Re: [FLASH-USERS] setup problems

Hi,

I don't have g95 so I only tested that the preprocessor didn't choke on
it. Sorry if it doesn't work. I think that, if the fortran compiler is
adding two underscores only to certain symbols, we can't work around it
in the preprocessor. I guess the solution for Steve is to change the
names by hand (or with a script) like you said, or give g95 the right
flags and rebuild mpich and company. Regards,

Alvaro

Alexander Heger wrote:
> Dear Alvaro,
>
> did you test that?
>
> The problem I had is that the second underscore needs to be added only
> if the name contains an underscore, not generically to all the calls. I
> don't know how that can be put into mangle_names.h, but if you know how
> to do that, this would likely fix the problem indeed.
>
> Best wishes,
> Alexander
>
> Alvaro Caceres wrote:
>
>>I think all of FLASH's c routines use mangle_names.h, so to fix the
>>linker problem it should look something like:
>>
>>#ifndef MANGLE_NAMES_H
>>#define MANGLE_NAMES_H
>>
>>#ifdef IBM
>>#define FTOC(x) x
>>
>>#else
>>#ifdef G95
>>#define FTOC(x) x##__
>>
>>#else /* default * /
>>#define FTOC(x) x##_
>>#endif /* G95 */
>>
>>#endif /* IBM */
>>
>>#endif /* MANGLE_NAMES_H */
>>
>>Alvaro
>>Tomasz Plewa wrote:
>>
>>>Thanks, Georgi. I reckon the same may work for hdf5.
>>>
>>>Still the permanent solution is to fix the FLASH source code. It is
>>>certainly doable.
>>>
>>>Tomek
>>>--
>>>On Tue, Oct 04, 2005 at 05:32:56PM +0100, Georgi Pavlovski wrote:
>>>
>>>
>>>>hi,
>>>>
>>>>I'm not sure if this is the cure, but it worth a try. While compiling
>>>>mpich set the environment variable,
>>>>
>>>>setenv pac_cv_prog_f77_name_mangle 'lower underscore'
>>>>
>>>>I remember having similar problems it the past, but I used intel fortran
>>>>compiler.
>>>>
>>>>- Georgi
>>>>
>>>>On Tue, 2005-10-04 at 10:06 -0500, Tomasz Plewa wrote:
>>>>
>>>>
>>>>>Steve -
>>>>>
>>>>>You can disregard warnings. However, compilation process fails during
>>>>>linking. Apparently, gcc/g95 combination produces object symbols with
>>>>>inconsistent number of underscores. There are some folks who are
>>>>>working on the workaround for this problem. (If someone has a working
>>>>>solution, please let us know!)
>>>>>
>>>>>Tomek
>>>>>--
>>>>>On Tue, Oct 04, 2005 at 03:48:54PM +0100, Steve Spreckley wrote:
>>>>>
>>>>>
>>>>>>Hi,
>>>>>>
>>>>>>I am new to flash, but I have been trying to install and run the
>>>>>>software
>>>>>>for some time with little success. I have used g95 compilers to
>>>>>>compile
>>>>>>and install mpich 1.2.7 and hdf5 1.6.4, and the example code for these
>>>>>>compiled and ran fine. However, when trying to compile flash 2.3 I
>>>>>>get a
>>>>>>number of warnings as follows:
>>>>>>
>>>>>>In file logfile.F90:625
>>>>>>
>>>>>>write (log_lun, "(X, A30, ' : ', A20)") trim(adjustl(str_array(i,
>>>>>> 1
>>>>>>Warning (100): X descriptor requires leading space count in format
>>>>>>string
>>>>>>at (1)
>>>>>>In file logfile.F90:628
>>>>>>
>>>>>> write (log_lun, "(X, A15, 22X, A8, 2X, A9, 4X, A8, 2X, A8)") trim
>>>>>> 1
>>>>>>Warning (100): X descriptor requires leading space count in format
>>>>>>string
>>>>>>at (1)
>>>>>>
>>>>>>
>>>>>>and then a number of lines like this for several functions:
>>>>>>
>>>>>>
>>>>>>dBaseAPI.c: In function write_logfile:
>>>>>>dBaseAPI.c:11: warning: incompatible implicit declaration of built-in
>>>>>>function strlen
>>>>>>dBaseAPI.c: In function stamp_logfile:
>>>>>>dBaseAPI.c:16: warning: incompatible implicit declaration of built-in
>>>>>>function strlen.
>>>>>>
>>>>>>
>>>>>>The code gets past these though and kicks out more F90 warnings:
>>>>>>
>>>>>>
>>>>>>In file checkpoint_re.F90:542
>>>>>>
>>>>>> write (num_to_str, '(I)') tot_blocks
>>>>>> 1
>>>>>>Warning (100): Nonnegative width required in format string at (1)
>>>>>>
>>>>>>In file checkpoint_re.F90:547
>>>>>>
>>>>>> write (num_to_str, '(I)') ReNumParts
>>>>>> 1
>>>>>>Warning (100): Nonnegative width required in format string at (1)
>>>>>>
>>>>>>mpif90 -c -r8 -i4 -I/home/sas/mpich-1.2.7/include -DN_DIM=2
>>>>>>-DMAXBLOCKS=1000 -DNXB=8 -DNYB=8 -DNZB=1 checkpoint_wr.F90
>>>>>>In file checkpoint_wr.F90:221
>>>>>>
>>>>>> write (num_to_str, '(I)') MAX_BLK_MSGS
>>>>>> 1
>>>>>>Warning (100): Nonnegative width required in format string at (1).
