[gentoo-dev] [RFC] multibuild.eclass -- a generic pluggable framework to handle multi-variant builds

[gentoo-dev] [RFC] multibuild.eclass -- a generic pluggable framework to handle multi-variant builds

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 2023 bytes --]

Hello,

Recently python-r1 and multilib-build started to share a few bits
of code related to running the build process for multiple 'variants'
of the same package. Over time, the code extended and it is a bit
cumbersome to maintain the two copies and keep them in sync.

Therefore, I'd like to propose a new multibuild.eclass which will
provide a generic framework for eclasses and ebuilds to perform
multi-variant builds.


The eclass currently is intended to provide two functions:
- multibuild_foreach,
- multibuild_parallel_foreach.

These functions are supposed to execute the child process for each
of the enabled variants with support for the following:

- setting BUILD_DIR to an unique location which can be used to build
  the sources,

- teeing logs to a variant-specific log file to make reading them
  easier,

- nesting multiple uses of multibuild.eclass -- e.g. fftw build 3-4
  'float type' variants + multilib.

It also fixes some of the issues that were noticed within
distutils-r1 / python-r1.


I will send patches in the reply to this message. The patches:

1) introduce the initial version of multibuild based on the code
from python-r1 and distutils-r1,

2) improve running 'tee' to a method not requiring subshelling
the called function,

3) fix an issue of trying to write ${WORKDIR%/}-${impl} if S=${WORKDIR},

4) convert multilib-build to use the new eclass,

5) provide python-r1 with a function to obtain enabled implementation
list to make the migration possible,

6) convert python-r1 and distutils-r1 to use the new eclass,

7) move run_in_build_dir() common function to the new eclass,

8) convert sci-libs/fftw to the new eclass as an example of nested use.


That's just the initial approach. If it works fine, the todo lists:

1) supporting nested parallel builds,

2) adding a function to run a snippet with just one ('default') variant,

and possibly more.


What are your thoughts?

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 966 bytes --]

[gentoo-dev] [PATCH 1/8] Initial version of multibuild eclass.

[Michał Górny]

Based on the code from python-r1.
---
 gx86/eclass/multibuild.eclass | 172 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 172 insertions(+)
 create mode 100644 gx86/eclass/multibuild.eclass

diff --git a/gx86/eclass/multibuild.eclass b/gx86/eclass/multibuild.eclass
new file mode 100644
index 0000000..716d34f
--- /dev/null
+++ b/gx86/eclass/multibuild.eclass
@@ -0,0 +1,172 @@
+# Copyright 1999-2013 Gentoo Foundation
+# Distributed under the terms of the GNU General Public License v2
+# $Header: $
+
+# @ECLASS: multibuild
+# @MAINTAINER:
+# Michał Górny <mgorny@gentoo.org>
+# @AUTHOR:
+# Author: Michał Górny <mgorny@gentoo.org>
+# @BLURB: A generic eclass for building multiple variants of packages.
+# @DESCRIPTION:
+# The multibuild eclass aims to provide a generic framework for building
+# multiple 'variants' of a package (e.g. multilib, Python
+# implementations).
+
+case "${EAPI:-0}" in
+	0|1|2|3|4)
+		die "Unsupported EAPI=${EAPI:-0} (too old) for ${ECLASS}"
+		;;
+	5)
+		;;
+	*)
+		die "Unsupported EAPI=${EAPI} (unknown) for ${ECLASS}"
+		;;
+esac
+
+if [[ ! ${_MULTIBUILD} ]]; then
+
+inherit multiprocessing
+
+# @ECLASS-VARIABLE: MULTIBUILD_VARIANTS
+# @DESCRIPTION:
+# An array specifying all enabled variants which multilib_foreach* will
+# execute the process for.
+#
+# In ebuild, it can be set in global scope. Eclasses should set it
+# locally in function scope to support nesting properly.
+#
+# Example:
+# @CODE
+# python_foreach_impl() {
+#	local MULTIBUILD_VARIANTS=( python{2_5,2_6,2_7} ... )
+#	multibuild_foreach_impl python_compile
+# }
+# @CODE
+
+# @ECLASS-VARIABLE: MULTIBUILD_VARIANT
+# @DESCRIPTION:
+# The current variant which the function was executed for.
+#
+# Example value:
+# @CODE
+# python2_6
+# @CODE
+
+# @ECLASS-VARIABLE: MULTIBUILD_ID
+# @DESCRIPTION:
+# The unique identifier for a multibuild run. In a simple run, it is
+# equal to MULTIBUILD_VARIANT. In a nested multibuild environment, it
+# contains the complete selection tree.
+#
+# It can be used to create variant-unique directories and files.
+#
+# Example value:
+# @CODE
+# amd64-double
+# @CODE
+
+# @ECLASS-VARIABLE: BUILD_DIR
+# @DESCRIPTION:
+# The current build directory. In global scope, it is supposed to
+# contain an 'initial' build directory; if unset, it defaults to ${S}.
+#
+# multibuild_foreach() sets BUILD_DIR locally to variant-specific build
+# directories based on the initial value of BUILD_DIR.
+#
+# Example value:
+# @CODE
+# ${WORKDIR}/foo-1.3-python2_6
+# @CODE
+
+# @FUNCTION: multibuild_foreach
+# @USAGE: [<argv>...]
+# @DESCRIPTION:
+# Run the passed command repeatedly for each of the enabled package
+# variants.
+#
+# Each of the runs will have variant-specific BUILD_DIR set, and output
+# teed to a separate log in ${T}.
+#
+# The function returns 0 if all commands return 0, or the first non-zero
+# exit status otherwise. However, it performs all the invocations
+# nevertheless. It is preferred to call 'die' inside of the passed
+# function.
+multibuild_foreach() {
+	debug-print-function ${FUNCNAME} "${@}"
+
+	[[ ${MULTIBUILD_VARIANTS} ]] \
+		|| die "MULTIBUILD_VARIANTS need to be set"
+
+	local bdir=${BUILD_DIR:-${S}}
+	local prev_id=${MULTIBUILD_ID:+${MULTIBUILD_ID}-}
+	local ret=0 lret=0 v
+
+	debug-print "${FUNCNAME}: initial build_dir = ${bdir}"
+
+	for v in "${MULTIBUILD_VARIANTS[@]}"; do
+		local MULTIBUILD_VARIANT=${v}
+		local MULTIBUILD_ID=${prev_id}${v}
+		local BUILD_DIR=${bdir%%/}-${v}
+
+		einfo "${v}: running ${@}" \
+			| tee -a "${T}/build-${MULTIBUILD_ID}.log"
+
+		# _multibuild_parallel() does redirection internally.
+		# this is a hidden API to avoid writing multilib_foreach twice.
+		if [[ ${1} == _multibuild_parallel ]]; then
+			"${@}"
+		else
+			"${@}" 2>&1 | tee -a "${T}/build-${MULTIBUILD_ID}.log"
+		fi
+		lret=${?}
+	done
+	[[ ${ret} -eq 0 && ${lret} -ne 0 ]] && ret=${lret}
+
+	return ${ret}
+}
+
+# @FUNCTION: multibuild_parallel_foreach
+# @USAGE: [<argv>...]
+# @DESCRIPTION:
+# Run the passed command repeatedly for each of the enabled package
+# variants alike multibuild_foreach. Multiple invocations of the command
+# will be performed in parallel, up to MULTIBUILD_JOBS tasks.
+#
+# The function returns 0 if all commands return 0, or the first non-zero
+# exit status otherwise. However, it performs all the invocations
+# nevertheless. It is preferred to call 'die' inside of the passed
+# function.
+multibuild_parallel_foreach() {
+	debug-print-function ${FUNCNAME} "${@}"
+
+	local ret lret
+
+	_multibuild_parallel() {
+		(
+			multijob_child_init
+			"${@}" 2>&1 | tee -a "${T}/build-${MULTIBUILD_ID}.log"
+			exit ${PIPESTATUS[0]}
+		) &
+		multijob_post_fork
+	}
+
+	local opts
+	if [[ ${MULTIBUILD_JOBS} ]]; then
+		opts=-j${MULTIBUILD_JOBS}
+	else
+		opts=${MAKEOPTS}
+	fi
+
+	multijob_init "${opts}"
+	multibuild_foreach _multibuild_parallel "${@}"
+	ret=${?}
+	multijob_finish
+	lret=${?}
+
+	[[ ${ret} -eq 0 ]] && ret=${lret}
+	return ${ret}
+}
+
+_MULTIBUILD=1
+fi
-- 
1.8.1.4

