Martin Costabel wrote: [] > I intend to commit a fixed version shortly (if nobody protests), I am > just finishing some more tests. On 10.4.11/ppc it already compiles OK, > on 10.5.1/intel it looks like it will.
I just did this now. This version compiles and runs for me on both configurations. I fixed two things: 1) --with-f77 is not recognized by configure, therefore g77 is used if present, although gfortran should be used. This leads to undefined symbols when linking dylibs. Simple fix: F77 instead of --with-f77. 2) In the test for the atlas variant, there is a sh syntax error: "if test [ ]" leads to the error message /var/tmp/tmp.3.ZN0uEn: line 5: test: too many arguments I don't know why this error is ignored; the -e flag is set, so this should bomb, but it doesn't. Fix: Remove the word "test". In this test, the first alternative does not work either, because of sh quoting problems. I fixed this, too. I have a couple of questions on this version, in case anyone who knows is present: a) The "-atlas" variant is desactivated by the "Type: -atlas (.)" field. I suppose this is intentional, because it does not work yet? b) The info file says: " Build with completely with gcc42 compilers on Macintel to avoid bus error in ov_fcn_handle.cc test case." I did not notice either of these, i.e. whereas gfortran from the gcc42 package is used, the C++ compiler is "g++" which is /usr/bin/g++ and not /sw/bin/g++-4; but there was no bus error in the test phase. The only test error was in the test_system phase, probably due to fink's special setting of $HOME. -- Martin ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Fink-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/fink-users