>>>>>>
>>>>>>
>>>>>>I then get a number of warnings associated with hdf5:
>>>>>>
>>>>>>
>>>>>>h5_write.c: In function h5_write_tree_info_:
>>>>>>h5_write.c:943: warning: pointer targets in passing argument 3 of
>>>>>>H5Sselect_hyperslab differ in signedness
>>>>>>h5_write.c:1030: warning: pointer targets in passing argument 3 of
>>>>>>H5Sselect_hyperslab differ in signedness
>>>>>>h5_write.c:1117: warning: pointer targets in passing argument 3 of
>>>>>>H5Sselect_hyperslab differ in signedness
>>>>>>h5_write.c:1204: warning: pointer targets in passing argument 3 of
>>>>>>H5Sselect_hyperslab differ in signedness
>>>>>>h5_write.c:1230: warning: pointer targets in passing argument 3 of
>>>>>>H5Sselect_hyperslab differ in signednes
>>>>>>
>>>>>>
>>>>>>and these continue for more h5_ funtions.
>>>>>>
>>>>>>The code finally crashes though when linking:
>>>>>>
>>>>>>
>>>>>>aseAPI.o(.text+0x1c): In function `write_logfile':
>>>>>>dBaseAPI.c: undefined reference to `xwrite_logfile_'
>>>>>>dBaseAPI.o(.text+0x44): In function `stamp_logfile':
>>>>>>dBaseAPI.c: undefined reference to `xstamp_logfile_'
>>>>>>dBaseAPI.o(.text+0x6f): In function `get_real_parm_from_glob_context':
>>>>>>dBaseAPI.c: undefined reference to `xget_real_parm_from_glob_contex_'
>>>>>>dBaseAPI.o(.text+0x9b): In function `get_int_parm_from_glob_context':
>>>>>>dBaseAPI.c: undefined reference to `xget_int_parm_from_glob_context_'
>>>>>>dBaseAPI.o(.text+0xd6): In function `get_str_parm_from_glob_context':
>>>>>>dBaseAPI.c: undefined reference to `xget_str_parm_from_glob_context_'
>>>>>>init_flash.o(.text+0x11): In function `init_flash__':
>>>>>>init_flash.F90: undefined reference to `init_signal_handler__'
>>>>>>driverAPI.o(.text+0x1c): In function `abort_flash':
>>>>>>driverAPI.c: undefined reference to `xabort_flash_'
>>>>>>checkpoint_re.o(.text+0x8fa): In function `checkpoint_filename_re__':
>>>>>>checkpoint_re.F90: undefined reference to `h5_open_file_for_read__'
>>>>>>checkpoint_re.o(.text+0x97e):checkpoint_re.F90: undefined reference to
>>>>>>`h5_read_timers__'
>>>>>>checkpoint_re.o(.text+0x9d3):checkpoint_re.F90: undefined reference to
>>>>>>`h5_read_header_info__'
>>>>>>checkpoint_re.o(.text+0x19e3):checkpoint_re.F90: undefined
>>>>>>reference to
>>>>>>`get_numparticles__'
>>>>>>checkpoint_re.o(.text+0x1cf5):checkpoint_re.F90: undefined
>>>>>>reference to
>>>>>>`h5_read_particles__'
>>>>>>checkpoint_re.o(.text+0x22a9):checkpoint_re.F90: undefined
>>>>>>reference to
>>>>>>`h5_read_tree_info__'
>>>>>>checkpoint_re.o(.text+0x24bc):checkpoint_re.F90: undefined
>>>>>>reference to
>>>>>>`h5_read_grid_info__'
>>>>>>checkpoint_re.o(.text+0x25fd):checkpoint_re.F90: undefined
>>>>>>reference to
>>>>>>`h5_read_unknowns__'
>>>>>>checkpoint_re.o(.text+0x3ab3):checkpoint_re.F90: undefined
>>>>>>reference to
>>>>>>`h5_close_file__'
>>>>>>checkpoint_wr.o(.text+0xd77): In function `checkpoint_wr__':
>>>>>>checkpoint_wr.F90: undefined reference to `h5_initialize_file__'
>>>>>>checkpoint_wr.o(.text+0x14f1):checkpoint_wr.F90: undefined
>>>>>>reference to
>>>>>>`h5_write_header_info__'
>>>>>>checkpoint_wr.o(.text+0x1730):checkpoint_wr.F90: undefined
>>>>>>reference to
>>>>>>`h5_write_runtime_parameters__'
>>>>>>checkpoint_wr.o(.text+0x299e):checkpoint_wr.F90: undefined
>>>>>>reference to
>>>>>>`h5_write_particles__'
>>>>>>checkpoint_wr.o(.text+0x2a16):checkpoint_wr.F90: undefined
>>>>>>reference to
>>>>>>`h5_write_tree_info__'
>>>>>>checkpoint_wr.o(.text+0x2a60):checkpoint_wr.F90: undefined
>>>>>>reference to
>>>>>>`h5_write_grid_info__'