[gentoo-dev] [PATCH 2/8] Use bash redirection to run 'tee' rather than simple pipes.

[Michał Górny]

This allows us to spawn 'tee' as separate process while keeping
the function code executed in the main shell.
---
 gx86/eclass/multibuild.eclass | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)

diff --git a/gx86/eclass/multibuild.eclass b/gx86/eclass/multibuild.eclass
index 716d34f..d42b8a7 100644
--- a/gx86/eclass/multibuild.eclass
+++ b/gx86/eclass/multibuild.eclass
@@ -108,18 +108,20 @@ multibuild_foreach() {
 		local MULTIBUILD_VARIANT=${v}
 		local MULTIBUILD_ID=${prev_id}${v}
 		local BUILD_DIR=${bdir%%/}-${v}
+		local log_fd
 
-		einfo "${v}: running ${@}" \
-			| tee -a "${T}/build-${MULTIBUILD_ID}.log"
+		# redirect_alloc_fd accepts files only. so we need to open
+		# a random file and then reuse the fd for logger process.
+		redirect_alloc_fd log_fd /dev/null
+		eval "exec ${log_fd}> "'>(exec tee -a "${T}/build-${MULTIBUILD_ID}.log")'
+
+		eval 'einfo "${v}: running ${@}" >&'${log_fd}' 2>&1'
 
 		# _multibuild_parallel() does redirection internally.
 		# this is a hidden API to avoid writing multilib_foreach twice.
-		if [[ ${1} == _multibuild_parallel ]]; then
-			"${@}"
-		else
-			"${@}" 2>&1 | tee -a "${T}/build-${MULTIBUILD_ID}.log"
-		fi
+		eval '"${@}" >&'${log_fd}' 2>&1'
 		lret=${?}
+		eval "exec ${log_fd}>&-"
 	done
 	[[ ${ret} -eq 0 && ${lret} -ne 0 ]] && ret=${lret}
 
@@ -145,8 +147,7 @@ multibuild_parallel_foreach() {
 	_multibuild_parallel() {
 		(
 			multijob_child_init
-			"${@}" 2>&1 | tee -a "${T}/build-${MULTIBUILD_ID}.log"
-			exit ${PIPESTATUS[0]}
+			"${@}"
 		) &
 		multijob_post_fork
 	}
-- 
1.8.1.4

[gentoo-dev] [PATCH 3/8] Avoid writing outside WORKDIR if S=${WORKDIR}.

[Michał Górny]

---
 gx86/eclass/multibuild.eclass | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/gx86/eclass/multibuild.eclass b/gx86/eclass/multibuild.eclass
index d42b8a7..a4d5d11 100644
--- a/gx86/eclass/multibuild.eclass
+++ b/gx86/eclass/multibuild.eclass
@@ -99,6 +99,10 @@ multibuild_foreach() {
 		|| die "MULTIBUILD_VARIANTS need to be set"
 
 	local bdir=${BUILD_DIR:-${S}}
+
+	# Avoid writing outside WORKDIR if S=${WORKDIR}.
+	[[ ${bdir%%/} == ${WORKDIR%%/} ]] && bdir=${WORKDIR}/build
+
 	local prev_id=${MULTIBUILD_ID:+${MULTIBUILD_ID}-}
 	local ret=0 lret=0 v
 
-- 
1.8.1.4

[gentoo-dev] [PATCH 4/8] Convert multilib-build to use multibuild.

[Michał Górny]

---
 gx86/eclass/multilib-build.eclass | 45 ++++++++++++++++++---------------------
 1 file changed, 21 insertions(+), 24 deletions(-)

diff --git a/gx86/eclass/multilib-build.eclass b/gx86/eclass/multilib-build.eclass
index b1afd85..4321e45 100644
--- a/gx86/eclass/multilib-build.eclass
+++ b/gx86/eclass/multilib-build.eclass
@@ -23,7 +23,7 @@ case ${EAPI:-0} in
 	*) die "EAPI=${EAPI} is not supported" ;;
 esac
 
-inherit multilib multiprocessing
+inherit multibuild multilib
 
 # @ECLASS-VARIABLE: _MULTILIB_FLAGS
 # @INTERNAL
@@ -85,6 +85,19 @@ multilib_get_enabled_abis() {
 	fi
 }
 
+# @FUNCTION: _multilib_multibuild_wrapper
+# @USAGE: <argv>...
+# @INTERNAL
+# @DESCRIPTION:
+# Initialize the environment for ABI selected for multibuild.
+_multilib_multibuild_wrapper() {
+	debug-print-function ${FUNCNAME} "${@}"
+
+	local ABI=${MULTIBUILD_VARIANT}
+	multilib_toolchain_setup "${ABI}"
+	"${@}"
+}
+
 # @FUNCTION: multilib_foreach_abi
 # @USAGE: <argv>...
 # @DESCRIPTION:
@@ -95,14 +108,11 @@ multilib_get_enabled_abis() {
 # If multilib support is disabled, it just runs the commands. No setup
 # is done.
 multilib_foreach_abi() {
-	local initial_dir=${BUILD_DIR:-${S}}
+	debug-print-function ${FUNCNAME} "${@}"
 
