[gentoo-dev] RFC: ruby-ng.eclass - new eclass for ruby-based packages

[gentoo-dev] RFC: ruby-ng.eclass - new eclass for ruby-based packages

[Hans de Graaff]

[-- Attachment #1.1: Type: text/plain, Size: 1060 bytes --]

Hi,

The ruby team is currently working on two new eclasses to better handle
our ruby packages. This first one is intended to be a replacement for
the current ruby.eclass. A second eclass is forthcoming and will provide
support for handling gems.

This new eclass has much better handling of multiple ruby target
implementations, fixing https://bugs.gentoo.org/show_bug.cgi?id=167029
and provides a better framework to include more functionality later.

We intend to move this eclass to the main tree and gradually move over
the packages currently using ruby.eclass. ruby.eclass will be deprecated
over time but we don't have a specific timeline for that yet.

There is a testbed with ebuilds here:
http://git.overlays.gentoo.org/gitweb/?p=proj/ruby-scripts.git;a=tree;f=ruby-ng-testbed;hb=HEAD

The eclass itself is attached to this message and here:
http://git.overlays.gentoo.org/gitweb/?p=proj/ruby-scripts.git;a=blob;f=ruby-ng-testbed/eclass/ruby-ng.eclass;h=ce9a37440600c7913a21b5780a18b5e8dd9d712f;hb=HEAD

Kind regards,

Hans


[-- Attachment #1.2: ruby-ng.eclass --]
[-- Type: text/plain, Size: 11887 bytes --]

# Copyright 1999-2004 Gentoo Foundation
# Distributed under the terms of the GNU General Public License v2
# $Header: /var/cvsroot/gentoo-x86/eclass/ruby.eclass,v 1.75 2009/03/03 15:51:54 a3li Exp $
#
# @ECLASS: ruby-ng.eclass
# @MAINTAINER:
# Ruby herd <ruby@gentoo.org>
#
# Author: Diego E. Pettenò <flameeyes@gentoo.org>
#
# Author: Alex Legler <a3li@gentoo.org>
#
# @BLURB: An eclass for installing Ruby packages with proper support for multiple Ruby slots.
# @DESCRIPTION:
# The Ruby eclass is designed to allow an easier installation of Ruby packages
# and their incorporation into the Gentoo Linux system.
#
# Currently available targets are:
#  * ruby18 - Ruby (MRI) 1.8.x
#  * ruby19 - Ruby (MRI) 1.9.x
#  * ree18  - Ruby Enterprise Edition 1.8.x
#  * jruby  - JRuby
#
# This eclass does not define the implementation of the configure,
# compile, test, or install phases. Instead, the default phases are
# used.  Specific implementations of these phases can be provided in
# the ebuild either to be run for each Ruby implementation, or for all
# Ruby implementations, as follows:
#
#  * each_ruby_configure
#  * all_ruby_configure

# @ECLASS-VARIABLE: USE_RUBY
# @DESCRIPTION:
# This variable contains a space separated list of targets (see above) a package
# is compatible to. It must be set before the `inherit' call. There is no
# default. All ebuilds are expected to set this variable.

# @ECLASS-VARIABLE: PATCHES
# @DESCRIPTION:
# A String or Array of filenames of patches to apply to all implementations.

# @ECLASS-VARIABLE: RUBY_OPTIONAL
# @DESCRIPTION:
# Set the value to "yes" to make the dependency on a Ruby interpreter optional.

inherit eutils toolchain-funcs

EXPORT_FUNCTIONS src_unpack src_prepare src_configure src_compile src_test src_install pkg_setup

case ${EAPI} in
	2) ;;
	*)
		die "Unsupported EAPI=${EAPI} for ruby-ng.eclass"
esac

# @FUNCTION: ruby_implementation_depend
# @USAGE: target [comparator [version]]
# @RETURN: Package atom of a Ruby implementation to be used in dependencies.
# @DESCRIPTION:
# This function returns the formal package atom for a Ruby implementation.
#
# `target' has to be one of the valid values for USE_RUBY (see above)
# 
# Set `comparator' and `version' to include a comparator (=, >=, etc.) and a
# version string to the returned string
ruby_implementation_depend() {
	local rubypn=
	local rubyslot=

