[Scons-users] Problem with Fortran Dependency Scanner
Damien
damien at khubla.com
Mon Apr 27 13:31:49 EDT 2015
Nikolaus,
No, I'm not sure. :-) 99% of our code is C++, we just have Fortran for
a sparse linear solver that we use as an option. It's a self-contained
compilation to a lib. You could modify the
env['FORTRANMODDIR'] = '${TARGET.dir}'
to put the mods somewhere else in a common directory I suppose, but
we've never needed it.
Damien
On 2015-04-27 11:11 AM, Nikolaus Rath wrote:
> Hi Damien,
>
> If I add '.' to FORTRANPATH, the build succeeds and the com.mod file is
> ends up in src2 (next to com.f90).
>
> Are you sure that your method also works if the dependencies cross
> different directories?
>
> Best,
> -Nikolaus
>
> On 04/27/2015 09:59 AM, Damien wrote:
>> Is the mod file actually being generated somewhere in the tree?
>>
>> This is what we do to set up our Fortran command line on Linux:
>>
>> env['FORTRAN'] = 'gfortran'
>> env['FORTRANMODDIRPREFIX'] = '-J'
>> env['FORTRANMODDIR'] = '${TARGET.dir}'
>> env['FORTINCPREFIX'] = '-I'
>> env['_FORTRANINCFLAGS'] = '$( ${_concat(FORTINCPREFIX, FORTRANPATH,
>> INCSUFFIX, __env__, RDirs, TARGET, SOURCE)} $)'
>> env['FORTRANCOM'] = '$FORTRAN -o$TARGET -c $FORTRANFLAGS
>> $FORTRANMODDIRPREFIX$FORTRANMODDIR $_FORTRANINCFLAGS $SOURCES'
>> env['SHFORTRANCOM'] = env['FORTRANCOM']
>> env['SHF90PPCOM'] = env['FORTRANCOM']
>> env['SHF95PPCOM'] = env['FORTRANCOM']
>> env['F90PPCOM'] = env['FORTRANCOM']
>> env['F95PPCOM'] = env['FORTRANCOM']
>>
>> We had a world of pain getting Fortran modules to work, and we did this
>> many years ago, with a similar blunt chainsaw on the Windows and OSX
>> builds, but it works.
>>
>> Damien
>>
>>
>>
>> On 2015-04-27 10:49 AM, Nikolaus Rath wrote:
>>> Hi Damien,
>>>
>>> No luck:
>>>
>>> scons: Building targets ...
>>> gfortran -o bar.o -c -g -Wall -O3 -DSETTING
>>> -I/home/nikratio/tmp/scons-test/inc -J. bar.F
>>> bar.F:2.9:
>>>
>>> USE com
>>> 1
>>> Fatal Error: Can't open module file 'com.mod' for reading at (1): No
>>> such file or directory
>>> scons: *** [bar.o] Error 1
>>>
>>>
>>> Best,
>>> -Nikolaus
>>>
>>>
>>> On 04/27/2015 08:10 AM, Damien wrote:
>>>> Try adding this to your flags:
>>>>
>>>> env['FORTRANMODDIRPREFIX'] = '-J'
>>>> env['FORTRANMODDIR'] = '${TARGET.dir}'
>>>>
>>>> Modules and directories can be a bit of a pain...
>>>>
>>>> Damien
>>>>
>>>> On 2015-04-27 8:50 AM, Nikolaus Rath wrote:
>>>>> Hello,
>>>>>
>>>>> I have the following tiny test project:
>>>>>
>>>>> $ find -type f
>>>>> ./src1/foo.f
>>>>> ./src1/bar.F
>>>>> ./src1/bar.mod
>>>>> ./src1/SConstruct
>>>>> ./src2/com.f90
>>>>> ./inc/params.f
>>>>>
>>>>> $ cat src1/bar.F
>>>>> PROGRAM bar
>>>>> USE com
>>>>> IMPLICIT NONE
>>>>>
>>>>> PRINT *, 'Starting bar'
>>>>> PRINT *, 'Value of SETTING is', SETTING
>>>>> call com_sub
>>>>> PRINT *, 'End of bar'
>>>>> END
>>>>>
>>>>> $ cat src2/com.f90
>>>>> MODULE com
>>>>> USE foo
>>>>> INTEGER:: com_int=17
>>>>> CONTAINS
>>>>> subroutine com_sub
>>>>> PRINT *, 'Starting com_sub'
>>>>> CALL foo_sub
>>>>> PRINT *, 'End of com_sub'
>>>>> end SUBROUTINE com_sub
>>>>> END MODULE com
>>>>>
>>>>> $ cat src1/foo.f
>>>>> MODULE foo
>>>>> IMPLICIT NONE
>>>>> INCLUDE 'params.f'
>>>>> INTEGER foo_int
>>>>>
>>>>> CONTAINS
>>>>>
>>>>> subroutine foo_sub
>>>>> PRINT *, 'In foo_sub'
>>>>> end SUBROUTINE
>>>>> END MODULE foo
>>>>>
>>>>> $ cat inc/params.f
>>>>> real,parameter::zero=0,one=1
>>>>>
>>>>>
>>>>> I'm trying to build it with the following Scons script:
>>>>>
>>>>> $ cat src1/SConstruct
>>>>> # Attention, emacs, this is a -*- python -*- file.