-	local ABI
-	for ABI in $(multilib_get_enabled_abis); do
-		multilib_toolchain_setup "${ABI}"
-		local BUILD_DIR=${initial_dir%%/}-${ABI}
-		"${@}" | tee -a "${T}/build-${ABI}.log"
-	done
+	local MULTIBUILD_VARIANTS=( $(multilib_get_enabled_abis) )
+
+	multibuild_foreach _multilib_multibuild_wrapper "${@}"
 }
 
 # @FUNCTION: multilib_parallel_foreach_abi
@@ -118,24 +128,11 @@ multilib_foreach_abi() {
 #
 # Useful for running configure scripts.
 multilib_parallel_foreach_abi() {
-	local initial_dir=${BUILD_DIR:-${S}}
-
-	multijob_init
-
-	local ABI
-	for ABI in $(multilib_get_enabled_abis); do
-		(
-			multijob_child_init
-
-			multilib_toolchain_setup "${ABI}"
-			local BUILD_DIR=${initial_dir%%/}-${ABI}
-			"${@}" 2>&1 | tee -a "${T}/build-${ABI}.log"
-		) &
+	debug-print-function ${FUNCNAME} "${@}"
 
-		multijob_post_fork
-	done
+	local MULTIBUILD_VARIANTS=( $(multilib_get_enabled_abis) )
 
-	multijob_finish
+	multibuild_parallel_foreach _multilib_multibuild_wrapper "${@}"
 }
 
 # @FUNCTION: multilib_check_headers
-- 
1.8.1.4

[gentoo-dev] [PATCH 5/8] python-r1: calculate final list of enabled impls for foreach.

[Michał Górny]

---
 gx86/eclass/python-r1.eclass | 61 +++++++++++++++++++++++++++-----------------
 1 file changed, 37 insertions(+), 24 deletions(-)

diff --git a/gx86/eclass/python-r1.eclass b/gx86/eclass/python-r1.eclass
index 310859e..a1d9228 100644
--- a/gx86/eclass/python-r1.eclass
+++ b/gx86/eclass/python-r1.eclass
@@ -586,6 +586,22 @@ _python_check_USE_PYTHON() {
 	fi
 }
 
+# @FUNCTION: _python_obtain_impls
+# @INTERNAL
+# @DESCRIPTION:
+# Set up the enabled implementation list.
+_python_obtain_impls() {
+	MULTIBUILD_VARIANTS=()
+
+	for impl in "${_PYTHON_ALL_IMPLS[@]}"; do
+		if has "${impl}" "${PYTHON_COMPAT[@]}" \
+			&& use "python_targets_${impl}"
+		then
+			MULTIBUILD_VARIANTS+=( "${impl}" )
+		fi
+	done
+}
+
 # @FUNCTION: python_foreach_impl
 # @USAGE: <command> [<args>...]
 # @DESCRIPTION:
@@ -608,33 +624,30 @@ python_foreach_impl() {
 	local impl
 	local bdir=${BUILD_DIR:-${S}}
 	local ret=0 lret=0
+	local MULTIBUILD_VARIANTS
 
 	debug-print "${FUNCNAME}: bdir = ${bdir}"
-	for impl in "${_PYTHON_ALL_IMPLS[@]}"; do
-		if has "${impl}" "${PYTHON_COMPAT[@]}" \
-			&& _python_impl_supported "${impl}" \
-			&& use "python_targets_${impl}"
-		then
-			local EPYTHON PYTHON
-			python_export "${impl}" EPYTHON PYTHON
-			local BUILD_DIR=${bdir%%/}-${impl}
-			export EPYTHON PYTHON
-
-			einfo "${EPYTHON}: running ${@}" \
-				| tee -a "${T}/build-${EPYTHON}.log"
-
-			# _python_parallel() does redirection internally.
-			# note: this is a hidden API to avoid writing python_foreach_impl
-			# twice. do *not* even think of using it anywhere else.
-			if [[ ${1} == _python_parallel ]]; then
-				"${@}"
-			else
-				"${@}" 2>&1 | tee -a "${T}/build-${EPYTHON}.log"
-			fi
-			lret=${?}
-
-			[[ ${ret} -eq 0 && ${lret} -ne 0 ]] && ret=${lret}
+	_python_obtain_impls
+	for impl in "${MULTIBUILD_VARIANTS[@]}"; do
+		local EPYTHON PYTHON
+		python_export "${impl}" EPYTHON PYTHON
+		local BUILD_DIR=${bdir%%/}-${impl}
+		export EPYTHON PYTHON
+
+		einfo "${EPYTHON}: running ${@}" \
+			| tee -a "${T}/build-${EPYTHON}.log"
+
+		# _python_parallel() does redirection internally.
+		# note: this is a hidden API to avoid writing python_foreach_impl
+		# twice. do *not* even think of using it anywhere else.
+		if [[ ${1} == _python_parallel ]]; then
+			"${@}"
+		else
+			"${@}" 2>&1 | tee -a "${T}/build-${EPYTHON}.log"
 		fi
+		lret=${?}
+
+		[[ ${ret} -eq 0 && ${lret} -ne 0 ]] && ret=${lret}
 	done
 
 	return ${ret}
-- 
1.8.1.4

[gentoo-dev] [PATCH 6/8] Convert python-r1 to use multibuild.

[Michał Górny]

---
 gx86/eclass/python-r1.eclass | 74 +++++++++++++-------------------------------
 1 file changed, 21 insertions(+), 53 deletions(-)

diff --git a/gx86/eclass/python-r1.eclass b/gx86/eclass/python-r1.eclass
index a1d9228..fb9032e 100644
--- a/gx86/eclass/python-r1.eclass
+++ b/gx86/eclass/python-r1.eclass
@@ -48,7 +48,7 @@ elif [[ ${_PYTHON_ANY_R1} ]]; then
 	die 'python-r1.eclass can not be used with python-any-r1.eclass.'
 fi
 
