cgns / cgns

The CFD General Notation System (CGNS) provides a standard for recording and recovering computer data associated with the numerical solution of fluid dynamics equations. All development work and bug fixes should be based off the 'develop' branch, CGNS uses the branching model Gitflow.

CGNS

About

The CFD General Notation System (CGNS) provides a standard for recording and recovering computer data associated with the numerical solution of fluid dynamics equations.

Bugs/Feature and issue tracking

https://cgnsorg.atlassian.net

Installation

Installation Instructions using cmake

1. Install HDF5 on your system.

   • HDF5 can use the standard GNU autotools, so ./configure, make, sudo make install should install HDF5 without problems on most systems.

2. Unpack the tar ball containing the source code into some directory.

3. Create a new directory in which to build the library.

4. Use cmake to initialize the build tree.

   user@hostname:build_path$ cmake /path/to/cgns/sources/

5. Use ccmake to edit the control variables as needed.

   user@hostname:build_path$ ccmake .

   • The path to the HDF5 library should be specified with CMAKE_PREFIX_PATH=$HDF_DIR for linking with a specific HDF5 version.
     • If HDF5 is built with parallel-IO support via MPI, the HDF5_NEED_MPI flag must be set to true.
     • If HDF5 is built with zlib and szip support, these need to be flagged with HDF5_NEED_ZLIB and HDF5_NEED_SZIP as well as the paths for those libraries.
   • Fortran can be enabled by toggling the CGNS_ENABLE_FORTRAN variable.
     • A view of the attempt to autodetect the correct interface between Fortran and C is show, setting the value of FORTRAN_NAMING.
     • For gfortran and pgf90 the value of FORTRAN_NAMING should be LOWERCASE_.
   • The build system must be reconfigured after variable changes by pressing c. Variables who's value has changed are marked with a * in the interface.
   • After configuration, the Makefiles must be generated by pressing g.

6. Use make to build the library.

   user@hostname:build_path$ make

   • A colorized review of the build process should follow.

7. Installation of the library is accomplished with the install target of the makefile.

   user@hostname:build_path$ make install

   • You must have permissions to alter the directory where CGNS will be installed.

Installation Instructions using make

1. Install HDF5 on your system.
   • HDF5 can use the standard GNU autotools, so ./configure, make, sudo make install should install HDF5 without problems on most systems.
2. Typically the standard ./configure, make, make install will suffice.
3. Sample scripts for building parallel CGNS can be found in src/SampleScripts.

Usage

License

The distribution and use of the CGNS software is covered by the following license:

---

This software is provided 'as-is', without any express or implied warranty. In no event will the authors be held liable for any damages arising from the use of this software.

Permission is granted to anyone to use this software for any purpose, including commercial applications, and to alter it and redistribute it freely, subject to the following restrictions:

1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required.

2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.

3. This notice may not be removed or altered from any source distribution.

---

This license is borrowed from the zlib/libpng License:

http://www.opensource.org/licenses/zlib-license.php

and supersedes the GNU Lesser General Public License (LGPL) which previously governed the use and distribution of the software.

For details on the policy governing the distribution of the CGNS standard and software see:

http://www.grc.nasa.gov/www/cgns/charter/principles.html

Development

CGNS uses the branching/release model as summarized at:

http://nvie.com/posts/a-successful-git-branching-model/

cgns's Issues

CMake needs to be updated to handle new fortran features.

Since Fortran source was added to the library, CMake needs to be updated to build a separate Fortran project and library.

[SOURCEFORGE-9] -lX11 not included in OS X

The linker does not include the -lX11 on OS X in the definition for OGLLIBS.

Current values are

OGLLIBS = -lGL -lGLU

but should be

OGLLIBS = -lGL -lGLU -lX11

Reported by: fommil

[CGNS-15] Implement travis CL testing on github

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Wed, 23 Sep 2015 10:29:04 -0500
• Resolved at: Tue, 2 Feb 2016 17:13:05 -0600

get automatic testing working with travis CL on github for both linux and mac.

