I wrote an ebuild for the ccpnmr suite. It's a software suite for macromolecular NMR spectroscopy from assignment to validation. I tested it with the supplied example data on an x86 arch, where it works. Amd64 should be tested. Reproducible: Always
Created attachment 143892 [details] ccpnmr-1.0.15.ebuild
Created attachment 144441 [details] license
Created attachment 144443 [details] ccpnmr-1.0.15.ebuild changed license corrected depencies
Created attachment 146037 [details] ccpnmr-1.0.15.ebuild Cleaned up.
> # Copyright 1999-2008 Gentoo Foundation > # Distributed under the terms of the GNU General Public License v2 > # $Header: $ > > inherit eutils toolchain-funcs python check-reqs > > SLOT="0" > LICENSE="CCPN" > KEYWORDS="~x86" > DESCRIPTION="The Collaborative Computing Project for NMR" > SRC_URI="ftp://www.bio.cam.ac.uk/pub/ccpnmr/analysis1.0.15.tar.gz > examples? ( ftp://www.bio.cam.ac.uk/pub/ccpnmr/analysisTutorialData.tar.gz )" > HOMEPAGE="http://www.ccpn.ac.uk/ccpn" > IUSE="examples" > RESTRICT="mirror" > DEPEND="${RDEPEND}" > RDEPEND=">=dev-lang/python-2.4 > virtual/glut > >=dev-lang/tk-8.3 Does python need to be built with USE=tk? > dev-tcltk/tix > x11-apps/mesa-progs" > > pkg_setup(){ > if use examples; then > ewarn "The examples are about 523MB large." > ewarn "Sure you want to have them?" > epause 5 > CHECKREQS_DISK_USR="1024" > CHECKREQS_DISK_VAR="1024" > check_reqs > fi > } > > src_unpack(){ > unpack ${A} > #epatch "${FILESDIR}/gentoo-CCPNPYTHONPATH.patch" No patch? > } > > src_compile(){ > python_version > cd ccpnmr/ccpnmr1.0/c > > echo "CC = "$(tc-getCC)>environment.txt The quote placement is weird here. echo "CC = $(tc-getCC)" > environment.txt Although why not do it in the below cat? > > #This could be outsourced to the file dir > cat >> environment.txt <<- EOF > MALLOC_FLAG = -DDO_NOT_HAVE_MALLOC > FPIC_FLAG = -fPIC > XOR_FLAG = > IGNORE_GL_FLAG = > GL_FLAG = -DUSE_GL_FALSE False? > GLUT_NEED_INIT = -DNEED_GLUT_INIT > GLUT_NOT_IN_GL = > GLUT_FLAG = $(GLUT_NEED_INIT) $(GLUT_NOT_IN_GL) > SHARED_FLAGS = -shared > MATH_LIB = -lm > X11_DIR = /usr/ > X11_LIB = -lX11 -lXext > X11_INCLUDE_FLAGS = -I$(X11_DIR)/include > X11_LIB_FLAGS = -L$(X11_DIR)/lib > TCL_DIR = /usr > TCL_LIB = -ltcl8.4 8.5 is in the tree now, does it work? See bug #173467. Same for tk. > TCL_INCLUDE_FLAGS = -I$(TCL_DIR)/include > TCL_LIB_FLAGS = -L$(TCL_DIR)/lib > TK_DIR = /usr > TK_LIB = -ltk8.4 > TK_INCLUDE_FLAGS = -I$(TK_DIR)/include > TK_LIB_FLAGS = -L$(TK_DIR)/lib > PYTHON_DIR = /usr > PYTHON_INCLUDE_FLAGS = -I$(PYTHON_DIR)/include/python${PYVER} > GL_DIR = /usr > GL_LIB = -lglut -lGLU -lGL > GL_INCLUDE_FLAGS = -I$(GL_DIR)/include > GL_LIB_FLAGS = -L$(GL_DIR)/lib > EOF > echo "CFLAGS = "$CFLAGS" $(MALLOC_FLAG) $(FPIC_FLAG) $(XOR_FLAG)">>environment.txt Why the extra quotes around "$CFLAGS", and why isn't this just part of the cat? > > > emake -j1 > emake -j1 links Parallel build is broken, I presume. Is it fixable? > } > > src_install(){ > python_version PYVER should be preserved from src_compile(). > > IN_PATH=/usr/lib/python${PYVER}/site-packages/ccpnmr Instead of lib, use get_libdir() from multilib.eclass > > einfo "Creating launch wrapper" > #doesnt work with env.d, because portage uses the PYTHONPATH variable > #and I couldn't manage renaming the variable inside ccpnmr correctly > # to CCPNPYTHONPATH. CCPNMR always breaks. > > #cat >> "${T}"/20ccpnmr << EOF > #CCPNMR_TOP_DIR=${IN_PATH} > #PYTHONPATH=${IN_PATH}/ccpnmr1.0/python > #LD_LIBRARY_PATH=/usr/lib > #TCL_LIBRARY=/usr/lib/tcl8.4 > #TK_LIBRARY=/usr/lib/tk8.4 > #EOF > > # doenvd "${T}"/20ccpnmr || die "Failed to install env.d." > > cat >> "${T}"/base << EOF > #!