	case $1 in
		ruby18)
			rubypn="dev-lang/ruby"
			rubyslot=":1.8"
			;;
		ruby19)
			rubypn="dev-lang/ruby"
			rubyslot=":1.9"
			;;
		ree18)
			rubypn="dev-lang/ruby-enterprise"
			rubyslot=":1.8"
			;;
		jruby)
			rubypn="dev-java/jruby"
			rubyslot=""
			;;
		*) die "$1: unknown Ruby implementation"
	esac

	echo "$2${rubypn}$3${rubyslot}"
}

# @FUNCTION: ruby_samelib
# @RETURN: use flag string with current ruby implementations
# @DESCRIPTION:
# Convenience function to output the use dependency part of a
# dependency. Used as a building block for ruby_add_rdepend() and
# ruby_add_bdepend(), but may also be useful in an ebuild to specify
# more complex dependencies.
ruby_samelib() {
	local res=
	for _ruby_implementation in $USE_RUBY; do
		has -${_ruby_implementation} $@ || \
			res="${res}ruby_targets_${_ruby_implementation}?,"
	done

	echo "[${res%,}]"
}

_ruby_implementation_depend() {
	echo "ruby_targets_${1}? ( ${2}[ruby_targets_${1}] )"
}

_ruby_add_bdepend() {
	local atom=$1
	local conditions=$2

	for condition in $conditions; do
		atom="${condition}? ( ${atom} )"
	done

	DEPEND="${DEPEND} ${atom}"
	RDEPEND="${RDEPEND}"
}

_ruby_add_rdepend() {
	local atom=$1
	local conditions=$2

	for condition in $conditions; do
		atom="${condition}? ( ${atom} )"
	done

	RDEPEND="${RDEPEND} ${atom}"
	_ruby_add_bdepend "$atom" test
}

# @FUNCTION: ruby_add_rdepend
# @USAGE: [conditions] atom
# @DESCRIPTION:
# Adds the specified atom(s) with optional use condition(s) to
# RDEPEND, taking the current set of ruby targets into account. This
# makes sure that all ruby dependencies of the package are installed
# for the same ruby targets. Use this function for all ruby
# dependencies instead of setting RDEPEND yourself. Both atom and
# conditions can be a space-separated list of atoms or conditions.
ruby_add_rdepend() {
	local atoms=
	local conditions=
	case $# in
		1)
			atoms=$1
			;;
		2)
			conditions=$1
			atoms=$2
			;;
		*)
			die "bad number of arguments to $0"
			;;
	esac

	for atom in $atoms; do
		_ruby_add_rdepend "${atom}$(ruby_samelib)" "$conditions"
	done
}

# @FUNCTION: ruby_add_bdepend
# @USAGE: [conditions] atom
# @DESCRIPTION:
# Adds the specified atom(s) with optional use condition(s) to both
# DEPEND and RDEPEND, taking the current set of ruby targets into
# account. This makes sure that all ruby dependencies of the package
# are installed for the same ruby targets. Use this function for all
# ruby dependencies instead of setting DEPEND and RDEPEND
# yourself. Both atom and conditions can be a space-separated list of
# atoms or conditions.
ruby_add_bdepend() {
	local atoms=
	local conditions=
	case $# in
		1)
			atoms=$1
			;;
		2)
			conditions=$1
			atoms=$2
			;;
		*)
			die "bad number of arguments to $0"
			;;
	esac

	for atom in $atoms; do
		_ruby_add_bdepend "${atom}$(ruby_samelib)" "$conditions"
	done
}

for _ruby_implementation in $USE_RUBY; do
	IUSE="${IUSE} ruby_targets_${_ruby_implementation}"

	# If you specify RUBY_OPTIONAL you also need to take care of
	# ruby useflag and dependency.
	if [[ ${RUBY_OPTIONAL} != "yes" ]]; then
		DEPEND="${DEPEND} ruby_targets_${_ruby_implementation}? ( $(ruby_implementation_depend $_ruby_implementation) )"
		RDEPEND="${RDEPEND} ruby_targets_${_ruby_implementation}? ( $(ruby_implementation_depend $_ruby_implementation) )"
	fi
done

_ruby_invoke_environment() {
	old_S=${S}
	sub_S=${S#${WORKDIR}}

	environment=$1; shift

	my_WORKDIR="${WORKDIR}"/${environment}
	S="${my_WORKDIR}"/"${sub_S}"

	if [[ -d "${S}" ]]; then
		pushd "$S" &>/dev/null
	elif [[ -d "${my_WORKDIR}" ]]; then
		pushd "${my_WORKDIR}" &>/dev/null
	else
		pushd "${WORKDIR}" &>/dev/null
	fi

	ebegin "Running ${_PHASE:-${EBUILD_PHASE}} phase for $environment"
	"$@"
	popd &>/dev/null