[CGNS-18] Test cgiof_F03 fails

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Wed, 23 Sep 2015 13:31:47 -0500
• Resolved at: Wed, 29 May 2019 16:00:43 -0500

Test fails on linux 32-bit machine with intel 15.0.3

[CGNS-14] Github CGNS version problem in parallel

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Wed, 3 Jun 2015 14:59:01 -0500

The Github version has problem when the CGNS is compiled for parallel and used in a parallel code, and when either of the following functions are called:

• cg_open
• cg_is_cgns

Note, that I made sure these files are called only on the master CPU. What happens is that the code easily runs when lunched with a single CPU, but with 2 or more CPUs, the master CPU gets glinched in these functions and never comes out.

The library on SourceForge doesn't seem to have this problem.

[CGNS-19] Test writing/reading large number of zones in parallel

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Wed, 23 Sep 2015 13:39:43 -0500
• Resolved at: Thu, 7 Jan 2016 10:20:32 -0600

From Forum:

Hi,

In case it's of use to anyone I got a working code that writes and reads
a large number of zones in parallel from CGNS. I've attached it to this
e-mail. I ended up writing the base and zone metadata from the rank
zero process, then using all the processes to define the co-ordinate
entries for all the zones, then writing and reading the specific zone
data from only the process that wanted that zone.

Note, the performance I get using 512 processes and 16384 zones divided
equally between them isn't great. It's about 100 times slower to read
the data than the code that does it with raw MPI-I/O. However, it does
complete and give the correct results. It may be the way we've setup
CGNS on our machine, or the way I've written the code that's causing
such slow performance (I should further note the writing of the data
takes even longer); any suggestions on performance improvements (other
than the excellent description I had earlier of why intra-zone I/O
currently performs better than many-zone I/O) would be welcome. I
tested this on 3.2.1, I've not tested on the beta release yet, I'll do
that once we've installed it.

Hope this is helpful.

thanks

adrianj

p.s. The code is not brilliant code, it's a sort of f77/f90 mix I'm not
proud of, in my defence it's written like this to integrate with an
existing application code that's the real reason I'm doing this
work....Honest, it's not my fault

original.zip

[SOURCEFORGE-19] configure should handle FC

Currently, configure only handles the env variable F77, but most people setenv FC or F90. Configure should accept FC env. variable.

*also the compiler flags FCFLAGS.

F77 is a misnomer.

Reported by: brtnfld

configure fails when --with-hdf5 is not set

On some systems that use system modules to load packages, the env. variables are usually set correctly to find the needed library. Therefore, we should not require the variable to be set (for example, --with-hdf5 -...). Make fails when these variables are not set.

Work around is to search around for where the needed libraries are installed and setting the paths at configure time.

[CGNS-16] Create build+test script for testing on HDF5 machines

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Wed, 23 Sep 2015 11:09:28 -0500
• Resolved at: Tue, 2 Feb 2016 17:14:29 -0600

Used script to test on hdf machines until an alternative place is found for testing.

[SOURCEFORGE-1] cgnslib2.2 compile error

I am using FreeBSD, gcc version 2.95.4

Here are the errors:

In file included from ../cg_ftoc.c:4:
../fortran_macros.h:566: syntax error before `need'
../fortran_macros.h:588: syntax error before `*'
../fortran_macros.h:588: warning: data definition has
no type or storage class
../fortran_macros.h:651: syntax error before `('
../fortran_macros.h:651: `FMNAME' declared as function
returning a function
../fortran_macros.h:651: warning: parameter names
(without types) in function declaration
../fortran_macros.h:652: syntax error before `icstr'
../fortran_macros.h:664: syntax error before `icstr'
../cg_ftoc.c:19: syntax error before `('
../cg_ftoc.c:19: `FMNAME' declared as function
returning a function
../cg_ftoc.c:19: warning: parameter names (without
types) in function declaration
../cg_ftoc.c:25: `filename' undeclared here (not in a
function)
../cg_ftoc.c:25: initializer element is not constant
../cg_ftoc.c:25: warning: data definition has no type
or storage class
../cg_ftoc.c:26: conflicting types for `c_name'
../cg_ftoc.c:23: previous declaration of `c_name'
../cg_ftoc.c:26: warning: initialization makes integer
from pointer without a cast
../cg_ftoc.c:26: initializer element is not constant
../cg_ftoc.c:26: warning: data definition has no type
or storage class
../cg_ftoc.c:28: syntax error before `('
../cg_ftoc.c:28: conflicting types for
`ADFI_string_2_C_string'
../adf/ADF_internals.h:639: previous declaration of
`ADFI_string_2_C_string'
../cg_ftoc.c:28: warning: parameter names (without
types) in function declaration
../cg_ftoc.c:28: syntax error before `&'
*** Error code 1

