guix-patches
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[bug#28045] [PATCH] gnu: Add openfoam


From: Marius Bakke
Subject: [bug#28045] [PATCH] gnu: Add openfoam
Date: Tue, 15 Aug 2017 01:07:28 +0200
User-agent: Notmuch/0.25 (https://notmuchmail.org) Emacs/25.2.1 (x86_64-unknown-linux-gnu)

Paul Garlick <address@hidden> writes:

> * gnu/packages/simulation.scm: New file
> * gnu/packages/patches/openfoam-cleanup.patch: New file
> * gnu/local.mk(GNU_SYSTEM_MODULES): Add module
>               (dist_patch_DATA): Add patch

Thanks a lot for this!  The patch does not apply for me, can you rebase
this on current 'master'?  Meanwhile, some comments...

> +diff -ur OpenFOAM-4.x-version-4.1.org/wmake/src/Makefile 
> OpenFOAM-4.x-version-4.1/wmake/src/Makefile
> +--- OpenFOAM-4.x-version-4.1.org/wmake/src/Makefile
> ++++ OpenFOAM-4.x-version-4.1/wmake/src/Makefile
> +@@ -33,7 +33,7 @@
> + # The Makefile use a POSIX shell
> + 
> #------------------------------------------------------------------------------
> + 
> +-SHELL      = /bin/sh
> ++# SHELL      = /bin/sh

This particular hunk is not necessary since gnu-build-system will search
for Makefiles and adjust SHELL as needed.

> +diff -ur OpenFOAM-4.x-version-4.1.org/wmake/makefiles/general 
> OpenFOAM-4.x-version-4.1/wmake/makefiles/general
> +--- OpenFOAM-4.x-version-4.1.org/wmake/makefiles/general
> ++++ OpenFOAM-4.x-version-4.1/wmake/makefiles/general
> +@@ -33,7 +33,7 @@
> + # The Makefile uses a POSIX shell
> + 
> #------------------------------------------------------------------------------
> + 
> +-SHELL           = /bin/sh
> ++# SHELL           = /bin/sh

...although this file named "general" would not be patched automatically.

> + 
> #------------------------------------------------------------------------------
> +diff -ur OpenFOAM-4.x-version-4.1.org/wmake/wmake 
> OpenFOAM-4.x-version-4.1/wmake/wmake
> +--- OpenFOAM-4.x-version-4.1.org/wmake/wmake
> ++++ OpenFOAM-4.x-version-4.1/wmake/wmake
> +@@ -163,7 +163,7 @@
> + then
> +     if [ "$WM_NCOMPPROCS" -gt 1 -a ! "$MAKEFLAGS" ]
> +     then
> +-        lockDir=$HOME/.$WM_PROJECT/.wmake
> ++        lockDir=$(cd $(dirname $BASH_SOURCE)/../.. && pwd 
> -P)/.$WM_PROJECT/.wmake
> + 
> +         if [ -d $lockDir ]
> +         then
> +diff -ur OpenFOAM-4.x-version-4.1.org/wmake/wmakeScheduler 
> OpenFOAM-4.x-version-4.1/wmake/wmakeScheduler
> +--- OpenFOAM-4.x-version-4.1.org/wmake/wmakeScheduler
> ++++ OpenFOAM-4.x-version-4.1/wmake/wmakeScheduler
> +@@ -53,7 +53,7 @@
> + # csh sets HOST, bash sets HOSTNAME
> + : ${HOST:=$HOSTNAME}
> + 
> +-lockDir=$HOME/.$WM_PROJECT/.wmake
> ++lockDir=$(cd $(dirname $BASH_SOURCE)/../.. && pwd -P)/.$WM_PROJECT/.wmake
> + 
> + # Fallback - 1 core on current host
> + : ${WM_HOSTS:=$HOST:1}
> +diff -ur OpenFOAM-4.x-version-4.1.org/wmake/wmakeSchedulerUptime 
> OpenFOAM-4.x-version-4.1/wmake/wmakeSchedulerUptime
> +--- OpenFOAM-4.x-version-4.1.org/wmake/wmakeSchedulerUptime
> ++++ OpenFOAM-4.x-version-4.1/wmake/wmakeSchedulerUptime
> +@@ -53,7 +53,7 @@
> + # csh sets HOST, bash sets HOSTNAME
> + : ${HOST:=$HOSTNAME}
> + 
> +-lockDir=$HOME/.$WM_PROJECT/.wmake
> ++lockDir=$(cd $(dirname $BASH_SOURCE)/../.. && pwd -P)/.$WM_PROJECT/.wmake
> + # Fallback - 1 core on current host
> + : ${WM_HOSTS:=$HOST:1}

Maybe these could be replaced with something like...