-inherit python-utils-r1
+inherit multibuild python-utils-r1
 
 # @ECLASS-VARIABLE: PYTHON_COMPAT
 # @REQUIRED
@@ -602,6 +602,21 @@ _python_obtain_impls() {
 	done
 }
 
+# @FUNCTION: _python_multibuild_wrapper
+# @USAGE: <command> [<args>...]
+# @INTERNAL
+# @DESCRIPTION:
+# Initialize the environment for Python implementation selected
+# for multibuild.
+_python_multibuild_wrapper() {
+	debug-print-function ${FUNCNAME} "${@}"
+
+	local -x EPYTHON PYTHON
+	python_export "${MULTIBUILD_VARIANT}" EPYTHON PYTHON
+
+	"${@}"
+}
+
 # @FUNCTION: python_foreach_impl
 # @USAGE: <command> [<args>...]
 # @DESCRIPTION:
@@ -621,36 +636,10 @@ python_foreach_impl() {
 	_python_validate_useflags
 	_python_check_USE_PYTHON
 
-	local impl
-	local bdir=${BUILD_DIR:-${S}}
-	local ret=0 lret=0
 	local MULTIBUILD_VARIANTS
-
-	debug-print "${FUNCNAME}: bdir = ${bdir}"
 	_python_obtain_impls
-	for impl in "${MULTIBUILD_VARIANTS[@]}"; do
-		local EPYTHON PYTHON
-		python_export "${impl}" EPYTHON PYTHON
-		local BUILD_DIR=${bdir%%/}-${impl}
-		export EPYTHON PYTHON
-
-		einfo "${EPYTHON}: running ${@}" \
-			| tee -a "${T}/build-${EPYTHON}.log"
-
-		# _python_parallel() does redirection internally.
-		# note: this is a hidden API to avoid writing python_foreach_impl
-		# twice. do *not* even think of using it anywhere else.
-		if [[ ${1} == _python_parallel ]]; then
-			"${@}"
-		else
-			"${@}" 2>&1 | tee -a "${T}/build-${EPYTHON}.log"
-		fi
-		lret=${?}
 
-		[[ ${ret} -eq 0 && ${lret} -ne 0 ]] && ret=${lret}
-	done
-
-	return ${ret}
+	multibuild_foreach _python_multibuild_wrapper "${@}"
 }
 
 # @FUNCTION: python_parallel_foreach_impl
@@ -672,32 +661,11 @@ python_foreach_impl() {
 python_parallel_foreach_impl() {
 	debug-print-function ${FUNCNAME} "${@}"
 
-	local ret lret
-
-	_python_parallel() {
-		(
-			multijob_child_init
-			"${@}" 2>&1 | tee -a "${T}/build-${EPYTHON}.log"
-			exit ${PIPESTATUS[0]}
-		) &
-		multijob_post_fork
-	}
-
-	local opts
-	if [[ ${DISTUTILS_JOBS} ]]; then
-		opts=-j${DISTUTILS_JOBS}
-	else
-		opts=${MAKEOPTS}
-	fi
-
-	multijob_init "${opts}"
-	python_foreach_impl _python_parallel "${@}"
-	ret=${?}
-	multijob_finish
-	lret=${?}
+	local MULTIBUILD_JOBS=${MULTIBUILD_JOBS:-${DISTUTILS_JOBS}}
+	local MULTIBUILD_VARIANTS
+	_python_obtain_impls
 
-	[[ ${ret} -eq 0 ]] && ret=${lret}
-	return ${ret}
+	multibuild_parallel_foreach _python_multibuild_wrapper "${@}"
 }
 
 # @FUNCTION: python_export_best
-- 
1.8.1.4

[gentoo-dev] [PATCH 7/8] Move run_in_build_dir() to multibuild.eclass.

[Michał Górny]

---
 gx86/eclass/multibuild.eclass | 19 +++++++++++++++++++
 gx86/eclass/python-r1.eclass  | 19 -------------------
 2 files changed, 19 insertions(+), 19 deletions(-)

diff --git a/gx86/eclass/multibuild.eclass b/gx86/eclass/multibuild.eclass
index a4d5d11..e15b74e 100644
--- a/gx86/eclass/multibuild.eclass
+++ b/gx86/eclass/multibuild.eclass
@@ -173,5 +173,24 @@ multibuild_parallel_foreach() {
 	return ${ret}
 }
 
+# @FUNCTION: run_in_build_dir
+# @USAGE: <argv>...
+# @DESCRIPTION:
+# Run the given command in the directory pointed by BUILD_DIR.
+run_in_build_dir() {
+	debug-print-function ${FUNCNAME} "${@}"
+	local ret
+
+	[[ ${#} -ne 0 ]] || die "${FUNCNAME}: no command specified."
+	[[ ${BUILD_DIR} ]] || die "${FUNCNAME}: BUILD_DIR not set."
+
+	pushd "${BUILD_DIR}" >/dev/null || die
+	"${@}"
+	ret=${?}
+	popd >/dev/null || die
+
+	return ${ret}
+}
+
 _MULTIBUILD=1
 fi
diff --git a/gx86/eclass/python-r1.eclass b/gx86/eclass/python-r1.eclass
index fb9032e..6c1dc3e 100644
--- a/gx86/eclass/python-r1.eclass
+++ b/gx86/eclass/python-r1.eclass
@@ -735,24 +735,5 @@ python_replicate_script() {
 	done
 }
 
-# @FUNCTION: run_in_build_dir
-# @USAGE: <argv>...
-# @DESCRIPTION:
-# Run the given command in the directory pointed by BUILD_DIR.
-run_in_build_dir() {
-	debug-print-function ${FUNCNAME} "${@}"
-	local ret
-
-	[[ ${#} -ne 0 ]] || die "${FUNCNAME}: no command specified."
-	[[ ${BUILD_DIR} ]] || die "${FUNCNAME}: BUILD_DIR not set."
-
-	pushd "${BUILD_DIR}" >/dev/null || die
-	"${@}"
-	ret=${?}
-	popd >/dev/null || die
-
-	return ${ret}
-}
-
 _PYTHON_R1=1
 fi
-- 
1.8.1.4

[gentoo-dev] [PATCH 8/8] fftw: example use of multibuild in ebuild.