/bin/sh > export CCPNMR_TOP_DIR=${IN_PATH} > export PYTHONPATH=${IN_PATH}/ccpnmr1.0/python > export LD_LIBRARY_PATH=/usr/lib > export TCL_LIBRARY=/usr/lib/tcl8.4 > export TK_LIBRARY=/usr/lib/tk8.4 > EOF > > cat "${T}"/base > "${T}"/analysis > echo "${python} -O -i ${CCPNMR_TOP_DIR}/ccpnmr1.0/python/ccpnmr/analysis/AnalysisGui.py $1 $2 $3 $4 $5" >> "${T}"/analysis > > cat "${T}"/base > "${T}"/dataShifter > echo "${python} -O ${PYTHONPATH}/ccpnmr/format/gui/DataShifter.py" >> "${T}"/dataShifter > > cat "${T}"/base > "${T}"/formatConverter > echo "${python} -O ${PYTHONPATH}/ccpnmr/format/gui/FormatConverter.py $1 $2" >> "${T}"/formatConverter > > cat "${T}"/base > "${T}"/pipe2azara > echo "${python} -O ${PYTHONPATH}/ccpnmr/analysis/NmrPipeData.py $1 $2 $3" >> "${T}"/pipe2azara > > #Perhaps the two Update wrapper shouldn't be. > cat "${T}"/base > "${T}"/updateAll > echo "${python} -O ${PYTHONPATH}/ccpnmr/update/UpdateAuto.py" >> "${T}"/updateAll > > cat "${T}"/base > "${T}"/updateCheck > echo "${python} -O ${PYTHONPATH}/ccpnmr/update/UpdatePopup.py" >> "${T}"/updateCheck Hmm, lots of duplication from the base file here. Could these other ones just source it instead? > > einfo "Installing wrapper" > exeinto ${IN_PATH}/bin > # doexe "${T}/{analysis,dataShifter,formatConverter,pipe2azara,updateAll,updateCheck}" > doexe "${T}"/analysis || die "Failed to install wrapper." > doexe "${T}"/dataShifter || die "Failed to install wrapper." > doexe "${T}"/formatConverter || die "Failed to install wrapper." > doexe "${T}"/pipe2azara || die "Failed to install wrapper." > doexe "${T}"/updateAll || die "Failed to install wrapper." > doexe "${T}"/updateCheck || die "Failed to install wrapper." > > for i in analysis dataShifter formatConverter pipe2azara > do > dosym ${IN_PATH}/bin/$i /usr/bin/$i > done > > insinto ${IN_PATH} > cd ccpnmr > > einfo "Installing main files" > insopts -v > doins -r * > > > einfo "Adjusting permissions" > #I do not know wether this is a must or not, but thats how the original install looks like Take a look at similar files from other packages. What permissions do shared libraries have? How about Python files that aren't directly executed? How about images? > local FILES="c/ccp/structure/StructUtil.so > c/ccp/structure/StructAtom.so > c/ccp/structure/StructBond.so > c/ccp/structure/StructStructure.so > c/ccpnmr/clouds/Midge.so > c/ccpnmr/clouds/Dynamics.so > c/ccpnmr/clouds/Bacus.so > c/ccpnmr/clouds/AtomCoord.so > c/ccpnmr/clouds/DistConstraintList.so > c/ccpnmr/clouds/DistConstraint.so > c/ccpnmr/clouds/CloudUtil.so > c/ccpnmr/clouds/DistForce.so > c/ccpnmr/clouds/AtomCoordList.so > c/ccpnmr/analysis/ContourFile.so > c/ccpnmr/analysis/ContourLevels.so > c/ccpnmr/analysis/SliceFile.so > c/ccpnmr/analysis/PeakList.so > c/ccpnmr/analysis/ContourStyle.so > c/ccpnmr/analysis/WinPeakList.so > c/memops/global/PdfHandler.so > c/memops/global/TkHandler.so > c/memops/global/FitMethod.so > c/memops/global/BlockFile.so > c/memops/global/MemCache.so > c/memops/global/StoreFile.so > c/memops/global/PsHandler.so > c/memops/global/StoreHandler.so > c/memops/global/GlHandler.so > doc/graphics/prev.gif > doc/graphics/up.gif > doc/graphics/next.gif > python/ccp/c/linkSharedObjs > python/ccpnmr/c/linkSharedObjs > python/ccpnmr/clouds/CloudsPopup.py > python/ccpnmr/clouds/HcloudsMdPopup.py > python/ccpnmr/clouds/PseudoResonances.py > python/ccpnmr/clouds/FilterCloudsPopup.py > python/ccpnmr/clouds/HydrogenDynamics.py > python/ccpnmr/clouds/FilterClouds.py > python/ccpnmr/clouds/Clouds.py > python/ccpnmr/clouds/MidgePopup.py > python/ccpnmr/clouds/NoeRelaxation.py > python/ccpnmr/clouds/FileIO.py > python/ccpnmr/clouds/BacusPopup.py > python/ccpnmr/clouds/ResonanceIdentification.py > python/ccpnmr/clouds/__init__.py > python/ccpnmr/clouds/_licenseInfo.py > python/memops/c/linkSharedObjs" > > for FILE in ${FILES} > do > fperms 755 ${IN_PATH}/ccpnmr1.0/${FILE} > done > > # I do not know how to make this more simple, because the relative paths have to be kept. > dodir /usr/share/doc/${PF}/html/ > cp -r --parents `find . -name doc` "${D}"usr/share/doc/${PF}/html/ Check out portability.eclass: treecopy() -- can that help? > rm -r `find "${D}"usr/lib/ -name doc` get_libdir() again > > if use examples; then > cd "${WORKDIR}" > einfo "Installing example files" > insopts -v > insinto /usr/share/${PF}/ > doins -r analysisTutorialData > fi > } > > pkg_postrm() { > python_version Shouldn't need to run this again, it should be saved in the environment.tbz2 file from when it was built (if it's a binpkg) or saved from previous phases (if source build/install on the same system). > python_mod_cleanup "${ROOT}"/usr/lib/python${PYVER}/site-packages/ccpnmr >}