Stop in /usr/apps/cgns/cgnslib_2.2.

Reported by: bijah

[SOURCEFORGE-3] Missing cg_elements_partial_read_f in cgnslib_2.5.2

The 2.5.2 version doesn't have cg_elements_partial_read_f.

Reported by: *anonymous

[CGNS-17] CGNS autoconfig needs to be upgraded from 2.13

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Wed, 23 Sep 2015 11:57:05 -0500
• Resolved at: Thu, 4 Feb 2016 14:25:28 -0600

configure.in should be upgraded to the newest autoconfig version.

[SOURCEFORGE-8] OpenGL includes not included on OS X

After running the ./configure script, the /usr/X11/include directory is not being included on OS X - this results in all OpenGL related files failing to compile. Workaround is to edit the CFLAGS to have -I/usr/X11/include

Reported by: fommil

[SOURCEFORGE-10] File is being opened as read/write when it is not needed

In ADF_interface.c line 1048 of the 2.5.4 version of the CGNS library the files are being opened as r+b when there is no need to write the file. This is causing old files not to be loaded as the previous libraries did not require write permision

Reported by: *anonymous

[CGNS-13] cgnsview fails to recognize files built in parallel

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Tue, 5 May 2015 19:10:54 -0500
• Resolved at: Tue, 22 Sep 2015 11:48:24 -0500

Fix:
src/cgnstools/cgnsview/cgiotcl.c
@@ -1032,7 +1032,7 @@ static int CGNSfile (ClientData data, Tcl_Interp *interp,
return (get_error (interp, "cgio_check_file"));
if (file_type == CGIO_FILE_ADF)
Tcl_SetResult (interp, "adf", TCL_STATIC);

• else if (file_type == CGIO_FILE_HDF5)
  + else if (file_type == CGIO_FILE_HDF5 || file_type == CGIO_FILE_PHDF5)
  Tcl_SetResult (interp, "hdf5", TCL_STATIC);
  else {
  Tcl_SetResult (interp, "unknown file type", TCL_STATIC);

cgnstools fail with tk/tcl 8.6

I get a segfault on Debian 8. Workaround is to use tk/tcl8.5 via

-DTCL_INCLUDE_PATH:PATH=/usr/include/tcl8.5 -DTCL_LIBRARY:FILEPATH=/usr/lib/x86_64-linux-gnu/libtcl8.5.so -DTK_INCLUDE_PATH:PATH=/usr/include/tcl8.5 -DTK_LIBRARY:FILEPATH=/usr/lib/x86_64-linux-gnu/libtk8.5.s

with cmake.

[SOURCEFORGE-2 Feature Request] install process

We should add an install process.
I suggest to use common things such as:

configure --exec-prefix=<target-directory>

add install a tree like:

<target-directory>/include/CGNS/*.h
<target-directory>/lib/*.a *.so *.sl etc...

We only require exec-prefix because the installation
has no real platform indepedant stuff. Well, headers
are, but I looks more complicated to define two
options.

-MP-

Reported by: poinot

[SOURCEFORGE-21] [AUTOTOOLS] make recompiles all the scr/tools files

make recompiles all the scr/tools files even when there are no code changes.

Reported by: brtnfld

[SOURCEFORGE-20] [CMAKE] in-source builds

CMAKE should not allow for in-source builds, but should instead issue an error message and stop.

