#!/bin/bash export SHORT_CIRCUIT_SMOKE if have_sudo ; then # Stop pre-started RDBMS, move their data back to disk (save RAM) # sync for some settle time (not available on all platforms) for d in mysql postgresql ; do # maybe not even running run_or_err "Stopping $d" "sudo /etc/init.d/$d stop || /bin/true" # no longer available on newer build systems if [[ -d /var/ramfs/$d ]] ; then sudo rm -rf /var/lib/$d sudo mv /var/ramfs/$d /var/lib/ sudo ln -s /var/lib/$d /var/ramfs/$d fi done /bin/sync fi # Sanity check VM before continuing echo " ============================================================================= = Startup Meminfo $(free -m -t) =============================================================================" CI_VM_MIN_FREE_MB=2000 if [[ "$(free -m | grep 'buffers/cache:' | perl -p -e '$_ = (split /\s+/, $_)[3]')" -lt "$CI_VM_MIN_FREE_MB" ]]; then SHORT_CIRCUIT_SMOKE=1 echo_err " ============================================================================= CI virtual machine stuck in a state with a lot of memory locked for no reason. Under Travis this state usually results in a failed build. Short-circuiting buildjob to avoid false negatives, please restart it manually. =============================================================================" # pull requests are always scrutinized after the fact anyway - run a # a simpler matrix elif [[ "$TRAVIS_PULL_REQUEST" != "false" ]]; then if [[ -n "$BREWVER" ]]; then # just don't brew anything SHORT_CIRCUIT_SMOKE=1 else # running PRs with 1 thread is non-sensical VCPU_USE="" fi fi if [[ -n "$SHORT_CIRCUIT_SMOKE" ]] ; then return ; fi # Previously we were going off the OpenVZ vcpu count and dividing by 3 # With the new infrastructure, somply go with "something high" export VCPU_AVAILABLE=10 if [[ -z "$VCPU_USE" ]] ; then export VCPU_USE="$VCPU_AVAILABLE" fi if [[ "$CLEANTEST" != "true" ]]; then if [[ -z "$(tail -n +2 /proc/swaps)" ]] ; then run_or_err "Configuring swap (for Oracle)" \ "sudo bash -c 'dd if=/dev/zero of=/swap.img bs=256M count=5 && chmod 600 /swap.img && mkswap /swap.img && swapon /swap.img'" fi # never installed, this looks like trusty if [[ ! -d /var/lib/mysql ]] ; then sudo dpkg --add-architecture i386 extra_debs+=( postgresql mysql-server ) fi # these APT sources do not mean anything to us anyway sudo rm -rf /etc/apt/sources.list.d/* # # FIXME these debconf lines should automate the firebird config but seem not to :((( sudo bash -c 'echo -e "firebird2.5-super\tshared/firebird/enabled\tboolean\ttrue" | debconf-set-selections' sudo bash -c 'echo -e "firebird2.5-super\tshared/firebird/sysdba_password/new_password\tpassword\t123" | debconf-set-selections' run_or_err "Updating APT sources" "sudo apt-get update" apt_install ${extra_debs[@]} libmysqlclient-dev memcached firebird2.5-super firebird2.5-dev expect # need to stop them again, in case we installed them above (trusty) for d in mysql postgresql ; do run_or_err "Stopping $d" "sudo /etc/init.d/$d stop || /bin/true" done export CACHE_DIR="/tmp/poormanscache" mkdir "$CACHE_DIR" # FIXME - by default db2 eats too much memory, we won't be able to test on legacy infra # someone needs to add a minimizing configuration akin to 9367d187 if [[ "$(free -m | grep 'Mem:' | perl -p -e '$_ = (split /\s+/, $_)[1]')" -gt 4000 ]] ; then run_or_err "Getting DB2 from poor man's cache github" ' wget -qO- https://github.com/poormanscache/poormanscache/archive/DB2_ExC/9.7.5_deb_x86-64.tar.gz \ | tar -C "$CACHE_DIR" -zx' # the actual package is built for lucid, installs fine on both precise and trusty manual_debs+=( "db2exc_9.7.5-0lucid0_amd64.deb" ) fi run_or_err "Getting Oracle from poor man's cache github" ' wget -qO- https://github.com/poormanscache/poormanscache/archive/OracleXE/10.2.0_deb_mixed.tar.gz \ | tar -C "$CACHE_DIR" -zx' manual_debs+=( "bc-multiarch-travis_1.0_all.deb" "oracle-xe_10.2.0.1-1.1_i386.deb" ) # reassemble chunked pieces ( working around github's filesize limit ) for reass in $CACHE_DIR/*/reassemble ; do /bin/bash "$reass" ; done run_or_err "Installing RDBMS debs manually: $( echo ${manual_debs[@]/#/$CACHE_DIR/*/*/} )" \ "sudo dpkg -i $( echo ${manual_debs[@]/#/$CACHE_DIR/*/*/} ) || sudo bash -c 'source maint/travis-ci_scripts/common.bash && apt_install -f'" # Needs to happen separately and *after* db2exc, as the former shits all over /usr/include (wtf?!) # For more info look at /opt/ibm/db2/V9.7/instance/db2iutil :: create_links() # The --reinstall is there in case it was already in place and got destroyed # (this is the case on newer trusty images) apt_install --reinstall unixodbc-dev ### config memcached run_or_err "Starting memcached" "sudo /etc/init.d/memcached start" export DBICTEST_MEMCACHED=127.0.0.1:11211 ### config mysql run_or_err "Installing minimizing MySQL config" "\ sudo bash -c 'rm /var/lib/mysql/ib*' \ && sudo cp maint/travis-ci_scripts/configs/minimal_mysql_travis.cnf /etc/mysql/conf.d/ \ && sudo chmod 644 /etc/mysql/conf.d/*.cnf \ " run_or_err "Starting MySQL" "sudo /etc/init.d/mysql start" run_or_err "Creating MySQL TestDB" "mysql -u root -e 'create database dbic_test;'" export DBICTEST_MYSQL_DSN='dbi:mysql:database=dbic_test;host=127.0.0.1' export DBICTEST_MYSQL_USER=root ### config pg run_or_err "Starting PostgreSQL" "sudo /etc/init.d/postgresql start" run_or_err "Creating PostgreSQL TestDB" "psql -c 'create database dbic_test;' -U postgres" export DBICTEST_PG_DSN='dbi:Pg:database=dbic_test;host=127.0.0.1' export DBICTEST_PG_USER=postgres ### conig firebird # poor man's deb config EXPECT_FB_SCRIPT=' spawn dpkg-reconfigure --frontend=text firebird2.5-super expect "Enable Firebird server?" send "\177\177\177\177yes\r" expect "Password for SYSDBA" send "123\r" sleep 2 expect eof ' # creating testdb # FIXME - this step still fails from time to time >:((( # has to do with the FB reconfiguration I suppose # for now if it fails twice - simply skip FB testing for i in 1 2 3 ; do run_or_err "Re-configuring Firebird" " sync sleep 5 DEBIAN_FRONTEND=text sudo expect -c '$EXPECT_FB_SCRIPT' " if run_or_err "Creating Firebird TestDB" \ "echo \"CREATE DATABASE '/var/lib/firebird/2.5/data/dbic_test.fdb';\" | sudo isql-fb -u sysdba -p 123" then # Do not upgrade to a newer ODBC driver - smoking on an old # and buggy POS is much more valuable # run_or_err "Fetching and building Firebird ODBC driver" ' cd "$(mktemp -d)" wget -qO- https://github.com/dbsrgits/Firebird-ODBC-driver/archive/2.0.2.153.tar.gz | tar -zx --strip-components 1 cd Builds/Gcc.lin make -f makefile.linux sudo make -f makefile.linux install ' sudo bash -c 'cat >> /etc/odbcinst.ini' <<< " [Firebird] Description = InterBase/Firebird ODBC Driver Driver = /usr/lib/x86_64-linux-gnu/libOdbcFb.so Setup = /usr/lib/x86_64-linux-gnu/libOdbcFb.so Threading = 1 FileUsage = 1 " export DBICTEST_FIREBIRD_DSN=dbi:Firebird:dbname=/var/lib/firebird/2.5/data/dbic_test.fdb export DBICTEST_FIREBIRD_USER=SYSDBA export DBICTEST_FIREBIRD_PASS=123 export DBICTEST_FIREBIRD_INTERBASE_DSN=dbi:InterBase:dbname=/var/lib/firebird/2.5/data/dbic_test.fdb export DBICTEST_FIREBIRD_INTERBASE_USER=SYSDBA export DBICTEST_FIREBIRD_INTERBASE_PASS=123 export DBICTEST_FIREBIRD_ODBC_DSN="dbi:ODBC:Driver=Firebird;Dbname=/var/lib/firebird/2.5/data/dbic_test.fdb" export DBICTEST_FIREBIRD_ODBC_USER=SYSDBA export DBICTEST_FIREBIRD_ODBC_PASS=123 break fi done ### config oracle SRV_ORA_HOME=/usr/lib/oracle/xe/app/oracle/product/10.2.0/server # without this some of the more zealous tests can exhaust the amount # of listeners and oracle is too slow to spin extras up :( sudo bash -c "echo -e '\nprocesses=150' >> $SRV_ORA_HOME/config/scripts/init.ora" EXPECT_ORA_SCRIPT=' spawn /etc/init.d/oracle-xe configure sleep 1 set send_slow {1 .005} expect "Specify the HTTP port that will be used for Oracle Application Express" sleep 0.5 send -s "8021\r" expect "Specify a port that will be used for the database listener" sleep 0.5 send -s "1521\r" expect "Specify a password to be used for database accounts" sleep 0.5 send -s "adminpass\r" expect "Confirm the password" sleep 0.5 send -s "adminpass\r" expect "Do you want Oracle Database 10g Express Edition to be started on boot" sleep 0.5 send -s "n\r" sleep 0.5 expect "Configuring Database" sleep 1 expect eof wait ' # if we do not redirect to some random file, but instead try to capture # into a var the way run_or_err does - everything hangs # FIXME: I couldn't figure it out after 3 hours of headdesking, # would be nice to know the reason eventually run_or_err "Configuring OracleXE" "sudo $(which expect) -c '$EXPECT_ORA_SCRIPT' &>/tmp/ora_configure_10.2.log" export DBICTEST_ORA_DSN=dbi:Oracle://localhost:1521/XE export DBICTEST_ORA_USER=dbic_test export DBICTEST_ORA_PASS=abc123456 export DBICTEST_ORA_EXTRAUSER_DSN="$DBICTEST_ORA_DSN" export DBICTEST_ORA_EXTRAUSER_USER=dbic_test_extra export DBICTEST_ORA_EXTRAUSER_PASS=abc123456 run_or_err "Create Oracle users" "ORACLE_SID=XE ORACLE_HOME=$SRV_ORA_HOME $SRV_ORA_HOME/bin/sqlplus -L -S system/adminpass @/dev/stdin <<< ' CREATE USER $DBICTEST_ORA_USER IDENTIFIED BY $DBICTEST_ORA_PASS; GRANT connect,resource TO $DBICTEST_ORA_USER; CREATE USER $DBICTEST_ORA_EXTRAUSER_USER IDENTIFIED BY $DBICTEST_ORA_EXTRAUSER_PASS; GRANT connect,resource TO $DBICTEST_ORA_EXTRAUSER_USER; '" export ORACLE_HOME="$CACHE_DIR/poormanscache-OracleXE-10.2.0_deb_mixed/ora_instaclient/x86-64/oracle_instaclient_10.2.0.5.0" ### config db2exc # we may have skipped installation due to low memory if dpkg -l db2exc &>/dev/null ; then # WTF is this world-writable? # Strip the write bit so it doesn't trip Ubuntu's symlink-in-/tmp attack mitigation sudo chmod -R o-w ~dasusr1/das export DB2_HOME=/opt/ibm/db2/V9.7 export DBICTEST_DB2_DSN=dbi:DB2:DATABASE=dbictest export DBICTEST_DB2_USER=db2inst1 export DBICTEST_DB2_PASS=abc123456 run_or_err "Set up DB2 users" \ "echo -e '$DBICTEST_DB2_PASS\n$DBICTEST_DB2_PASS' | sudo passwd $DBICTEST_DB2_USER" run_or_err "Create DB2 database" \ "sudo -u $DBICTEST_DB2_USER -i db2 'CREATE DATABASE dbictest' && sudo -u $DBICTEST_DB2_USER -i db2 'ACTIVATE DATABASE dbictest'" fi fi