Opened 11 years ago

Closed 11 years ago

Last modified 11 years ago

#37362 closed defect (invalid)

atlas Portfile missing dash on force_cpusubtype_ALL option

Reported by: jdgleeson Owned by: Veence (Vincent)
Priority: Normal Milestone:
Component: ports Version: 2.1.2
Keywords: Cc:
Port: atlas

Description

I get the following error when I try to build atlas (OS X 10.5.8 PPC):


:debug:configure Executing org.macports.configure (atlas)
:debug:configure Environment: CPATH='/opt/local/include' CFLAGS='-pipe -O2 -m32' CPPFLAGS='-I/opt/local/include' CXXFLAGS='-pipe -O2 -m32' LIBRARY_PATH='/opt/local/lib' MACOSX_DEPLOYMENT_TARGET='10.5' CXX='/opt/local/bin/g++-mp-4.5' CC_PRINT_OPTIONS_FILE='/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_math_atlas/atlas/work/.CC_PRINT_OPTIONS' F90FLAGS='-pipe -O2 -m32' LDFLAGS='-L/opt/local/lib' FCFLAGS='-pipe -O2 -m32' OBJC='/opt/local/bin/gcc-mp-4.5' INSTALL='/usr/bin/install -c' F90='/opt/local/bin/gfortran-mp-4.5' FC='/opt/local/bin/gfortran-mp-4.5' FFLAGS='-pipe -O2 -m32' OBJCFLAGS='-pipe -O2 -m32' F77='/opt/local/bin/gfortran-mp-4.5' CC_PRINT_OPTIONS='YES' CC='/opt/local/bin/gcc-mp-4.5'
:debug:configure Assembled command: 'cd "/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_math_atlas/atlas/work/atlas-3.10.0/build" && ../configure --prefix=/opt/local -C ac /opt/local/bin/gcc-mp-4.5 --with-netlib-lapack-tarfile=/opt/local/var/macports/distfiles/atlas/lapack-3.4.2.tgz -O 12 -v 5 -b 32 --dylibs -t 2 --cc='/opt/local/bin/gcc-mp-4.5 -m32 force_cpusubtype_ALL' -A 4 -Fa alg '-fPIC -m32 -force_cpusubtype_ALL' -C if /opt/local/bin/gfortran-mp-4.5 -Ss f77lib /opt/local/lib/gcc45/libgfortran.dylib -D c -DWALL'
:debug:configure Executing command line:  cd "/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_math_atlas/atlas/work/atlas-3.10.0/build" && ../configure --prefix=/opt/local -C ac /opt/local/bin/gcc-mp-4.5 --with-netlib-lapack-tarfile=/opt/local/var/macports/distfiles/atlas/lapack-3.4.2.tgz -O 12 -v 5 -b 32 --dylibs -t 2 --cc='/opt/local/bin/gcc-mp-4.5 -m32 force_cpusubtype_ALL' -A 4 -Fa alg '-fPIC -m32 -force_cpusubtype_ALL' -C if /opt/local/bin/gfortran-mp-4.5 -Ss f77lib /opt/local/lib/gcc45/libgfortran.dylib -D c -DWALL 
:info:configure /opt/local/bin/gcc-mp-4.5 -m32 force_cpusubtype_ALL -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_math_atlas/atlas/work/atlas-3.10.0/build/..//CONFIG/include  -g -w -DATL_DYLIBS -o xconfig /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_math_atlas/atlas/work/atlas-3.10.0/build/..//CONFIG/src/config.c atlconf_misc.o 
:info:configure gcc-mp-4.5: force_cpusubtype_ALL: No such file or directory
:info:configure make: *** [xconfig] Error 1

I changed Portfile line 227 from

configure.args-append       --cc='${configure.cc} -m32 force_cpusubtype_ALL' \

to

configure.args-append       --cc='${configure.cc} -m32 -force_cpusubtype_ALL' \

and the build succeeded. (Well, it succeeded eventually. I got an Internal Compiler Error from gcc 4.5. I noticed that Clint Whaley (the atlas maintainer) uses gcc 4.7 whenever possible. I've switched to gcc 4.7 and atlas builds!!)

Change History (6)

comment:1 Changed 11 years ago by jmroot (Joshua Root)

Owner: changed from macports-tickets@… to vince@…
Port: atlas added

Please remember to fill in the Port field and Cc the maintainer.

comment:2 Changed 11 years ago by Veence (Vincent)

Resolution: invalid
Status: newclosed

This was already fixed in r100652. It means you have not refreshed your tree. Worse, you have probably built Atlas with the threaded libraries. It builds fine, but later you will get random assertion failures due to OpenMP not being able to find enough resources. According to Clint, threading has no interest on PowerPC single-processor machine, so I removed it altogether. To wrap up: please uninstall your version, update, and install again… :-|

comment:3 Changed 11 years ago by Veence (Vincent)

PS: it’s not a ‘dash’ – or —, but a ‘hyphen’ -. :-–—)

comment:4 in reply to:  2 Changed 11 years ago by jdgleeson

Replying to vince@…:

This was already fixed in r100652. It means you have not refreshed your tree. Worse, you have probably built Atlas with the threaded libraries. It builds fine, but later you will get random assertion failures due to OpenMP not being able to find enough resources. According to Clint, threading has no interest on PowerPC single-processor machine, so I removed it altogether. To wrap up: please uninstall your version, update, and install again… :-|

I last refreshed just before r100652 was checked in. I did check the repository just before submitting the ticket, but I must have somehow viewed an earlier version without realizing it. Sorry for the mistake.

There are dual processor PowerPCs. I'm using one right now.

comment:5 in reply to:  3 Changed 11 years ago by jdgleeson

Replying to vince@…:

PS: it’s not a ‘dash’ – or —, but a ‘hyphen’ -. :-–—)

Quite right. I was speaking informally. I had no difficulty finding others as careless. For example, a Google search shows that phrases similar to "Options are preceded by a dash" are over 30 times as frequent as "Options are preceded by a hyphen". Nonetheless, I'll try to be more careful in the future; I found many people championing your preference, and none defending my informality. ;-)

comment:6 Changed 11 years ago by Veence (Vincent)

Well, I didn’t mean to be rude in either messages; my apologies if I seemed to. I wasn't aware of dual PPC machines. Threading might give you a better performance, but it is painful to set up, as some atomic lock routines seem to work erratically on PPC. So I guess the best thing is to stick with a single threaded approach.

As for the -, that wee dash of dandyism derives directly from my professional milieu… I work for a French newsmagazine :) Cheers!

Note: See TracTickets for help on using tickets.