Reported by: brtnfld

[SOURCEFORGE-7] syntax error in p3dout.c

The multi-line macro beginning on line 22 of p3dout.c is missing an end bracket. It currently reads (ignoring whitespace)

fprintf (stderr, "Fortran unformatted output not supported for %s machine\n", bf_machname (MACH_LOCAL);

but should read

fprintf (stderr, "Fortran unformatted output not supported for %s machine\n", bf_machname (MACH_LOCAL));

Reported by: fommil

[CGNS-6] 3.3.0alpha release on github

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Mon, 21 Sep 2015 11:50:46 -0500
• Resolved at: Mon, 21 Sep 2015 16:13:44 -0500

[SOURCEFORGE-2] Get the wrong version info

This is happened at line 380 of cgnslib.c (version 2.2.1):
cg->version = (int)(1000*(*FileVersion));

if *FileVersion=2.09999999999999999999999999
cg->version will be 2099 not 2.100

the correct way should be
cg->version = (int)(1000*(*FileVersion)+0.5);

Please correct it.

Reported by: fzhu88

[SOURCEFORGE-23] cgnslib_f.h compatibility with extended source f77 files

The current cgnslib_f.h does not work (at least with ifort) with extended source (80 or 132 length) f77 programs as the line break characters for the data declarations are not beyond the end of the extended source line, i.e.:

    data MassUnitsName /'Null','UserDefined','Kilogram','Gram',     &
 &                      'Slug','PoundMass'/

The & at the end of the first line causes compilation errors if included with files that are using extended source.

This can be fixed by moving the & character further down the line so it is beyond the 132 character. This should allow it to compile with all line length fixed source and also with free source FORTRAN programs.

Reported by: ajackso3

[SOURCEFORGE-6] Tcl framework not referenced in OS X build

On OS X, in the file make.defs, the definition for the TKLIBS variable begins

TKLIBS = -framework Tk -framework Tk ...

which results in a failed compile. I believe a simple solution is to replace this with

TKLIBS = -framework Tk -framework Tcl

(i.e. the second framework is Tcl, not a repeat of Tk)

Reported by: fommil

[CGNS-12] cgnstools fail with tk/tcl 8.6

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Tue, 5 May 2015 19:09:06 -0500
• Resolved at: Mon, 21 Sep 2015 16:17:27 -0500

I get a segfault on Debian 8. Workaround is to use tk/tcl8.5 via

-DTCL_INCLUDE_PATH:PATH=/usr/include/tcl8.5 -DTCL_LIBRARY:FILEPATH=/usr/lib/x86_64-linux-gnu/libtcl8.5.so -DTK_INCLUDE_PATH:PATH=/usr/include/tcl8.5 -DTK_LIBRARY:FILEPATH=/usr/lib/x86_64-linux-gnu/libtk8.5.s

with cmake.

[CGNS-3] Contact atlassian to get opensource license.

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Mon, 21 Sep 2015 11:47:29 -0500
• Resolved at: Mon, 21 Sep 2015 16:16:40 -0500

[CGNS-1] Need to test in cmake/configure Parallel is enabled in HDF5

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Mon, 21 Sep 2015 11:45:48 -0500
• Resolved at: Thu, 13 Jan 2022 10:33:26 -0600

When building CGNS with parallel, we need to make sure the HDF5 has parallel I/O enabled.

[CGNS-9] remove obsolete functions parallel

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Mon, 21 Sep 2015 11:52:55 -0500
• Resolved at: Thu, 19 Nov 2015 09:23:41 -0600

remove flush routines.

[SOURCEFORGE-17] cgp_mpi_comm passes wrong type

using openmpi 1.8.4 and gfortran,gcc on the Mac cgp_mpi_comm segfaults.

ier = cgp_mpi_comm(int mpicomm);
should be
ier = cgp_mpi_comm(MPI_Comm mpicomm);

In the calls made by cgp_mpi_comm, we should not be converting MPI_comm variables to size_t and then back again to MPI_Comm.

We also need to check the warnings passing MPI_Comm from fortran (which is an integer) to C.