[Michał Górny]

Just a quick, dirty example. Not even tested thoroughly ;).
---
 gx86/sci-libs/fftw/fftw-3.3.3-r1.ebuild | 38 +++++++++++++--------------------
 1 file changed, 15 insertions(+), 23 deletions(-)

diff --git a/gx86/sci-libs/fftw/fftw-3.3.3-r1.ebuild b/gx86/sci-libs/fftw/fftw-3.3.3-r1.ebuild
index 18554f0..ddca8e4 100644
--- a/gx86/sci-libs/fftw/fftw-3.3.3-r1.ebuild
+++ b/gx86/sci-libs/fftw/fftw-3.3.3-r1.ebuild
@@ -7,7 +7,7 @@ EAPI=5
 #AUTOTOOLS_AUTORECONF=1
 FORTRAN_NEEDED=fortran
 
-inherit autotools-multilib eutils flag-o-matic fortran-2 toolchain-funcs versionator
+inherit autotools-multilib eutils flag-o-matic fortran-2 multibuild toolchain-funcs versionator
 
 DESCRIPTION="Fast C library for the Discrete Fourier Transform"
 HOMEPAGE="http://www.fftw.org/"
@@ -24,6 +24,8 @@ DEPEND="${RDEPEND}
 	test? ( dev-lang/perl )"
 
 pkg_setup() {
+	# XXX: this looks like it should be used with BUILD_TYPE!=binary
+
 	if use openmp; then
 		if [[ $(tc-getCC) == *gcc ]] && ! tc-has-openmp; then
 			ewarn "OpenMP is not available in your current selected gcc"
@@ -32,13 +34,13 @@ pkg_setup() {
 		FORTRAN_NEED_OPENMP=1
 	fi
 	fortran-2_pkg_setup
-	FFTW_DIRS="single double longdouble"
+	MULTIBUILD_VARIANTS=( single double longdouble )
 	if use quad; then
 		if [[ $(tc-getCC) == *gcc ]] && ! version_is_at_least 4.6 $(gcc-version); then
 			ewarn "quad precision only available for gcc >= 4.6"
 			die "need quad precision capable gcc"
 		fi
-		FFTW_DIRS+=" quad"
+		MULTIBUILD_VARIANTS+=( quad )
 	fi
 }
 
@@ -57,7 +59,9 @@ src_configure() {
 	# filter -Os according to docs
 	replace-flags -Os -O2
 
-	for x in ${FFTW_DIRS}; do
+	my_configure() {
+		local x=${MULTIBUILD_VARIANT}
+
 		myeconfargs=(
 			$(use_enable fma)
 			$(use_enable fortran)
@@ -93,42 +97,30 @@ src_configure() {
 			die "${x} precision not implemented in this ebuild"
 		fi
 
-		einfo "Configuring for ${x} precision"
-		BUILD_DIR="${S}-${x}" \
-			autotools-multilib_src_configure
-	done
+		autotools-multilib_src_configure
+	}
+
+	multibuild_foreach my_configure
 }
 
 src_compile() {
-	for x in ${FFTW_DIRS}; do
-		einfo "Compiling for ${x} precision"
-		BUILD_DIR="${S}-${x}" \
-			autotools-multilib_src_compile
-	done
+	multibuild_foreach autotools-multilib_src_compile
 }
 
 src_test () {
-	do_smalltest() { cd "${BUILD_DIR}" && emake -C tests smallcheck; }
 	# We want this to be a reasonably quick test, but that is still hard...
 	ewarn "This test series will take 30 minutes on a modern 2.5Ghz machine"
 	# Do not increase the number of threads, it will not help your performance
 	#local testbase="perl check.pl --nthreads=1 --estimate"
 	#		${testbase} -${p}d || die "Failure: $n"
-	for x in ${FFTW_DIRS}; do
-		einfo "Testing ${x} precision"
-		BUILD_DIR="${S}-${x}" \
-			multilib_foreach_abi do_smalltest
-	done
+	multibuild_foreach autotools-multilib_src_compile -C tests smallcheck
 }
 
 src_install () {
 	local u x
 	DOCS=( AUTHORS ChangeLog NEWS README TODO COPYRIGHT CONVENTIONS )
 	HTML_DOCS=( doc/html/ )
-	for x in ${FFTW_DIRS}; do
-		BUILD_DIR="${S}-${x}" \
-			autotools-multilib_src_install
-	done
+	multibuild_foreach autotools-multilib_src_install
 
 	if use doc; then
 		dodoc doc/*.pdf
-- 
1.8.1.4

Re: [gentoo-dev] [PATCH 2/8] Use bash redirection to run 'tee' rather than simple pipes.

[Alec Warner]

On Wed, Feb 27, 2013 at 1:43 PM, Michał Górny <mgorny@gentoo.org> wrote:
> This allows us to spawn 'tee' as separate process while keeping
> the function code executed in the main shell.

Can you explain why this is 'better'? I'd prefer way more
documentation in the code itself. You are using a number of what I'd
consider 'banned' constructs and you don't go very far out of your way
to explain why you cannot live without them.

> ---
>  gx86/eclass/multibuild.eclass | 19 ++++++++++---------
>  1 file changed, 10 insertions(+), 9 deletions(-)
>
> diff --git a/gx86/eclass/multibuild.eclass b/gx86/eclass/multibuild.eclass
> index 716d34f..d42b8a7 100644
> --- a/gx86/eclass/multibuild.eclass
> +++ b/gx86/eclass/multibuild.eclass
> @@ -108,18 +108,20 @@ multibuild_foreach() {
>                 local MULTIBUILD_VARIANT=${v}
>                 local MULTIBUILD_ID=${prev_id}${v}
>                 local BUILD_DIR=${bdir%%/}-${v}
> +               local log_fd
>
> -               einfo "${v}: running ${@}" \
> -                       | tee -a "${T}/build-${MULTIBUILD_ID}.log"
> +               # redirect_alloc_fd accepts files only. so we need to open
> +               # a random file and then reuse the fd for logger process.
> +               redirect_alloc_fd log_fd /dev/null
> +               eval "exec ${log_fd}> "'>(exec tee -a "${T}/build-${MULTIBUILD_ID}.log")'
The quoting here is a nightmare.
Are the single quotes required?
Do the leading single quotes have to be *right next to* the double quotes?
Why is there a space between the first > and the "?
vim hi-lighting makes this more readable, but still weird looking.
Also please document clearly why 'eval' is required (it is not clear
to me..or I suspect most readers ;p)
Are we not just making an fd and pointing the 'read end' at a tee subprocess?

> +
> +               eval 'einfo "${v}: running ${@}" >&'${log_fd}' 2>&1'
And again here, why can't we just redirect directly to log_fd?

>
>                 # _multibuild_parallel() does redirection internally.
>                 # this is a hidden API to avoid writing multilib_foreach twice.
We do not use the _multibuild_parellel stuff anymore...is this comment relevant?

> -               if [[ ${1} == _multibuild_parallel ]]; then
> -                       "${@}"
> -               else
> -                       "${@}" 2>&1 | tee -a "${T}/build-${MULTIBUILD_ID}.log"
> -               fi
> +               eval '"${@}" >&'${log_fd}' 2>&1'
>                 lret=${?}
> +               eval "exec ${log_fd}>&-"
>         done
>         [[ ${ret} -eq 0 && ${lret} -ne 0 ]] && ret=${lret}
>
> @@ -145,8 +147,7 @@ multibuild_parallel_foreach() {
>         _multibuild_parallel() {
>                 (
>                         multijob_child_init
> -                       "${@}" 2>&1 | tee -a "${T}/build-${MULTIBUILD_ID}.log"
> -                       exit ${PIPESTATUS[0]}
> +                       "${@}"
>                 ) &
>                 multijob_post_fork
>         }
> --
> 1.8.1.4
>
>

Re: [gentoo-dev] [PATCH 2/8] Use bash redirection to run 'tee' rather than simple pipes.