	S=${old_S}
}

_ruby_each_implementation() {
	local invoked=no
	for _ruby_implementation in ${USE_RUBY}; do
		# only proceed if it's requested
		use ruby_targets_${_ruby_implementation} || continue

		RUBY=$(type -p $_ruby_implementation 2>/dev/null)
		invoked=yes

		if [[ -n "$1" ]]; then
			_ruby_invoke_environment $_ruby_implementation "$@"
		fi

		unset RUBY
	done

	[[ ${invoked} == "no" ]] && die "You need to select at least one Ruby implementation by setting RUBY_TARGETS in /etc/make.conf."
}

# @FUNCTION: ruby-ng_pkg_setup
# @DESCRIPTION:
# Check whether at least one ruby target implementation is present.
ruby-ng_pkg_setup() {
	# This only checks that at least one implementation is present
	# before doing anything; by leaving the parameters empty we know
	# it's a special case.
	_ruby_each_implementation
}

# @FUNCTION: ruby-ng_src_unpack
# @DESCRIPTION:
# Unpack the source archive, including gems.
ruby-ng_src_unpack() {
	mkdir "${WORKDIR}"/all
	pushd "${WORKDIR}"/all &>/dev/null

	# We don't support an each-unpack, it's either all or nothing!
	if type all_ruby_unpack &>/dev/null; then
		_ruby_invoke_environment all all_ruby_unpack
	else
		[[ -n ${A} ]] && unpack ${A}
	fi

	popd &>/dev/null
}

_ruby_apply_patches() {
	for x in "${PATCHES[@]}"; do
		epatch "${x}"
	done

	# This is a special case: instead of executing just in the special
	# "all" environment, this will actually copy the effects on _all_
	# the other environments, and is thus executed before the copy
	type all_ruby_prepare &>/dev/null && all_ruby_prepare
}

_ruby_source_copy() {
	# Until we actually find a reason not to, we use hardlinks, this
	# should reduce the amount of disk space that is wasted by this.
	cp -prl all ${_ruby_implementation} \
		|| die "Unable to copy ${_ruby_implementation} environment"
}

# @FUNCTION: ruby-ng_src_prepare
# @DESCRIPTION:
# Apply patches and prepare versions for each ruby target
# implementation. Also carry out common clean up tasks.
ruby-ng_src_prepare() {
	# Way too many Ruby packages are prepared on OSX without removing
	# the extra data forks, we do it here to avoid repeating it for
	# almost every other ebuild.
	find . -name '._*' -delete

	_ruby_invoke_environment all _ruby_apply_patches

	_ruby_each_implementation _ruby_source_copy

	if type each_ruby_prepare &>/dev/null; then
		_ruby_each_implementation each_ruby_prepare
	fi
}

# @FUNCTION: ruby-ng_src_configure
# @DESCRIPTION:
# Configure the package.
ruby-ng_src_configure() {
	if type each_ruby_configure &>/dev/null; then
		_ruby_each_implementation each_ruby_configure
	fi

	type all_ruby_configure &>/dev/null && \
		_ruby_invoke_environment all all_ruby_configure
}

# @FUNCTION: ruby-ng_src_compile
# @DESCRIPTION:
# Compile the package.
ruby-ng_src_compile() {
	if type each_ruby_compile &>/dev/null; then
		_ruby_each_implementation each_ruby_compile
	fi

	type all_ruby_compile &>/dev/null && \
		_ruby_invoke_environment all all_ruby_compile
}

# @FUNCTION: ruby-ng_src_test
# @DESCRIPTION:
# Run tests for the package.
ruby-ng_src_test() {
	if type each_ruby_test &>/dev/null; then
		_ruby_each_implementation each_ruby_test
	fi

	type all_ruby_test &>/dev/null && \
		_ruby_invoke_environment all all_ruby_test
}

_each_ruby_check_install() {
	local libruby_basename=$(${RUBY} -rrbconfig -e 'puts Config::CONFIG["LIBRUBY_SO"]')
	local libruby_soname=$(scanelf -qS "/usr/$(get_libdir)/${libruby_basename}" | awk '{ print $1 }')