/Users/brtnfld/packages/cgns/trunk/src/cg_ftoc.c:3643:17: warning: cast to 'MPI_Comm' (aka 'struct ompi_communicator_t *') from smaller integer type
'int' [-Wint-to-pointer-cast]
mpi_comm_c = MPI_Comm_f2c((int)*mpi_comm_f);
^
/Users/brtnfld/packages/cgns/trunk/src/cg_ftoc.c:35:28: note: expanded from macro 'MPI_Comm_f2c'
#define MPI_Comm_f2c(comm) (MPI_Comm)(comm)
^
/Users/brtnfld/packages/cgns/trunk/src/cg_ftoc.c:3655:20: warning: cast to 'MPI_Info' (aka 'struct ompi_info_t *') from smaller integer type 'int'
[-Wint-to-pointer-cast]
pcg_mpi_info_c = MPI_Info_f2c((int)*pcg_mpi_info_f);
^
/Users/brtnfld/packages/cgns/trunk/src/cg_ftoc.c:36:28: note: expanded from macro 'MPI_Info_f2c'
#define MPI_Info_f2c(info) (MPI_Info)(info)

Reported by: brtnfld

[CGNS-10] CMake needs to be updated to handle new fortran features.

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Tue, 28 Apr 2015 11:35:22 -0500
• Resolved at: Thu, 23 Feb 2017 12:01:33 -0600

Since Fortran source was added to the library, CMake needs to be updated to build a separate Fortran project and library.

[SOURCEFORGE-5] ADFI_string_2_C_String oversteps character string size

This isn't a bug in ADFI_String_2_C_String per say, but rather many of the routines that call it. What happens is that this function is being passed character strings which are smaller than the specified string_length parameter and the internal for loop is accessing memory beyond the size of the passed in string argument. In many places you are calling cgi_new_node(...) with a fixed character string of "R4" for instance which is turn may call cgio_set_dimensions(...), eventually calling ADFI_string_2_C_string(...) with a value of string_length equal to ADF_DATA_TYPE_LENGTH, but you passed in a character constant of length 2, so you access memory outside the bounds of string. So to repeat, the call path is:

cgi_new_node -> cgio_set_dimensions -> ADF_Put_Dimension_Information -> ADF_evaluate_datatype -> ADFI_string_2_C_string

If cgi_new_node is called with data_type="R4" or some such, this will be a problem as you will exceed the memory region of string in ADFI_string_2_C_string(...) when performing the copy. An easy fix is to not use a character constant in the actual calls, but to define a ADF_DATA_TYPE_LENGTH length character constant in the header for things such as "I4", "R4", etc... Also, they should be used in the cgi_adf_datatype(...) function as well. Thanks.

Reported by: skinny_boy

[SOURCEFORGE-1 Feature Request] Shared libs

Add shared libs to production.

On SGI, the command to add looks like:

cc -shared -o libcgns.so *.o

But it certainly has to be cleverly set for
other platforms. I guess the best is to see
what is done for other open source packages
and it up in the configure stuff.

-MP-

Reported by: poinot

[SOURCEFORGE-1 Support Request] CRAY X1

I'm trying to compile CGNS 2.3 and 2.4 on the ARSC Cray
X1, Klondike. I get errors in "fortran_macros.h" even
though I'm configuring with the --without-fortran flag
and --with-system=CRAY. Has anyone compiled this
library on a Cray X1?

jlovric at u.washington.edu

Reported by: *anonymous

[SOURCEFORGE-14] Cmake failed in cgnslib_3.1.4-2

Hello,

I try to compile cgnslib 3.1.4-2 with cmake but it failed.
I have exactly the same problem with the first cgnslib 3.1 release.
I use cmake 2.6-patch4 and gcc 4.1.2 and Linux Redhat RHEL 5.4

#cmake --version
cmake version 2.6-patch 4

Installation steps :
cd /home/user
tar xzf cgnslib_3.1.4-2.tar.gz
cd cgnslib_3.1.4
cmake /home/webadm//cgnslib_3.1.4/src

-- The C compiler identification is GNU
-- The CXX compiler identification is GNU
-- Check for working C compiler: /usr/bin/gcc
-- Check for working C compiler: /usr/bin/gcc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
CMake Error at CMakeLists.txt:36 (if):
if had incorrect arguments: ${SIZE_OF_LONG} GREATER 4 (Unknown arguments
specified).

