Mercurial > hg > orthanc
view OrthancFramework/Resources/CMake/BoostConfiguration.sh @ 4666:3294385a245a
typo
author | Alain Mazy <am@osimis.io> |
---|---|
date | Fri, 21 May 2021 11:25:35 +0200 |
parents | d25f4c0fa160 |
children | 6156a970a2e1 |
line wrap: on
line source
#!/bin/bash set -e set -u ## Starting with version 0.6.2, Orthanc is shipped with a subset of the ## Boost libraries that is generated with the BCP tool: ## ## http://www.boost.org/doc/libs/1_54_0/tools/bcp/doc/html/index.html ## ## This script generates this subset. ## ## History: ## - Orthanc between 0.6.2 and 0.7.3: Boost 1.54.0 ## - Orthanc between 0.7.4 and 0.9.1: Boost 1.55.0 ## - Orthanc between 0.9.2 and 0.9.4: Boost 1.58.0 ## - Orthanc between 0.9.5 and 1.0.0: Boost 1.59.0 ## - Orthanc between 1.1.0 and 1.2.0: Boost 1.60.0 ## - Orthanc 1.3.0: Boost 1.64.0 ## - Orthanc 1.3.1: Boost 1.65.1 ## - Orthanc 1.3.2: Boost 1.66.0 ## - Orthanc between 1.4.0 and 1.4.2: Boost 1.67.0 ## - Orthanc between 1.5.0 and 1.5.4: Boost 1.68.0 ## - Orthanc >= 1.5.5: Boost 1.69.0 BOOST_VERSION=1_69_0 ORTHANC_VERSION=1.5.6 rm -rf /tmp/boost_${BOOST_VERSION} rm -rf /tmp/bcp/boost_${BOOST_VERSION} cd /tmp echo "Uncompressing the sources of Boost ${BOOST_VERSION}..." tar xfz ./boost_${BOOST_VERSION}.tar.gz echo "Generating the subset..." mkdir -p /tmp/bcp/boost_${BOOST_VERSION} bcp --boost=/tmp/boost_${BOOST_VERSION} thread system locale date_time filesystem math/special_functions algorithm uuid atomic iostreams program_options numeric/ublas geometry polygon signals2 chrono /tmp/bcp/boost_${BOOST_VERSION} echo "Removing documentation..." rm -rf /tmp/bcp/boost_${BOOST_VERSION}/libs/locale/doc/html rm -rf /tmp/bcp/boost_${BOOST_VERSION}/libs/algorithm/doc/html rm -rf /tmp/bcp/boost_${BOOST_VERSION}/libs/geometry/doc/html rm -rf /tmp/bcp/boost_${BOOST_VERSION}/libs/geometry/doc/doxy/doxygen_output/html rm -rf /tmp/bcp/boost_${BOOST_VERSION}/libs/filesystem/example/ # https://stackoverflow.com/questions/1655372/longest-line-in-a-file LONGEST_FILENAME=`find /tmp/bcp/ | awk '{print length, $0}' | sort -nr | head -1` LONGEST=`echo "$LONGEST_FILENAME" | cut -d ' ' -f 1` echo echo "Longest filename (${LONGEST} characters):" echo "${LONGEST_FILENAME}" echo if [ ${LONGEST} -ge 128 ]; then echo "ERROR: Too long filename for Windows!" echo exit -1 fi echo "Compressing the subset..." cd /tmp/bcp tar cfz boost_${BOOST_VERSION}_bcpdigest-${ORTHANC_VERSION}.tar.gz boost_${BOOST_VERSION} ls -l boost_${BOOST_VERSION}_bcpdigest-${ORTHANC_VERSION}.tar.gz md5sum boost_${BOOST_VERSION}_bcpdigest-${ORTHANC_VERSION}.tar.gz readlink -f boost_${BOOST_VERSION}_bcpdigest-${ORTHANC_VERSION}.tar.gz