# Copyright 1999-2010 Gentoo Foundation # Distributed under the terms of the GNU General Public License v2 # $Header: /var/cvsroot/gentoo-x86/dev-lang/ghc/ghc-6.10.4-r1.ebuild,v 1.11 2010/09/19 16:45:24 slyfox Exp $ # Brief explanation of the bootstrap logic: # # Previous ghc ebuilds have been split into two: ghc and ghc-bin, # where ghc-bin was primarily used for bootstrapping purposes. # From now on, these two ebuilds have been combined, with the # binary USE flag used to determine whether or not the pre-built # binary package should be emerged or whether ghc should be compiled # from source. If the latter, then the relevant ghc-bin for the # arch in question will be used in the working directory to compile # ghc from source. # # This solution has the advantage of allowing us to retain the one # ebuild for both packages, and thus phase out virtual/ghc. # Note to users of hardened gcc-3.x: # # If you emerge ghc with hardened gcc it should work fine (because we # turn off the hardened features that would otherwise break ghc). # However, emerging ghc while using a vanilla gcc and then switching to # hardened gcc (using gcc-config) will leave you with a broken ghc. To # fix it you would need to either switch back to vanilla gcc or re-emerge # ghc (or ghc-bin). Note that also if you are using hardened gcc-3.x and # you switch to gcc-4.x that this will also break ghc and you'll need to # re-emerge ghc (or ghc-bin). People using vanilla gcc can switch between # gcc-3.x and 4.x with no problems. inherit base autotools bash-completion eutils flag-o-matic multilib toolchain-funcs ghc-package versionator pax-utils DESCRIPTION="The Glasgow Haskell Compiler" HOMEPAGE="http://www.haskell.org/ghc/" # discover if this is a snapshot release IS_SNAPSHOT="$(get_version_component_range 4)" # non-empty if snapshot EXTRA_SRC_URI="${PV}" [[ "${IS_SNAPSHOT}" ]] && EXTRA_SRC_URI="stable/dist" arch_binaries="" arch_binaries="$arch_binaries alpha? ( http://code.haskell.org/~ivanm/ghc-bin-${PV}-alpha.tbz2 )" arch_binaries="$arch_binaries x86? ( http://code.haskell.org/~ivanm/ghc-bin-${PV}-x86.tbz2 )" arch_binaries="$arch_binaries amd64? ( http://haskell.org/~kolmodin/ghc-bin-${PV}-amd64.tbz2 )" arch_binaries="$arch_binaries ia64? ( http://code.haskell.org/~slyfox/ghc-ia64/ghc-bin-${PV}-ia64.tbz2 )" arch_binaries="$arch_binaries sparc? ( http://haskell.org/~duncan/ghc/ghc-bin-${PV}-sparc.tbz2 )" arch_binaries="$arch_binaries ppc64? ( http://code.haskell.org/~slyfox/ghc-ppc64/ghc-bin-${PV}-ppc64.tbz2 )" arch_binaries="$arch_binaries ppc? ( mirror://gentoo/ghc-bin-${PV}-ppc.tbz2 )" SRC_URI="!binary? ( http://haskell.org/ghc/dist/${EXTRA_SRC_URI}/${P}-src.tar.bz2 ) !ghcbootstrap? ( $arch_binaries )" LICENSE="BSD" SLOT="0" KEYWORDS="~alpha ~amd64 ~ia64 ~ppc ~ppc64 ~sparc ~x86" IUSE="binary doc ghcbootstrap" RDEPEND=" !dev-lang/ghc-bin >=sys-devel/gcc-2.95.3 >=sys-devel/binutils-2.17 >=dev-lang/perl-5.6.1 >=dev-libs/gmp-4.1 !=ghc-5.04.3 on their $PATH already PDEPEND="!ghcbootstrap? ( =app-admin/haskell-updater-0.9* )" append-ghc-cflags() { local flag compile assemble link for flag in $*; do case ${flag} in compile) compile="yes";; assemble) assemble="yes";; link) link="yes";; *) [[ ${compile} ]] && GHC_CFLAGS="${GHC_CFLAGS} -optc${flag}" [[ ${assemble} ]] && GHC_CFLAGS="${GHC_CFLAGS} -opta${flag}" [[ ${link} ]] && GHC_CFLAGS="${GHC_CFLAGS} -optl${flag}";; esac done } ghc_setup_cflags() { # We need to be very careful with the CFLAGS we ask ghc to pass through to # gcc. There are plenty of flags which will make gcc produce output that # breaks ghc in various ways. The main ones we want to pass through are # -mcpu / -march flags. These are important for arches like alpha & sparc. # We also use these CFLAGS for building the C parts of ghc, ie the rts. strip-flags strip-unsupported-flags GHC_CFLAGS="" for flag in ${CFLAGS}; do case ${flag} in # Ignore extra optimisation (ghc passes -O to gcc anyway) # -O2 and above break on too many systems -O*) ;; # Arch and ABI flags are