(substitute* (find-files "." "^\\.wmake$")
  (("\\$HOME) (getenv "PWD")))

$PWD is initially the source directory, which I think is what the
$(dirname $BASH_SOURCE ...) is aiming for(?).

> +diff -ur 
> OpenFOAM-4.x-version-4.1.org/src/parallel/decompose/metisDecomp/metisDecomp.C 
> OpenFOAM-4.x-version-4.1/src/parallel/decompose/metisDecomp/metisDecomp.C
> +--- 
> OpenFOAM-4.x-version-4.1.org/src/parallel/decompose/metisDecomp/metisDecomp.C
> ++++ OpenFOAM-4.x-version-4.1/src/parallel/decompose/metisDecomp/metisDecomp.C
> +@@ -67,7 +67,7 @@
> + 
> +     // Processor weights initialised with no size, only used if specified in
> +     // a file
> +-    Field<scalar> processorWeights;
> ++    Field<floatScalar> processorWeights;

This seems unrelated to unbundling :) Is this a known issue?

> diff --git a/gnu/packages/simulation.scm b/gnu/packages/simulation.scm
> new file mode 100644
> index 000000000..5aa5f8f03
> --- /dev/null
> +++ b/gnu/packages/simulation.scm
> @@ -0,0 +1,197 @@
> +;;; GNU Guix --- Functional package management for GNU
> +;;; Copyright © 2017 Paul Garlick <address@hidden>
> +;;;
> +;;; This file is part of GNU Guix.
> +;;;
> +;;; GNU Guix is free software; you can redistribute it and/or modify it
> +;;; under the terms of the GNU General Public License as published by
> +;;; the Free Software Foundation; either version 3 of the License, or (at
> +;;; your option) any later version.
> +;;;
> +;;; GNU Guix is distributed in the hope that it will be useful, but
> +;;; WITHOUT ANY WARRANTY; without even the implied warranty of
> +;;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> +;;; GNU General Public License for more details.
> +;;;
> +;;; You should have received a copy of the GNU General Public License
> +;;; along with GNU Guix.  If not, see <http://www.gnu.org/licenses/>.
> +
> +(define-module (gnu packages simulation)
> +  #:use-module (gnu packages)
> +  #:use-module (gnu packages base)
> +  #:use-module (gnu packages bash)
> +  #:use-module (gnu packages bison)
> +  #:use-module (gnu packages boost)
> +  #:use-module (gnu packages compression)
> +  #:use-module (gnu packages flex)
> +  #:use-module (gnu packages gettext)
> +  #:use-module (gnu packages gcc)
> +  #:use-module (gnu packages gl)
> +  #:use-module (gnu packages graphics)
> +  #:use-module (gnu packages gtk)
> +  #:use-module (gnu packages linux)
> +  #:use-module (gnu packages m4)
> +  #:use-module (gnu packages maths)
> +  #:use-module (gnu packages mpi)
> +  #:use-module (gnu packages multiprecision)
> +  #:use-module (gnu packages ncurses)
> +  #:use-module (gnu packages readline)
> +  #:use-module (gnu packages tls)
> +  #:use-module (gnu packages version-control)
> +  #:use-module (gnu packages xml)
> +  #:use-module (gnu packages xorg)
> +  #:use-module (guix download)
> +  #:use-module (guix build utils)
> +  #:use-module (guix build-system gnu)
> +  #:use-module ((guix licenses) #:prefix license:)
> +  #:use-module (guix packages)
> +  #:use-module (guix utils)
> +  #:use-module (ice-9 ftw)
> +  #:use-module (ice-9 regex)
> +  #:use-module (srfi srfi-1))
> +
> +(define-public openfoam
> +  (package
> +    (name "openfoam")
> +    (version "4.1")
> +    (source
> +     (origin
> +      (method url-fetch)
> +      (uri (string-append
> +             "http://dl.openfoam.org/source/";
> +             (string-replace-substring version "." "-")))
> +      (file-name (string-append name "-" version ".tar.gz"))
> +      (sha256
> +       (base32 "1hgh2kw7fqv4bfachhbjhw16v8j35jbv6s576fmz4ka2q4ks4wg3"))
> +      (patches (search-patches "openfoam-cleanup.patch"))
> +      (modules '((guix build utils)))))
> +    (build-system gnu-build-system)
> +    (native-inputs
> +     `(("bison" ,bison)))
> +    (propagated-inputs
> +     `(("boost" ,boost)
> +       ("cgal" ,cgal)
> +       ("flex" ,flex)
> +       ("git" ,git)
> +       ("gmp" ,gmp)
> +       ("gzip" ,gzip)
> +       ("gnuplot" ,gnuplot)
> +       ("libxt" ,libxt)
> +       ("metis" ,metis)
> +       ("mpfr" ,mpfr)
> +       ("ncurses" ,ncurses)
> +       ("openmpi" ,openmpi)
> +       ("readline" ,readline)
> +       ("scotch" ,pt-scotch)
> +       ("zlib" ,zlib)))

