gfortran on MacOS X

Bill Northcott w.northcott@unsw.edu.au
Wed Jul 6 23:54:00 GMT 2005


I just built gfortran on MacOS X 10.4.1 as part of an Apple style  
compiler.

I started with the gcc-5026 sources for gcc-4 as shipped with Xcode  
2.1.  I diffed the four Fortran directories against the gcc-4_0- 
branch as at 4 July and applied all the updates.  I also applied the  
uncommitted fix for PR17917.  I added f95 to the --enable-languages  
option on line 102 of Apple's build_gcc script.

With these mods a standard Apple build built gfortran including 64  
bit libraries.  The only problem is that libgfortran.a and  
libgfortranbegin.a were not copied from the staging directory to the  
final distribution.

This compiler was used to build recent versions of R and Octave which  
passed all their own 'make check' tests.  Thanks to everyone involved.

The build also included 64 bit libraries, but I do not have access to  
a G5 to test these.

The dejagnu testsuite gave the following results:
         === gfortran Summary ===

# of expected passes        6910
# of unexpected failures    9
# of expected failures        10
# of unsupported tests        17
/Users/billn/Public/Darwin/gcc-5026patched/build/obj/obj-powerpc- 
powerpc/gcc/testsuite/../gfortran  version 4.0.0 (Apple Computer,  
Inc. build 5026)

The unexpected failures were:
FAIL: gfortran.dg/dev_null.f90  -O0  execution test
FAIL: gfortran.dg/dev_null.f90  -O1  execution test
FAIL: gfortran.dg/dev_null.f90  -O2  execution test
FAIL: gfortran.dg/dev_null.f90  -O3 -fomit-frame-pointer  execution test
FAIL: gfortran.dg/dev_null.f90  -O3 -fomit-frame-pointer -funroll- 
loops  execution test
FAIL: gfortran.dg/dev_null.f90  -O3 -fomit-frame-pointer -funroll-all- 
loops -finline-functions  execution test
FAIL: gfortran.dg/dev_null.f90  -O3 -g  execution test
FAIL: gfortran.dg/dev_null.f90  -Os  execution test
FAIL: gfortran.dg/eoshift.f90  -O0  execution test

These seem be PR18958 and PR19478.  I have checked that the 'fixes'  
for these problems were in the sources I compiled.  Also PR19273 does  
not seem to be relevant as all sources were updated prior to any code  
being built.  So maybe there is some problem with these fixes.

Bill Northcott



More information about the Fortran mailing list