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 2984 bytes --]

On Thu, 28 Feb 2013 05:09:15 -0800
Alec Warner <antarus@gentoo.org> wrote:

> On Wed, Feb 27, 2013 at 1:43 PM, Michał Górny <mgorny@gentoo.org> wrote:
> > This allows us to spawn 'tee' as separate process while keeping
> > the function code executed in the main shell.
> 
> Can you explain why this is 'better'? I'd prefer way more
> documentation in the code itself. You are using a number of what I'd
> consider 'banned' constructs and you don't go very far out of your way
> to explain why you cannot live without them.

Well, if we used simple pipes, like:

  foo | tee ...

Then any variable exported by foo would be discarded. This broke a few
ebuilds already.

By banned constructs, do you mean 'eval'? I hate it too but wasn't able
to solve it anyhow better.

> >  gx86/eclass/multibuild.eclass | 19 ++++++++++---------
> >  1 file changed, 10 insertions(+), 9 deletions(-)
> >
> > diff --git a/gx86/eclass/multibuild.eclass b/gx86/eclass/multibuild.eclass
> > index 716d34f..d42b8a7 100644
> > --- a/gx86/eclass/multibuild.eclass
> > +++ b/gx86/eclass/multibuild.eclass
> > @@ -108,18 +108,20 @@ multibuild_foreach() {
> >                 local MULTIBUILD_VARIANT=${v}
> >                 local MULTIBUILD_ID=${prev_id}${v}
> >                 local BUILD_DIR=${bdir%%/}-${v}
> > +               local log_fd
> >
> > -               einfo "${v}: running ${@}" \
> > -                       | tee -a "${T}/build-${MULTIBUILD_ID}.log"
> > +               # redirect_alloc_fd accepts files only. so we need to open
> > +               # a random file and then reuse the fd for logger process.
> > +               redirect_alloc_fd log_fd /dev/null
> > +               eval "exec ${log_fd}> "'>(exec tee -a "${T}/build-${MULTIBUILD_ID}.log")'
> The quoting here is a nightmare.
> Do the leading single quotes have to be *right next to* the double quotes?
> Why is there a space between the first > and the "?
> vim hi-lighting makes this more readable, but still weird looking.

Yes, I know. I'm thinking of trying to put it all into a single eval,
and just use \$ rather than mixing single and double quotes.

> Also please document clearly why 'eval' is required (it is not clear
> to me..or I suspect most readers ;p)

Well, for the very simple reason that 'fd=2; echo 11 >&${fd}' doesn't
work. The eval simply changes that into 'echo 11>&2'.

> Are we not just making an fd and pointing the 'read end' at a tee subprocess?

Yes, exactly that.

> > +
> > +               eval 'einfo "${v}: running ${@}" >&'${log_fd}' 2>&1'
> And again here, why can't we just redirect directly to log_fd?
> 
> >
> >                 # _multibuild_parallel() does redirection internally.
> >                 # this is a hidden API to avoid writing multilib_foreach twice.
> We do not use the _multibuild_parellel stuff anymore...is this comment relevant?

You are correct, my mistake.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 966 bytes --]

[gentoo-dev] Further changes to multibuild.eclass

[Michał Górny]

Four more patches so far:

1) print only 'public' part of command-line

That is, shift over all the arguments starting with an underscore
for einfo.

  multibuild_foreach_variant _python_multibuild_wrapper foo

will be printed as:

  pythonX.Y: foo


2) add a function to run the command for the 'best' (default, preferred)
variant.


3) use it to implement multilib_for_best_abi().

A few ebuilds will happily use it instead of the current hacks.


4) use it to run *_all() phases in distutils-r1.

This fixes bug 460016 (python_configure_all() doesn't have best-impl if
there's no python_configure()).

[gentoo-dev] [PATCH 1/4] multibuild: print only 'public' part of command-line.

[Michał Górny]

Shift the unnecessary 'private' commands from the printed commands when
executing.

That is:

	python_parallel_foreach_impl foo

will print:

	* pythonX.Y: foo

rather than:

	* pythonX.Y: _multibuild_parallel _python_multibuild_wrapper ...
---
 gx86/eclass/multibuild.eclass | 15 +++++++++++++--
 1 file changed, 13 insertions(+), 2 deletions(-)

diff --git a/gx86/eclass/multibuild.eclass b/gx86/eclass/multibuild.eclass
index 1c9058a..1cc33a9 100644
--- a/gx86/eclass/multibuild.eclass
+++ b/gx86/eclass/multibuild.eclass
@@ -118,12 +118,23 @@ multibuild_foreach_variant() {
 		# redirect_alloc_fd accepts files only. so we need to open
 		# a random file and then reuse the fd for logger process.
 		redirect_alloc_fd log_fd /dev/null
+
+		_multibuild_run() {
+			# find the first non-private command
+			local i=1
+			while [[ ${!i} == _* ]]; do
+				(( i += 1 ))
+			done
+
+			einfo "${v}: running ${@:${i}}"
+			"${@}"
+		}
+
 		# bash can't handle ${log_fd} in redirections,
 		# we need to use eval to pass fd numbers directly.
 		eval "
 			exec ${log_fd}> >(exec tee -a \"\${T}/build-\${MULTIBUILD_ID}.log\")
-			einfo \"\${v}: running \${@}\" >&${log_fd} 2>&1
-			\"\${@}\" >&${log_fd} 2>&1
+			_multibuild_run \"\${@}\" >&${log_fd} 2>&1
 			lret=\${?}
 			exec ${log_fd}>&-
 		"
-- 
1.8.1.4

[gentoo-dev] [PATCH 2/4] multibuild: add multibuild_for_best_variant().