We try to avoid propagating inputs to minimize risk of profile
conflicts.  Typically, the linker will add things to RUNPATH as
necessary and then we only have to fully qualify command-line references
and e.g. dlopen()'d solibs.  Can you try making these normal inputs?

If the dependencies show up in "guix gc -R $(./pre-inst-env guix build
openfoam)" when added as regular inputs, propagation is usually not
necessary.

> +    (arguments
> +     `(;; executable files and shared libraries are located in the 
> 'platforms'
> +       ;; subdirectory
> +       #:strip-directories (list (string-append
> +                                  "OpenFOAM-" ,version
> +                                  "/platforms/linux64GccDPInt32Opt/bin")
> +                                 (string-append
> +                                  "OpenFOAM-" ,version
> +                                  "/platforms/linux64GccDPInt32Opt/lib"))
> +       ;; '#:elf-directories' is not recognised here so skip validation phase
> +       #:validate-runpath? #f

I don't suppose it's easy to move everything up to the top store level,
which is already in a versioned directory and would allow these phases
to proceed normally?

> +       #:phases (modify-phases %standard-phases
> +        (add-after 'unpack 'rename-build-directory
> +          (lambda _
> +            (use-modules (ice-9 ftw))
> +            (use-modules (ice-9 regex))
> +            (chdir "..")
> +            ;; use 'OpenFOAM-version' convention to match path name
> +            ;; expectations in the build phase
> +            (let ((unpack-dir (string-append
> +                               (getcwd) "/"
> +                               (list-ref (scandir (getcwd) (lambda (name)
> +                                 (string-match "^OpenFOAM" name))) 0)))
> +                  (build-dir (string-append
> +                              (getcwd) "/OpenFOAM-" ,version)))
> +             ;; re-name build directory
> +             (rename-file unpack-dir build-dir)
> +             ;; move to build directory
> +             (chdir (basename build-dir)))))

Nit-pick: (chdir ...) (and setenv below) have unspecified return values,
so please add an explicit #t at the end of such phases for determinism.
Consult the Guile manual when unsure if a return value can be "trusted".

> +        (add-after 'patch-source-shebangs 'set-environment-variables
> +          (lambda _
> +            (let ((libraries '("boost" "cgal" "gmp" "metis" "mpfr" 
> "scotch")))
> +             ;; set variables to define store paths
> +             (for-each (lambda (library)
> +                        (setenv (string-append
> +                                 (string-upcase library) "_ROOT")
> +                                (assoc-ref %build-inputs library))) 
> libraries))
> +            ;; set variables to define package versions
> +            (setenv "SCOTCHVERSION" ,(package-version scotch))
> +            (setenv "METISVERSION" ,(package-version metis))))
> +        (delete 'configure) ; no configure phase
> +        (delete 'patch-generated-file-shebangs) ; no generated files

Does this phase fail somehow, or is it deleted simply because it's not
needed?  Just curious :-)