>>>>> import os
>>>>>
>>>>> opts = Variables(None, ARGUMENTS)
>>>>> opts.Add(BoolVariable('DEBUG', 'Enable debugging', False))
>>>>> opts.Add(BoolVariable('SETTING', 'Enable setting', True))
>>>>>
>>>>> env = DefaultEnvironment(variables = opts,
>>>>> ENV = {'PATH' : os.environ['PATH']})
>>>>>
>>>>> # Use the same settings for all Fortran dialects
>>>>> for dialect in ('F77', 'F90', 'F95', 'F03'):
>>>>> for var_suffix in ('', 'FLAGS', 'PATH'):
>>>>> env['%s%s' % (dialect, var_suffix)] = \
>>>>> '${FORTRAN%s}' % var_suffix
>>>>>
>>>>> opts = Split('-g -Wall')
>>>>> if env['DEBUG']:
>>>>> opts += Split('-O0 -g')
>>>>> else:
>>>>> opts += Split('-O3')
>>>>> env.Append(FORTRANFLAGS=opts, CPPDEFINES=['SETTING'],
>>>>> FORTRANPATH=['#/../inc'])
>>>>>
>>>>> sources = [ 'bar.F',
>>>>> 'foo.f',
>>>>> '../src2/com.f90' ]
>>>>>
>>>>> # Object method returns both .mod and .o, so we filter for the latter.
>>>>> objects = [ x for x in Object(sources) if x.get_suffix() == '.o' ]
>>>>> Program('test', objects)
>>>>>
>>>>>
>>>>> However, this fails with:
>>>>>
>>>>> $ scons
>>>>> scons: Reading SConscript files ...
>>>>> scons: done reading SConscript files.
>>>>> scons: Building targets ...
>>>>> gfortran -o /home/nikratio/tmp/scons-test/src2/com.o -c -g -Wall -O3
>>>>> -I/home/nikratio/tmp/scons-test/inc
>>>>> /home/nikratio/tmp/scons-test/src2/com.f90
>>>>> /home/nikratio/tmp/scons-test/src2/com.f90:2.6:
>>>>>
>>>>> USE foo
>>>>> 1
>>>>> Fatal Error: Can't open module file 'foo.mod' for reading at (1): No
>>>>> such file or directory
>>>>> scons: *** [/home/nikratio/tmp/scons-test/src2/com.o] Error 1
>>>>> scons: building terminated because of errors.
>>>>>
>>>>>
>>>>> It seems that Scons did not notice that src2/com.f90 depends on src1/foo.f.
>>>>>
>>>>>
>>>>> Why is that?
>>>>>
>>>>>
>>>>> With experimenting, I found out that it helps if I add '.' to
>>>>> FORTRANPATH - but that seems wrong, because I don't want to include any
>>>>> files from src1/.
>>>>>
>>>>>
>>>>> Best,
>>>>> -Nikolaus
>>>>>
>>>> _______________________________________________
>>>> Scons-users mailing list
>>>> Scons-users at scons.org
>>>> https://pairlist4.pair.net/mailman/listinfo/scons-users
>>>>
>> _______________________________________________
>> Scons-users mailing list
>> Scons-users at scons.org
>> https://pairlist4.pair.net/mailman/listinfo/scons-users
>>
>
More information about the Scons-users
mailing list