2009/8/6 Dr. David Kirkby <david.kir...@onetel.net> > > The README.txt says > > > SUPPORTED COMPILERS: > * Sage builds with GCC >= 3.x and GCC >= 4.1.x. > * Sage will not build with GCC 2.9.x. > * WARNING: Don't build with GCC 4.0.0, which is very buggy. > * Sage has never been built without using GCC compiler. > > > Yet from what I gather, people on OS X are using 4.0.1 with success.
True. > > I've also herd it said that gcc < 3.4 will not work. True. In fact, GCC 3.x will not work anymore for any value of x, due to ratpoints. This should be considered a bug... but it's still a fact, and as far as I know nobody is even working on fixing it. I wish we wouldn't have let ratpoints get through refereeing without this getting fixed. > > I'm trying to put some sensible tests for the compilers - the current > 'configure' script is lacking in places. I'll show you what I have to date. > > > My aim is to do these things. > > 1) Allows non-GNU compilers, but issues a warning when compilers are not > GNU > 2) Checks if gcc is version is 4.0.0 and if so exits > 3) Checks if version is less than 3.4.0 and if so exits. (but is 3.4.0 > the sensible number)? I don't think any 3.x works anymore. See above. > > 4) Checks if gcc4 is at least 4.0.1 > 5) Checks that if the C compiler is gcc, then the C++ compiler is g++, > and will now allow mixing compilers. > 6) Checks the versions of gcc and g++ are the same, if the GNU tools are > used. > > i.e. this throws an error > > checking gcc version... 3.4.3 > checking if gcc accepts -dumpversion option... yes > checking gxx version... 4.4.1 > configure: error: gcc and g++ must be the same version > > 7) I hope to extend this to fortran - somewhat more difficult to do in > fact. Awesome. That's a good step toward getting rid of shipping g95. > > > I'm defining: > > Buggy_gcc_version="4.0.0" > Sage_minimum_gcc3_version="3.4.0" > Sage_minimum_gcc4_version="4.0.1" > > Are those ok? If not, what should they be? Should they be different for > OS X? > > > This is what I have as a basic configure.ac (this uses M4 macros, do > don't try to build it, but you should get the idea what it does). > > I really need to know what compiler versions to accept, and what to > reject, as things that are said on sage-devel and things that it says in > README.txt are not the same. > > If we are going to permit 4.0.1 on OS X, but not anywhere else, then I'd > need to come up with a test for that. I can't recall why we wouldn't allow 4.0.1 on Linux... > > > > Dave > > AC_PREREQ(2.62) > AC_COPYRIGHT([GPL version 2]) > AC_INIT([Sage],[version], sage-devel@googlegroups.com) > AC_CONFIG_SRCDIR([Sage.c]) > > AM_INIT_AUTOMAKE([1.9.6 foreign]) > > #--------------------------------------------------------- > # Hardware architecture > # > AC_CANONICAL_HOST > > #--------------------------------------------------------- > # C/C++ compilers > > AC_LANG(C) > AC_LANG(C++) > AC_LANG(Fortran) > AC_PROG_CC() > AC_PROG_CXX() > AC_PROG_F77([gfortran g95]) > AC_PROG_CPP > > Buggy_gcc_version="4.0.0" > Sage_minimum_gcc3_version="3.4.0" > Sage_minimum_gcc4_version="4.0.1" > if test x$GCC = xyes > then > dnl Check if C++ compiler is g++. If not, there is a problem. > dnl as mixing GNU and non-GNU compilers is likely to cause problems. > if test x$GXX != xyes > then > AC_MSG_WARN([You are trying to use gcc but not g++]) > AC_MSG_ERROR([The mixing of GNU and non-GNU compilers is not > permitted]) > fi > dnl do likewise with the Fortran compiler (not working yet) > if test x$G77 != xyes > then > AC_MSG_WARN([You are trying to use gcc but not a GNU Fortran > compiler]) > AC_MSG_ERROR([The mixing of GNU and non-GNU compilers is not > permitted]) > fi > > dnl Thank you to Andrew W. Nosenko andrew.w.nose...@gmail.com who > answered my query about > dnl testing of gcc versions on the autoc...@gnu.org mailing list. > dnl AS_VERSION_COMPARE(ver-1, ver-2, [action-if-less], > [action-if-eq], [action-if-greater]) > AX_GCC_VERSION > AX_GXX_VERSION > AS_VERSION_COMPARE([$GCC_VERSION], [$GXX_VERSION], > [AC_MSG_ERROR([gcc and g++ must be the same version])], > [], > [AC_MSG_ERROR([gcc and g++ must be the same version])]) > > dnl Exit if the version of GCC is known to be buggy. > AS_VERSION_COMPARE([$GCC_VERSION], [$Buggy_gcc_version],[], > [ > AC_MSG_WARN([GCC $Buggy_gcc_version is very buggy and can not > build AC_PACKAGE_NAME. Please use]) > AC_MSG_WARN([a GCC 3.x version of at least > $Sage_minimum_gcc3_version or a GCC 4.x version]) > AC_MSG_WARN([of at least $Sage_minimum_gcc4_version. (GCC > $Buggy_gcc_version is too bug ridden)]) > AC_MSG_ERROR([Exiting due to the use of a buggy GCC]) > ] > , []) > dnl Exit if the version of GCC is known to be too old > AS_VERSION_COMPARE([$GCC_VERSION], [$Sage_minimum_gcc3_version],[ > AC_MSG_WARN([GCC $GCC_VERSION is too old and can not build > AC_PACKAGE_NAME. Please use]) > AC_MSG_WARN([a GCC 3.x version of at least > $Sage_minimum_gcc3_version or a GCC 4.x version]) > AC_MSG_WARN([of at least $Sage_minimum_gcc4_version. (GCC > $Buggy_gcc_version is too bug ridden)]) > AC_MSG_ERROR([Exiting due to the use of a version of GCC that > is too old]) > ], [],[]) > > dnl AS_VERSION_COMPARE(ver-1, ver-2, [action-if-less], > [action-if-eq], [action-if-greater]) > else > AC_MSG_WARN([You have a non-GNU compiler, but AC_PACKAGE_NAME has > never been built]) > AC_MSG_WARN([successfully with any compiler other than GCC, despite]) > AC_MSG_WARN([some attempts made on Solaris to use Sun's compiler, > which]) > AC_MSG_WARN([produces faster code than GCC. However, the > AC_PACKAGE_NAME developers]) > AC_MSG_WARN([want AC_PACKAGE_NAME to work with other compilers, so > please try.]) > AC_MSG_WARN([The developers would welcome any feedback you can give.]) > AC_MSG_WARN([If you just want to use AC_PACKAGE_NAME, we suggest > the use of]) > AC_MSG_WARN([a GCC 3.x version of at least > $Sage_minimum_gcc3_version or a GCC 4.x version]) > AC_MSG_WARN([of at least $Sage_minimum_gcc4_version. (GCC > $Buggy_gcc_version is too bug ridden)]) > fi > > AC_CONFIG_HEADERS([config.h]) > AC_CONFIG_FILES([Makefile]) > > AC_OUTPUT > > AC_CONFIG_MACRO_DIR([m4]) > > > > -- William Stein Associate Professor of Mathematics University of Washington http://wstein.org --~--~---------~--~----~------------~-------~--~----~ To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