> +        (replace 'build
> +          (lambda _
> +            ;; compile OpenFOAM libraries and applications
> +            (zero? (system (format #f
> +                             "source ./etc/bashrc && ./Allwmake -j~a"
> +                                (parallel-job-count))))))
> +        (add-after 'build 'update-configuration-files
> +          (lambda _ 
> +            ;; record store paths and package versions in configuration files
> +            (zero? (system (format #f "sed -i 's|~a|'~:*~s'|' ~a"
> +                       "$CGAL_ROOT" "etc/config.sh/CGAL")))
> +            (zero? (system (format #f "sed -i 's|~a|'~:*~s'|' ~a"
> +                       "$BOOST_ROOT" "etc/config.sh/CGAL")))
> +            (zero? (system (format #f "sed -i 's|~a|'~:*~s'|' ~a"
> +                       "$METIS_ROOT" "etc/config.sh/metis")))
> +            (zero? (system (format #f "sed -i 's|~a|'~:*~s'|' ~a"
> +                       "$METISVERSION" "etc/config.sh/metis")))
> +            (zero? (system (format #f "sed -i 's|~a|'~:*~s'|' ~a"
> +                       "$SCOTCH_ROOT" "etc/config.sh/scotch")))
> +            (zero? (system (format #f "sed -i 's|~a|'~:*~s'|' ~a"
> +                       "$SCOTCHVERSION" "etc/config.sh/scotch")))
> +            (zero? (system (format #f "sed -i 's|~a|'~:*~s'|' ~a"
> +                       "$GMP_ROOT" "etc/config.sh/settings")))
> +            (zero? (system (format #f "sed -i 's|~a|'~:*~s'|' ~a"
> +                       "$MPFR_ROOT" "etc/config.sh/settings")))))

Can you try to use (substitute* ...) here instead?  I failed to parse
the four single quotes in the sed expression in the format string :-/

> +        (delete 'check) ; no tests to run

You can set #:tests? #f instead of deleting this phase.

> +        (replace 'install
> +          (lambda _
> +            ;; use 'OpenFOAM-version' convention
> +            (let ((install-dir (string-append
> +                                %output "/OpenFOAM-" ,version)))
> +              ;; create install directory
> +              (mkdir-p install-dir)
> +              ;; move contents of build directory to install directory
> +              (copy-recursively "." install-dir)))))))

Related to the earlier question, but it would be nice if we could just
copy the contents of platforms/linux64GccDPInt32Opt to %output directly.

> +        ;; Note: tutorial files are installed read-only in /gnu/store.
> +        ;; To allow write permissions on files copied from the store a
> +        ;; 'chmod' step is needed before running the applications.  For
> +        ;; example, from a user's login:
> +        ;; $ source $GUIX_PROFILE/OpenFOAM-4.1/etc/bashrc
> +        ;; $ mkdir -p $FOAM_RUN
> +        ;; $ cd $FOAM_RUN
> +        ;; $ cp -r $FOAM_TUTORIALS/incompressible/simpleFoam/pitzdaily .
> +        ;; $ cd pitzdaily
> +        ;; $ chmod -R u+w .
> +        ;; $ blockMesh
> +    (native-search-paths
> +     ;; define the FOAM_INST_DIR variable
> +     (list (search-path-specification
> +            (variable "FOAM_INST_DIR")
> +            (files '(".")))))

...maybe by utilizing this variable?  It would also be better to wrap
the binaries with it, instead of depending on it being in a profile.

> +    (synopsis "Framework for numerical simulation of fluid flow")
> +    (description "OpenFOAM provides a set of solvers and methods for tackling
> +problems in the field of Computational Fluid Dynamics (CFD).  It is written 
> in
> +C++.  Complex governing equations such as the Navier-Stokes equations can be
> +solved in integral form.  Physical processes such as phase change, droplet
> +transport and chemical reaction can be modelled.  Numerical methods are
> +included to deal with sharp gradients, such as those encountered in flows 
> with
> +shock waves and flows with gas/liquid interfaces.
> +
> +OpenFOAM implements the Finite Volume Method for discretizing the spacial
> +dimensions.  Computational grids defining geometries in two- and 
> +three-dimensional domains may be generated by the included blockMesh and
> +snappyHexMesh routines or imported from external grid-generation packages.
> +Grids may be unstructured in layout and may contain any combination of
> +polyhedral cells.  Large problems may be split into smaller, connected
> +problems for efficient solution on parallel systems")
Missing ".'  --->                                    ^^^

Having packaged a fair share of academic software myself, I fully expect
that some of the suggestions above are overly optimistic, but had to ask
regardless ;-) Hopefully it didn't sound too demanding.

Can you send an updated patch please? :-)

Attachment: signature.asc
Description: PGP signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]