(In reply to comment #5) > Does python need to be built with USE=tk? Yes indeed, corrected this. > > #epatch "${FILESDIR}/gentoo-CCPNPYTHONPATH.patch" > > No patch? I tryed to make a patch, to avoid resetting the PYTHONPATH variable, but I didn't get this to work, so I dropped this. > > echo "CC = "$(tc-getCC)>environment.txt > > The quote placement is weird here. > > echo "CC = $(tc-getCC)" > environment.txt > > Although why not do it in the below cat? At the time I wrote this I wasn't that confirm with several things. Your solution is the better one. > > GL_FLAG = -DUSE_GL_FALSE > > False? I don't really know: From the enviroment-default: [Quote] # special GL flag, should have either USE_GL_TRUE or USE_GL_FALSE # (-D means this gets defined by the compiler so can be checked in source code) # (this relates to glXCreateContext() call in ccpnmr/global/gl_handler.c) # if have problems with USE_GL_TRUE then try GL_FALSE, or vice versa GL_FLAG = -DUSE_GL_FALSE # use below for Linux? #GL_FLAG = -DUSE_GL_FALSE # use below for everything else? #GL_FLAG = -DUSE_GL_TRUE [/Quote] I tryed both and on my system both settings are working. Do you understand the -D thing? Another thing is the setting GLUT_NEED_INIT = -DNEED_GLUT_INIT. Here the installer says [Quote] Does your glut need explicit initialisation (OSX and freeglut probably y, ordinary glut probably n) (y or n)? [/Quote] What do you think about this? We can let this setting depend on the version of glut which is used. I played around with combinations of the glut version and the init type but can not found any difference. > > TCL_LIB = -ltcl8.4 > > 8.5 is in the tree now, does it work? See bug #173467. Same for tk. > It seems working. First testing doesn't show an abnormality. > > Why the extra quotes around "$CFLAGS", and why isn't this just part of the cat? Don't know. Gone! > > emake -j1 > > emake -j1 links > > Parallel build is broken, I presume. Is it fixable? Once I wrote the ebuild the parallel build seems broken, I tried it again and everything is fine. I removed the -j1 until there is problem again. But nevertheless, how can I find out how to fix the make if parallel build is broken? > > cat "${T}"/base > "${T}"/updateCheck > > echo "${python} -O ${PYTHONPATH}/ccpnmr/update/UpdatePopup.py" >> "${T}"/updateCheck > > Hmm, lots of duplication from the base file here. Could these other ones just > source it instead? Sure, I just did a replay of the original install process. > Take a look at similar files from other packages. What permissions do shared > libraries have? How about Python files that aren't directly executed? How about > images? Once again, just the replay of the original installation process. Corrected, but which perms would you expact for shared libraries? Most are 775. > > Check out portability.eclass: treecopy() -- can that help? Yeah, but there must be a better documentation of the eclasses. For me it is impossible to find those functions. And I think this fits to everyone who is new in ebuild writing.
Created attachment 147003 [details, diff] ccpnmr-1.0.15.ebuild diff
Created attachment 152427 [details] ccpnmr-1.0.15-r2.ebuild I rewrote some Stuff. It is more straight now and still working.
Created attachment 152429 [details] ccpnmr.diff Diff between last and current version
Created attachment 152477 [details] ccpnmr.diff