[Michał Górny]

---
 gx86/eclass/multibuild.eclass | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/gx86/eclass/multibuild.eclass b/gx86/eclass/multibuild.eclass
index 1cc33a9..3ffe9e8 100644
--- a/gx86/eclass/multibuild.eclass
+++ b/gx86/eclass/multibuild.eclass
@@ -185,6 +185,26 @@ multibuild_parallel_foreach_variant() {
 	return ${ret}
 }
 
+# @FUNCTION: multibuild_for_best_variant
+# @USAGE: [<argv>...]
+# @DESCRIPTION:
+# Run the passed command once, for the best of the enabled package
+# variants.
+#
+# The run will have a proper, variant-specificBUILD_DIR set, and output
+# teed to a separate log in ${T}.
+#
+# The function returns command exit status.
+multibuild_for_best_variant() {
+	debug-print-function ${FUNCNAME} "${@}"
+
+	[[ ${MULTIBUILD_VARIANTS} ]] \
+		|| die "MULTIBUILD_VARIANTS need to be set"
+
+	local MULTIBUILD_VARIANTS=( "${MULTIBUILD_VARIANTS[-1]}" )
+	multibuild_foreach_variant "${@}"
+}
+
 # @FUNCTION: run_in_build_dir
 # @USAGE: <argv>...
 # @DESCRIPTION:
-- 
1.8.1.4

[gentoo-dev] [PATCH 3/4] multilib-build: introduce multilib_for_best_abi().

[Michał Górny]

---
 gx86/eclass/multilib-build.eclass | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/gx86/eclass/multilib-build.eclass b/gx86/eclass/multilib-build.eclass
index b6409d8..0fe46a0 100644
--- a/gx86/eclass/multilib-build.eclass
+++ b/gx86/eclass/multilib-build.eclass
@@ -139,6 +139,18 @@ multilib_parallel_foreach_abi() {
 	multibuild_parallel_foreach_variant _multilib_multibuild_wrapper "${@}"
 }
 
+# @FUNCTION: multilib_for_best_abi
+# @USAGE: <argv>...
+# @DESCRIPTION:
+# Runs the given command with setup for the 'best' (usually native) ABI.
+multilib_for_best_abi() {
+	debug-print-function ${FUNCNAME} "${@}"
+
+	local MULTIBUILD_VARIANTS=( $(multilib_get_enabled_abis) )
+
+	multibuild_for_best_variant _multilib_multibuild_wrapper "${@}"
+}
+
 # @FUNCTION: multilib_check_headers
 # @DESCRIPTION:
 # Check whether the header files are consistent between ABIs.
-- 
1.8.1.4

[gentoo-dev] [PATCH 4/4] distutils-r1: reuse multibuild_for_best_variant.

[Michał Górny]

---
 gx86/eclass/distutils-r1.eclass | 25 ++++---------------------
 1 file changed, 4 insertions(+), 21 deletions(-)

diff --git a/gx86/eclass/distutils-r1.eclass b/gx86/eclass/distutils-r1.eclass
index 40dae36..ba48a35 100644
--- a/gx86/eclass/distutils-r1.eclass
+++ b/gx86/eclass/distutils-r1.eclass
@@ -549,11 +549,6 @@ distutils-r1_run_phase() {
 	then
 		popd >/dev/null || die
 	fi
-
-	# Store them for reuse.
-	_DISTUTILS_BEST_IMPL=(
-		"${EPYTHON}" "${PYTHON}" "${BUILD_DIR}" "${PYTHONPATH}"
-	)
 }
 
 # @FUNCTION: _distutils-r1_run_common_phase
@@ -567,23 +562,11 @@ distutils-r1_run_phase() {
 _distutils-r1_run_common_phase() {
 	local DISTUTILS_ORIG_BUILD_DIR=${BUILD_DIR}
 
-	local EPYTHON=${_DISTUTILS_BEST_IMPL[0]}
-	local PYTHON=${_DISTUTILS_BEST_IMPL[1]}
-	local BUILD_DIR=${_DISTUTILS_BEST_IMPL[2]}
-	local PYTHONPATH=${_DISTUTILS_BEST_IMPL[3]}
-
-	export EPYTHON PYTHON PYTHONPATH
-
-	if [[ ${DISTUTILS_IN_SOURCE_BUILD} ]]; then
-		pushd "${BUILD_DIR}"/.. >/dev/null || die
-	fi
+	local MULTIBUILD_VARIANTS
+	_python_obtain_impls
 
-	einfo "common: running ${1}"
-	"${@}"
-
-	if [[ ${DISTUTILS_IN_SOURCE_BUILD} ]]; then
-		popd >/dev/null || die
-	fi
+	multibuild_for_best_variant _python_multibuild_wrapper \
+		distutils-r1_run_phase "${@}"
 }
 
 # @FUNCTION: _distutils-r1_run_foreach_impl
-- 
1.8.1.4

Re: [gentoo-dev] [PATCH 1/4] multibuild: print only 'public' part of command-line.

[Alec Warner]

On Sat, Mar 2, 2013 at 1:42 PM, Michał Górny <mgorny@gentoo.org> wrote:
> Shift the unnecessary 'private' commands from the printed commands when
> executing.
>
> That is:
>
>         python_parallel_foreach_impl foo
>
> will print:
>
>         * pythonX.Y: foo
>
> rather than:
>
>         * pythonX.Y: _multibuild_parallel _python_multibuild_wrapper ...
> ---
>  gx86/eclass/multibuild.eclass | 15 +++++++++++++--
>  1 file changed, 13 insertions(+), 2 deletions(-)
>
> diff --git a/gx86/eclass/multibuild.eclass b/gx86/eclass/multibuild.eclass
> index 1c9058a..1cc33a9 100644
> --- a/gx86/eclass/multibuild.eclass
> +++ b/gx86/eclass/multibuild.eclass
> @@ -118,12 +118,23 @@ multibuild_foreach_variant() {
>                 # redirect_alloc_fd accepts files only. so we need to open
>                 # a random file and then reuse the fd for logger process.
>                 redirect_alloc_fd log_fd /dev/null
> +
> +               _multibuild_run() {
> +                       # find the first non-private command
> +                       local i=1
> +                       while [[ ${!i} == _* ]]; do
> +                               (( i += 1 ))
> +                       done
> +
> +                       einfo "${v}: running ${@:${i}}"

So this is an einfo with an assignment side-effect? Can we perhaps
make the assignment explicit?

-A

> +                       "${@}"
> +               }
> +
>                 # bash can't handle ${log_fd} in redirections,
>                 # we need to use eval to pass fd numbers directly.
>                 eval "
>                         exec ${log_fd}> >(exec tee -a \"\${T}/build-\${MULTIBUILD_ID}.log\")
> -                       einfo \"\${v}: running \${@}\" >&${log_fd} 2>&1
> -                       \"\${@}\" >&${log_fd} 2>&1
> +                       _multibuild_run \"\${@}\" >&${log_fd} 2>&1
>                         lret=\${?}
>                         exec ${log_fd}>&-
>                 "
> --
> 1.8.1.4
>
>

Re: [gentoo-dev] [PATCH 1/4] multibuild: print only 'public' part of command-line.