CMake Warning (dev) in CMakeLists.txt:
No cmake_minimum_required command is present. A line of code such as

cmake_minimum_required(VERSION 2.6)

should be added at the top of the file. The version specified may be lower
if you wish to support older CMake versions for this project. For more
information run "cmake --help-policy CMP0000".
This warning is for project developers. Use -Wno-dev to suppress it.

-- Configuring incomplete, errors occurred!

Can you fix the the error :
CMake Error at CMakeLists.txt:36 (if):
if had incorrect arguments: ${SIZE_OF_LONG} GREATER 4 (Unknown arguments
specified). ?

Or can you explain me a workaround to bypass this error ?

Best Regards.
Thanks.

Reported by: yannick-56

[CGNS-7] Send release annoucement of CGNS 3.3.0 alpha

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Mon, 21 Sep 2015 11:51:21 -0500
• Resolved at: Mon, 21 Sep 2015 16:11:02 -0500

cgnsview fails to recognize files built in parallel

Fix:
src/cgnstools/cgnsview/cgiotcl.c
@@ -1032,7 +1032,7 @@ static int CGNSfile (ClientData data, Tcl_Interp *interp,
return (get_error (interp, "cgio_check_file"));
if (file_type == CGIO_FILE_ADF)
Tcl_SetResult (interp, "adf", TCL_STATIC);

• else if (file_type == CGIO_FILE_HDF5)
• else if (file_type == CGIO_FILE_HDF5 || file_type == CGIO_FILE_PHDF5)
  Tcl_SetResult (interp, "hdf5", TCL_STATIC);
  else {
  Tcl_SetResult (interp, "unknown file type", TCL_STATIC);

[SOURCEFORGE-16] cgnsutil.c:1747: bad loop ?

cgnslib_3.2.1/src/cgnstools/utilities/cgnsutil.c:1747:44: warning: iteration 2u invokes undefined behavior [-Waggressive-loop-optimizations]

       float exp[5];
        for (n = 0; n < 5; n++)
            exp[n] = (float)f->dataconv[n];

but

double dataconv[2];

Reported by: dcb314

[SOURCEFORGE-18] [CMake] error for lseek64 and open64

I think there is an error in CMakeLists.txt on line 84 of the trunk version:

81 if (HAVE_OPEN64)
82 add_definitions(-DHAVE_LSEEK64)
83 endif (HAVE_OPEN64)
84 if (HAVE_OPEN64)
85 add_definitions(-DHAVE_LSEEK64)
86 endif (HAVE_LSEEK64)

it should be:

81 if (HAVE_OPEN64)
82 add_definitions(-DHAVE_LSEEK64)
83 endif (HAVE_OPEN64)
84 if (HAVE_LSEEK64)
85 add_definitions(-DHAVE_LSEEK64)
86 endif (HAVE_LSEEK64)

Reported by: brtnfld

[SOURCEFORGE-15] gfortran compatibility cgnslib_f.h

CGNS V3.2.1, compiled with gcc compiler (version 4.6.3 ). Gfortran compiler has problems with including the cgnslib_f.h, since header comments are in f77 style (c at the beginning). Can be fixed by replacing c with ! in cgnslib_f.h.in (! is actually used for all other comments in this file)

Reported by: hindenlang

[CGNS-2] Set up cloud atlassian project

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Mon, 21 Sep 2015 11:46:47 -0500
• Resolved at: Tue, 2 Feb 2016 17:12:44 -0600

Set-up cloud atlassian project so that we can apply for a Open-source license.

[SOURCEFORGE-13] CGNS fails to create node when using multiple files

When CGNS is configured with HDF5 I am unable to create a new node (e.g., write a base) when working with multiple files. It only occurs if I close one of the files while writing to the other. For example, to repeat the issue:

1. Open (new) file A for writing
2. Open file B as read only
3. Close file B
4. Write base to file A

When attempting the write to file A the following error is obtained: cgio_create_node:H5Gcreate:node group creation failed

This issue is not observed when using the ADF backend.

Attached are a sample code to reproduce the issue along with the file the sample code opens as read only.