	# The current implementation lacks libruby (i.e.: jruby)
	[[ -z ${libruby_soname} ]] && return 0

	scanelf -qnR "${D}"/$(dirname $(${RUBY} -rrbconfig -e 'puts Config::CONFIG["sitedir"]')) \
		| fgrep -v "${libruby_soname}" \
		> "${T}"/ruby-ng-${_ruby_implementation}-mislink.log

	if [[ -s "${T}"/ruby-ng-${_ruby_implementation}-mislink.log ]]; then
		ewarn "Extensions installed for ${_ruby_implementation} with missing links to ${libruby}"
		ewarn $(< "${T}"/ruby-ng-${_ruby_implementation}-mislink.log )
		die "Missing links to ${libruby}"
	fi
}

# @FUNCTION: ruby-ng_src_install
# @DESCRIPTION:
# Install the package for each ruby target implementation.
ruby-ng_src_install() {
	if type each_ruby_install &>/dev/null; then
		_ruby_each_implementation each_ruby_install
	fi

	type all_ruby_install &>/dev/null && \
		_ruby_invoke_environment all all_ruby_install

	_PHASE="check install" \
		_ruby_each_implementation _each_ruby_check_install
}

# @FUNCTION: doruby
# @USAGE: file [file...]
# @DESCRIPTION:
# Installs the specified file(s) into the sitelibdir of the Ruby interpreter in ${RUBY}. 
doruby() {
	( # don't want to pollute calling env
		insinto $(${RUBY} -r rbconfig -e 'print Config::CONFIG["sitelibdir"]')
		insopts -m 0644
		doins "$@"
	) || die "failed to install $@"
}

# @FUNCTION: ruby_get_libruby
# @RETURN: The location of libruby*.so belonging to the Ruby interpreter in ${RUBY}.
ruby_get_libruby() {
	${RUBY} -rrbconfig -e 'puts File.join(Config::CONFIG["libdir"], Config::CONFIG["LIBRUBY"])'
}

# @FUNCTION: ruby_get_hdrdir
# @RETURN: The location of the header files belonging to the Ruby interpreter in ${RUBY}.
ruby_get_hdrdir() {
	local rubyhdrdir=$(${RUBY} -rrbconfig -e 'puts Config::CONFIG["rubyhdrdir"]')

	if [[ "${rubyhdrdir}" = "nil" ]] ; then
		rubyhdrdir=$(${RUBY} -rrbconfig -e 'puts Config::CONFIG["archdir"]')
	fi

	echo "${rubyhdrdir}"
}

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

[gentoo-dev] Re: RFC: ruby-ng.eclass - new eclass for ruby-based packages

[Christian Faulhammer]

[-- Attachment #1: Type: text/plain, Size: 866 bytes --]

Hans de Graaff <graaff@gentoo.org>:

> The eclass itself is attached to this message and here:
> http://git.overlays.gentoo.org/gitweb/?p=proj/ruby-scripts.git;a=blob;f=ruby-ng-testbed/eclass/ruby-ng.eclass;h=ce9a37440600c7913a21b5780a18b5e8dd9d712f;hb=HEAD

 At the moment I don't remember for what reason exactly, but we renamed
the PATCHES variable to EBZR_PATCHES in bzr.eclass.  You should do
something similar.

EXPORT_FUNCTIONS src_unpack src_prepare src_configure src_compile
src_test src_install pkg_setup

case ${EAPI} in
	2) ;;
	*)
		die "Unsupported EAPI=${EAPI} for ruby-ng.eclass"
esac

The sense of this I don't understand.  You should not restrict EAPI
like this.

V-Li

-- 
Christian Faulhammer, Gentoo Lisp project
<URL:http://www.gentoo.org/proj/en/lisp/>, #gentoo-lisp on FreeNode

<URL:http://gentoo.faulhammer.org/>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] Re: RFC: ruby-ng.eclass - new eclass for ruby-based packages

[Tomáš Chvátal]

[-- Attachment #1: Type: Text/Plain, Size: 731 bytes --]

