloadError` if
something bad happens to `curl` execution.