Reported by: jgraha8

[SOURCEFORGE-4] Prameters in cg_gopath_f

There is a mismatch in parameters in cg_gopath_f.
Documentation says:
cg_gopath_f(fn, ier, path)

however, the actual (2.5r3) implementation is
cg_gopath_f(fn, path, ier)

Reported by: *anonymous

Github CGNS version problem in parallel

The Github version has problem when the CGNS is compiled for parallel and used in a parallel code, and when either of the following functions are called:

• cg_open
• cg_is_cgns

Note, that I made sure these files are called only on the master CPU. What happens is that the code easily runs when lunched with a single CPU, but with 2 or more CPUs, the master CPU gets glinched in these functions and never comes out.

The library on SourceForge doesn't seem to have this problem.

[SOURCEFORGE-11] Problem opening files that have the same name

Steps to reproduce:

1. Create two directories (let’s say dir A and dir B).
2. In dir A create two files (say, file 1 and file 2). File 1 is linked to file 2.
3. In dir B create two different files but give them the same name. File 1 is also linked to file 2.
4. Change the current working directory to one of the previously created directories, let’s say dir A.
5. Open dir A/file 1.
6. Try to open dir B/file 2. This step will fail.

Reported by: *anonymous

[CGNS-5] Update web docs about fortran Integer changes

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Mon, 21 Sep 2015 11:49:47 -0500
• Resolved at: Fri, 5 Feb 2016 15:10:32 -0600

need to update in git

[CGNS-8] update configure with mpi.

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Mon, 21 Sep 2015 11:52:13 -0500
• Resolved at: Tue, 22 Sep 2015 11:39:46 -0500

Absolutely - I will let you know how we get on with that compiler/MPI combination. BTW, in the source code config file of CGNS trunk you might want to add "impi" (Intel MPI lib) to the list of supported MPI libraries:

echo checking for MPI library ...
for %%l in ( mpi mpich libmpi mpid mpichd libmpid impi ) do (
if exist %mpidir%\lib%%l.lib (
echo %mpidir%\lib%%l.lib
set mpilibs=%mpidir%\lib%%l.lib
goto next
)
)

[SOURCEFORGE-22] test/cgiof.F fails due to fixed-format on some systems.

Need to add the flag for fixed format on mira (ANL, IBM Blue Gene/Q) to get the Fortran programs to compile. needed to specify -qfixed.

Reported by: brtnfld

[SOURCEFORGE-12] pcgns_util.c fails to compile

cgnslib_3.1.3 fails to compile in Ubuntu 11.10 with parallel support.

Compiler gives following error message:

/home/user/cgnslib_3.1.3/pcgns/src/pcgns_util.c: Function ”new_str_attb”:
/home/user/cgnslib_3.1.3/pcgns/src/pcgns_util.c:264:2: error: too many arguments to function ”H5Acreate1”
/usr/include/H5Apublic.h:105:16: note: declared here

and two similar errors follow.

With Ubuntu 11.10 HDF5 library is compiled with option --with-default-api-version=v16 as there are these two lines in libhdf5.settings:

...
Default API Mapping: v16
With Deprecated Public Symbols: yes
...

It seems that pcgns_util.c uses old functions by default even if it should use newer ones. With hard-coded to use function H5Acreate2 compiling finishes without errors.

Expected behaviour: Compiling finishes without errors.
Current behaviour: Wrong version from function seems to be selected for pcgns_util.c and build process fails.

Reported by: grinda

[CGNS-4] Fix CGNS and cmake on windows.

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Mon, 21 Sep 2015 11:48:26 -0500
• Resolved at: Tue, 15 Nov 2016 16:00:08 -0600

Build on windows with cmake does not work.

[CGNS-11] configure fails when --with-hdf5 is not set

This issue has been imported from JIRA. Read the original ticket here.

• Created at: Tue, 28 Apr 2015 11:40:07 -0500

On some systems that use system modules to load packages, the env. variables are usually set correctly to find the needed library. Therefore, we should not require the variable to be set (for example, --with-hdf5 -...). Make fails when these variables are not set.

Work around is to search around for where the needed libraries are installed and setting the paths at configure time.