Dne středa 02 Prosinec 2009 09:51:25 Christian Faulhammer napsal(a):
>  At the moment I don't remember for what reason exactly, but we renamed
> the PATCHES variable to EBZR_PATCHES in bzr.eclass.  You should do
> something similar.
No reason to rename it if they use base_src_prepare
> 
> EXPORT_FUNCTIONS src_unpack src_prepare src_configure src_compile
> src_test src_install pkg_setup
> 
> case ${EAPI} in
> 	2) ;;
> 	*)
> 		die "Unsupported EAPI=${EAPI} for ruby-ng.eclass"
> esac
> 
> The sense of this I don't understand.  You should not restrict EAPI
> like this.
> 
> V-Li
> 
Why not. Quite standart way.
 19 case ${EAPI:-0} in
 20     2) : ;;
 21     *) DEPEND="EAPI-TOO-OLD" ;;
 22 esac

Tom

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] Re: RFC: ruby-ng.eclass - new eclass for ruby-based packages

[Fabian Groffen]

On 02-12-2009 09:57:37 +0100, Tomáš Chvátal wrote:
> Dne středa 02 Prosinec 2009 09:51:25 Christian Faulhammer napsal(a):
> >  At the moment I don't remember for what reason exactly, but we renamed
> > the PATCHES variable to EBZR_PATCHES in bzr.eclass.  You should do
> > something similar.
> No reason to rename it if they use base_src_prepare
> > 
> > EXPORT_FUNCTIONS src_unpack src_prepare src_configure src_compile
> > src_test src_install pkg_setup
> > 
> > case ${EAPI} in
> > 	2) ;;
> > 	*)
> > 		die "Unsupported EAPI=${EAPI} for ruby-ng.eclass"
> > esac
> > 
> > The sense of this I don't understand.  You should not restrict EAPI
> > like this.
> > 
> > V-Li
> > 
> Why not. Quite standart way.
>  19 case ${EAPI:-0} in
>  20     2) : ;;
>  21     *) DEPEND="EAPI-TOO-OLD" ;;
>  22 esac

I guess you better add a section for known eapis that are too old to
your liking, and use the *) section for an oops message saying you don't
know about ${EAPI} yet, and to please file a bug.



-- 
Fabian Groffen
Gentoo on a different level

[gentoo-dev] Re: RFC: ruby-ng.eclass - new eclass for ruby-based packages

[Christian Faulhammer]

[-- Attachment #1: Type: text/plain, Size: 754 bytes --]

Hi,

Tomáš Chvátal <scarabeus@gentoo.org>:
> > The sense of this I don't understand.  You should not restrict EAPI
> > like this.
> > 
> > V-Li
> > 
> Why not. Quite standart way.
>  19 case ${EAPI:-0} in
>  20     2) : ;;
>  21     *) DEPEND="EAPI-TOO-OLD" ;;
>  22 esac

 Microsoft Windows is standard software, too....you want to use it?  And
if an ebuild uses EAPI 3 it will fail badly, while it isn't too old.  A
sane assumption is that EAPI 3 will at least contain all features from
EAPI 2 with very few adjustments to existing syntax and behaviour if at
all.

V-Li

-- 
Christian Faulhammer, Gentoo Lisp project
<URL:http://www.gentoo.org/proj/en/lisp/>, #gentoo-lisp on FreeNode

<URL:http://gentoo.faulhammer.org/>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] Re: RFC: ruby-ng.eclass - new eclass for ruby-based packages

[Hans de Graaff]

[-- Attachment #1: Type: text/plain, Size: 336 bytes --]

On Wed, 2009-12-02 at 11:46 +0100, Fabian Groffen wrote:

> I guess you better add a section for known eapis that are too old to
> your liking, and use the *) section for an oops message saying you don't
> know about ${EAPI} yet, and to please file a bug.

Yes, good idea. We've gone with this solution.

Kind regards,

Hans

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] Re: RFC: ruby-ng.eclass - new eclass for ruby-based packages

[Hans de Graaff]

[-- Attachment #1: Type: text/plain, Size: 580 bytes --]

On Wed, 2009-12-02 at 09:57 +0100, Tomáš Chvátal wrote:
> Dne středa 02 Prosinec 2009 09:51:25 Christian Faulhammer napsal(a):
> >  At the moment I don't remember for what reason exactly, but we renamed
> > the PATCHES variable to EBZR_PATCHES in bzr.eclass.  You should do
> > something similar.
> No reason to rename it if they use base_src_prepare

We handle the patches ourselves and also in a non-standard way since we
may apply them to multiple copies of the source tree, so I have renamed
this to RUBY_PATCHES to be on the safe side.

Kind regards,

Hans

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 198 bytes --]