>>>>>>checkpoint_wr.o(.text+0x2b97):checkpoint_wr.F90: undefined
>>>>>>reference to
>>>>>>`h5_write_unknowns__'
>>>>>>checkpoint_wr.o(.text+0x3e38):checkpoint_wr.F90: undefined
>>>>>>reference to
>>>>>>`h5_write_timers__'
>>>>>>checkpoint_wr.o(.text+0x4022):checkpoint_wr.F90: undefined
>>>>>>reference to
>>>>>>`h5_close_file__'
>>>>>>plotfile.o(.text+0xc6a): In function `plotfile_':
>>>>>>plotfile.F90: undefined reference to `h5_initialize_file__'
>>>>>>plotfile.o(.text+0x12ce):plotfile.F90: undefined reference to
>>>>>>`h5_write_header_info_sp__'
>>>>>>plotfile.o(.text+0x1504):plotfile.F90: undefined reference to
>>>>>>`h5_write_runtime_parameters__'
>>>>>>plotfile.o(.text+0x1e8f):plotfile.F90: undefined reference to
>>>>>>`h5_write_tree_info_sp__'
>>>>>>plotfile.o(.text+0x220c):plotfile.F90: undefined reference to
>>>>>>`h5_write_grid_info_sp__'
>>>>>>plotfile.o(.text+0x2be4):plotfile.F90: undefined reference to
>>>>>>`h5_write_unknowns_sp__'
>>>>>>plotfile.o(.text+0x2cca):plotfile.F90: undefined reference to
>>>>>>`h5_write_unknowns_sp__'
>>>>>>plotfile.o(.text+0x2da3):plotfile.F90: undefined reference to
>>>>>>`h5_write_unknowns_sp__'
>>>>>>plotfile.o(.text+0x3a78):plotfile.F90: undefined reference to
>>>>>>`h5_close_file__'
>>>>>>init_global_parms_checkpoint.o(.text+0x2cd): In function
>>>>>>`init_global_parms_checkpoint__':
>>>>>>init_global_parms_checkpoint.F90: undefined reference to
>>>>>>`h5_open_file_for_read__'
>>>>>>init_global_parms_checkpoint.o(.text+0x3ac):init_global_parms_checkpoint.F90:
>>>>>>
>>>>>>undefined reference to `h5_read_runtime_parameters__'
>>>>>>init_global_parms_checkpoint.o(.text+0x3b8):init_global_parms_checkpoint.F90:
>>>>>>
>>>>>>undefined reference to `h5_close_file__'
>>>>>>multifluidAPI.o(.text+0x1f): In function `find_fluid_index':
>>>>>>multifluidAPI.c: undefined reference to `xfind_fluid_index_'
>>>>>>memory_report.o(.text+0x61): In function `memory_report__':
>>>>>>memory_report.F90: undefined reference to `memory_usage__'
>>>>>>gmake[1]: *** [flash2] Error 1
>>>>>>gmake[1]: Leaving directory
>>>>>>`/home/sas/work/exoplanets/MHD/FLASH2.5/object'
>>>>>>gmake: *** [default] Error 2
>>>>>>
>>>>>>As can be seen the majority of the problms are with the c code, and
>>>>>>there
>>>>>>is a correlation between the functions that cause warnings early on
>>>>>>and
>>>>>>the undefined reference problems later. I thought that maybe the
>>>>>>problem
>>>>>>could be to do with name mangling but I am not sure, and if it is I
>>>>>>am not
>>>>>>entirely sure how to fix the problem.
>>>>>>
>>>>>>Any help is greatly appreciated.
>>>>>>
>>>>>>Steve Spreckley
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>-----------------------------------------------------------------------
>>>>>>
>>>>>>Steve
>>>>>>Spreckley
>>>>>>Astrophysics and Space Research Group School
>>>>>>of Physics and Astronomy The University
>>>>>>of Birmingham Edgbaston,
>>>>>>Birmingham, UK Tel: +44 (0)1214158410 B15
>>>>>>2TT E-mail: sas@star.sr.bham.ac.uk
>>>>>>-----------------------------------------------------------------------
>>>>>>
>>>>>
>>>>--
>>>>===================================================================
>>>> Dr Georgi B. Pavlovski School of Physics and Astronomy
>>>> Tel: +44-(0)-23-8059-2446 University of Southampton
>>>> Fax: +44-(0)-23-8059-3910 Southampton SO17 1BJ, U.K.
>>>> Email: gbp@phys.soton.ac.uk Web: www.astro.soton.ac.uk/~gbp
>>>>===================================================================
>>>>
>>>
>>>
>>!DSPAM:4342bb04210382220925225!
>>
>>
Received on Tue Oct 4 13:37:22 2005