what we're really after -m*) append-ghc-cflags compile assemble ${flag};; # Debugging flags don't help either. You can't debug Haskell code # at the C source level and the mangler discards the debug info. -g*) ;; # Ignore all other flags, including all -f* flags esac done # hardened-gcc needs to be disabled, because the mangler doesn't accept # its output. gcc-specs-pie && append-ghc-cflags compile link -nopie gcc-specs-ssp && append-ghc-cflags compile -fno-stack-protector # prevent from failind building unregisterised ghc: # http://www.mail-archive.com/debian-bugs-dist@lists.debian.org/msg171602.html use ppc64 && append-ghc-cflags compile -mminimal-toc # We also add -Wa,--noexecstack to get ghc to generate .o files with # non-exectable stack. This it a hack until ghc does it itself properly. append-ghc-cflags assemble "-Wa,--noexecstack" } pkg_setup() { if use ghcbootstrap; then ewarn "You requested ghc bootstrapping, this is usually only used" ewarn "by Gentoo developers to make binary .tbz2 packages for" ewarn "use with the ghc ebuild's USE=\"binary\" feature." use binary && \ die "USE=\"ghcbootstrap binary\" is not a valid combination." [[ -z $(type -P ghc) ]] && \ die "Could not find a ghc to bootstrap with." fi } src_unpack() { # Create the ${S} dir if we're using the binary version use binary && mkdir "${S}" base_src_unpack source "${FILESDIR}/ghc-apply-gmp-hack" "$(get_libdir)" ghc_setup_cflags if ! use ghcbootstrap; then # Modify the wrapper script from the binary tarball to use GHC_CFLAGS. # See bug #313635. sed -i -e "s|wrapped|wrapped ${GHC_CFLAGS}|" \ "${WORKDIR}/usr/bin/ghc-${PV}" # allow hardened users use vanilla biary to bootstrap ghc # ghci uses mmap with rwx protection at it implements dynamic # linking on it's own (bug #299709) pax-mark -m "${WORKDIR}/usr/$(get_libdir)/${P}/ghc" fi if use binary; then # Move unpacked files to the expected place mv "${WORKDIR}/usr" "${S}" else if ! use ghcbootstrap; then # Relocate from /usr to ${WORKDIR}/usr sed -i -e "s|/usr|${WORKDIR}/usr|g" \ "${WORKDIR}/usr/bin/ghc-${PV}" \ "${WORKDIR}/usr/bin/ghci-${PV}" \ "${WORKDIR}/usr/bin/ghc-pkg-${PV}" \ "${WORKDIR}/usr/bin/hsc2hs" \ "${WORKDIR}/usr/$(get_libdir)/${P}/package.conf" \ || die "Relocating ghc from /usr to workdir failed" fi # Hack to prevent haddock being installed, remove when ./configure # supports something better to not build docs or haddock. sed -i -e 's/DO_NOT_INSTALL =/DO_NOT_INSTALL = haddock/' \ "${S}/utils/Makefile" # Modify the ghc driver script to use GHC_CFLAGS sed -i -e "s|wrapped|wrapped ${GHC_CFLAGS}|" \ "${S}/ghc/ghc.wrapper" cd "${S}" # disable cabal built binary stripping (solves bug #299492) # as installPackage strips them by default epatch "${FILESDIR}/ghc-6.10.4-disable-strip.patch" # patch aclocal.m4 and configure.ac to work with >=autoconf-2.64 epatch "${FILESDIR}/${P}-autoconf.patch" # fix configure.ac to detect libm need # http://bugs.gentoo.org/show_bug.cgi?id=293208 # http://hackage.haskell.org/trac/ghc/ticket/3730 epatch "${FILESDIR}/ghc-6.10.4-libm-detection.patch" # -r and --relax are incompatible epatch "${FILESDIR}/ghc-6.10.4-ia64-fixed-relax.patch" # prevent from wiping upper address bits used in cache lookup epatch "${FILESDIR}/ghc-6.10.4-ia64-storage-manager-fix.patch" # as we have changed the build system eautoreconf fi } src_compile() { if ! use binary; then # initialize build.mk echo '# Gentoo changes' > mk/build.mk # Put docs into the right place, ie /usr/share/doc/ghc-${PV} echo "docdir = /usr/share/doc/${P}" >> mk/build.mk echo "htmldir = /usr/share/doc/${P}" >> mk/build.mk # We also need to use the GHC_CFLAGS flags when building ghc itself echo "SRC_HC_OPTS+=${GHC_CFLAGS}" >> mk/build.mk echo "SRC_CC_OPTS+=${CFLAGS} -Wa,--noexecstack" >> mk/build.mk # We can't depend on haddock except when bootstrapping when we # must build docs and include them into the binary .tbz2 package if use ghcbootstrap && use doc; then echo XMLDocWays="html" >> mk/build.mk echo HADDOCK_DOCS=YES >> mk/build.mk