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 1735 bytes --]

On Sat, 2 Mar 2013 14:52:25 -0800
Alec Warner <antarus@gentoo.org> wrote:

> On Sat, Mar 2, 2013 at 1:42 PM, Michał Górny <mgorny@gentoo.org> wrote:
> > Shift the unnecessary 'private' commands from the printed commands when
> > executing.
> >
> > That is:
> >
> >         python_parallel_foreach_impl foo
> >
> > will print:
> >
> >         * pythonX.Y: foo
> >
> > rather than:
> >
> >         * pythonX.Y: _multibuild_parallel _python_multibuild_wrapper ...
> > ---
> >  gx86/eclass/multibuild.eclass | 15 +++++++++++++--
> >  1 file changed, 13 insertions(+), 2 deletions(-)
> >
> > diff --git a/gx86/eclass/multibuild.eclass b/gx86/eclass/multibuild.eclass
> > index 1c9058a..1cc33a9 100644
> > --- a/gx86/eclass/multibuild.eclass
> > +++ b/gx86/eclass/multibuild.eclass
> > @@ -118,12 +118,23 @@ multibuild_foreach_variant() {
> >                 # redirect_alloc_fd accepts files only. so we need to open
> >                 # a random file and then reuse the fd for logger process.
> >                 redirect_alloc_fd log_fd /dev/null
> > +
> > +               _multibuild_run() {
> > +                       # find the first non-private command
> > +                       local i=1
> > +                       while [[ ${!i} == _* ]]; do
> > +                               (( i += 1 ))
> > +                       done
> > +
> > +                       einfo "${v}: running ${@:${i}}"
> 
> So this is an einfo with an assignment side-effect? Can we perhaps
> make the assignment explicit?

Assignment side-effect? I don't understand.

Maybe I'm missing something but by design it was supposed to print ${@}
starting from ${i}-th item.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 966 bytes --]

Re: [gentoo-dev] [PATCH 8/8] fftw: example use of multibuild in ebuild.

[Christoph Junghans]

+1, feel free to commit, when multibuild.eclass was added.

2013/2/27 Michał Górny <mgorny@gentoo.org>:
> Just a quick, dirty example. Not even tested thoroughly ;).
> ---
>  gx86/sci-libs/fftw/fftw-3.3.3-r1.ebuild | 38 +++++++++++++--------------------
>  1 file changed, 15 insertions(+), 23 deletions(-)
>
> diff --git a/gx86/sci-libs/fftw/fftw-3.3.3-r1.ebuild b/gx86/sci-libs/fftw/fftw-3.3.3-r1.ebuild
> index 18554f0..ddca8e4 100644
> --- a/gx86/sci-libs/fftw/fftw-3.3.3-r1.ebuild
> +++ b/gx86/sci-libs/fftw/fftw-3.3.3-r1.ebuild
> @@ -7,7 +7,7 @@ EAPI=5
>  #AUTOTOOLS_AUTORECONF=1
>  FORTRAN_NEEDED=fortran
>
> -inherit autotools-multilib eutils flag-o-matic fortran-2 toolchain-funcs versionator
> +inherit autotools-multilib eutils flag-o-matic fortran-2 multibuild toolchain-funcs versionator
>
>  DESCRIPTION="Fast C library for the Discrete Fourier Transform"
>  HOMEPAGE="http://www.fftw.org/"
> @@ -24,6 +24,8 @@ DEPEND="${RDEPEND}
>         test? ( dev-lang/perl )"
>
>  pkg_setup() {
> +       # XXX: this looks like it should be used with BUILD_TYPE!=binary
> +
>         if use openmp; then
>                 if [[ $(tc-getCC) == *gcc ]] && ! tc-has-openmp; then
>                         ewarn "OpenMP is not available in your current selected gcc"
> @@ -32,13 +34,13 @@ pkg_setup() {
>                 FORTRAN_NEED_OPENMP=1
>         fi
>         fortran-2_pkg_setup
> -       FFTW_DIRS="single double longdouble"
> +       MULTIBUILD_VARIANTS=( single double longdouble )
>         if use quad; then
>                 if [[ $(tc-getCC) == *gcc ]] && ! version_is_at_least 4.6 $(gcc-version); then
>                         ewarn "quad precision only available for gcc >= 4.6"
>                         die "need quad precision capable gcc"
>                 fi
> -               FFTW_DIRS+=" quad"
> +               MULTIBUILD_VARIANTS+=( quad )
>         fi
>  }
>
> @@ -57,7 +59,9 @@ src_configure() {
>         # filter -Os according to docs
>         replace-flags -Os -O2
>
> -       for x in ${FFTW_DIRS}; do
> +       my_configure() {
> +               local x=${MULTIBUILD_VARIANT}
> +
>                 myeconfargs=(
>                         $(use_enable fma)
>                         $(use_enable fortran)
> @@ -93,42 +97,30 @@ src_configure() {
>                         die "${x} precision not implemented in this ebuild"
>                 fi
>
> -               einfo "Configuring for ${x} precision"
> -               BUILD_DIR="${S}-${x}" \
> -                       autotools-multilib_src_configure
> -       done
> +               autotools-multilib_src_configure
> +       }
> +
> +       multibuild_foreach my_configure
>  }
>
>  src_compile() {
> -       for x in ${FFTW_DIRS}; do
> -               einfo "Compiling for ${x} precision"
> -               BUILD_DIR="${S}-${x}" \
> -                       autotools-multilib_src_compile
> -       done
> +       multibuild_foreach autotools-multilib_src_compile
>  }
>
>  src_test () {
> -       do_smalltest() { cd "${BUILD_DIR}" && emake -C tests smallcheck; }
>         # We want this to be a reasonably quick test, but that is still hard...
>         ewarn "This test series will take 30 minutes on a modern 2.5Ghz machine"
>         # Do not increase the number of threads, it will not help your performance
>         #local testbase="perl check.pl --nthreads=1 --estimate"
>         #               ${testbase} -${p}d || die "Failure: $n"
> -       for x in ${FFTW_DIRS}; do
> -               einfo "Testing ${x} precision"
> -               BUILD_DIR="${S}-${x}" \
> -                       multilib_foreach_abi do_smalltest
> -       done
> +       multibuild_foreach autotools-multilib_src_compile -C tests smallcheck
>  }
>
>  src_install () {
>         local u x
>         DOCS=( AUTHORS ChangeLog NEWS README TODO COPYRIGHT CONVENTIONS )
>         HTML_DOCS=( doc/html/ )
> -       for x in ${FFTW_DIRS}; do
> -               BUILD_DIR="${S}-${x}" \
> -                       autotools-multilib_src_install
> -       done
> +       multibuild_foreach autotools-multilib_src_install
>
>         if use doc; then
>                 dodoc doc/*.pdf
> --
> 1.8.1.4
>
>



--
Christoph Junghans
http://dev.gentoo.org/~ottxor/