else echo XMLDocWays="" >> mk/build.mk echo HADDOCK_DOCS=NO >> mk/build.mk fi # circumvent a very strange bug that seems related with ghc producing # too much output while being filtered through tee (e.g. due to # portage logging) reported as bug #111183 echo "SRC_HC_OPTS+=-w" >> mk/build.mk # some arches do not support ELF parsing for ghci module loading # PPC64: never worked (should be easy to implement) # alpha: never worked if use alpha || use ppc64; then echo "GhcWithInterpreter=NO" >> mk/build.mk fi # we have to tell it to build unregisterised on some arches # ppc64: EvilMangler currently does not understand some TOCs # ia64: EvilMangler bitrot if use alpha || use ia64 || use ppc64; then echo "GhcUnregisterised=YES" >> mk/build.mk echo "GhcWithNativeCodeGen=NO" >> mk/build.mk echo "SplitObjs=NO" >> mk/build.mk echo "GhcRTSWays := debug" >> mk/build.mk echo "GhcNotThreaded=YES" >> mk/build.mk fi # Have "ld -r --relax" problem with split-objs on sparc: if use sparc; then echo "SplitObjs=NO" >> mk/build.mk fi # Get ghc from the unpacked binary .tbz2 # except when bootstrapping we just pick ghc up off the path if ! use ghcbootstrap; then export PATH="${WORKDIR}/usr/bin:${PATH}" fi econf || die "econf failed" # LC_ALL needs to workaround ghc's ParseCmm failure on some (es) locales # bug #202212 / http://hackage.haskell.org/trac/ghc/ticket/4207 LC_ALL=C emake all || die "make failed" fi # ! use binary } src_install() { if use binary; then mv "${S}/usr" "${D}" # Remove the docs if not requested if ! use doc; then rm -rf "${D}/usr/share/doc/${P}/*/" \ "${D}/usr/share/doc/${P}/*.html" \ || die "could not remove docs (P vs PF revision mismatch?)" fi else local insttarget="install" # We only built docs if we were bootstrapping, otherwise # we copy them out of the unpacked binary .tbz2 if use doc; then if use ghcbootstrap; then insttarget="${insttarget} install-docs" else mkdir -p "${D}/usr/share/doc" mv "${WORKDIR}/usr/share/doc/${P}" "${D}/usr/share/doc" \ || die "failed to copy docs" fi fi emake -j1 ${insttarget} \ DESTDIR="${D}" \ || die "make ${insttarget} failed" # ghci uses mmap with rwx protection at it implements dynamic # linking on it's own (bug #299709) # so mark resulting binary pax-mark -m "${D}/usr/$(get_libdir)/${P}/ghc" dodoc "${S}/README" "${S}/ANNOUNCE" "${S}/LICENSE" "${S}/VERSION" dobashcompletion "${FILESDIR}/ghc-bash-completion" cp -p "${D}/usr/$(get_libdir)/${P}/package.conf"{,.shipped} \ || die "failed to copy package.conf" fi } pkg_postinst() { # 'ghc-pkg check' fails in ghc 6.10.2, with the error message: # There are problems in package rts-1.0: # include-dirs: PAPI_INCLUDE_DIR doesn't exist or isn't a directory # Upstream suggests this solution to fix it: export PATH="/usr/bin:${PATH}" $(ghc-getghcpkg) describe rts | sed 's/PAPI_INCLUDE_DIR//' | $(ghc-getghcpkg) update - ghc-reregister ewarn "IMPORTANT:" ewarn "If you have upgraded from another version of ghc," ewarn "once app-admin/haskell-updater has installed please run:" ewarn " /usr/sbin/haskell-updater --upgrade" ewarn "to re-build all ghc-based Haskell libraries." bash-completion_pkg_postinst } pkg_prerm() { # Be very careful here... Call order when upgrading is (according to PMS): # * src_install for new package # * pkg_preinst for new package # * pkg_postinst for new package # * pkg_prerm for the package being replaced # * pkg_postrm for the package being replaced # so you'll actually be touching the new packages files, not the one you # uninstall, due to that or installation directory ${P} will be the same for # both packages. # Call order for reinstalling is (according to PMS): # * src_install # * pkg_preinst # * pkg_prerm for the package being replaced # * pkg_postrm for the package being replaced # * pkg_postinst # Overwrite the (potentially) modified package.conf with a copy of the # original one, so that it will be removed during uninstall. PKG="${ROOT}/usr/$(get_libdir)/${P}/package.conf" cp -p "${PKG}"{.shipped,} [[ -f ${PKG}.old ]] && rm "${PKG}.old" }