Re: [ESPResSo-users] Question about the two compilation methods given in the manual

[Christoph Junghans]

Hi Axel,

2012/7/5 Axel Arnold <address@hidden>:
> Hi!
>
> I think the answer is actually very short - it is a bug in the tar file.
You are 100% right, I reported this issue as bug #36341 a while ago:
<http://savannah.nongnu.org/bugs/?36431>

Christoph

>
> For some reason, the latest tar.gz contains a myconfig-final.h in the src
> directory with a lot of features active. If you compile in the source
> directory, of course it gets overwritten, but not in a build directory. And
> unfortunately, the source directory has priority over the build directory, so
> you end up with the wrong file there.
>
> @Olaf: Looks like the set that is left over is myconfig-maxset.h, can be that
> it is left from Jenkins testing?
>
> Regards,
> Axel
>
> On Thursday 05 July 2012, Olaf Lenz wrote:
>> Hi Kai Yang!
>>
>> Thanks for the very cleanly written problem report! It's a pleasure to
>> reply to such a report - it contains all required information, and there
>> are no demands. I hope that the speed and length of the reply honor this.
>>
>> On 07/05/2012 10:29 AM, Kai Yang wrote:
>> >> code_info
>> >
>> > ESPResSo-3.1.0 { Compilation status { FFTW } { CONSTRAINTS } {
>> > COMFIXED } { NPT } { COMFORCE } { PARTIAL_PERIODIC } { LENNARD_JONES
>> >
>> >  } { EXTERNAL_FORCES } { TABULATED } { BUCKINGHAM } { SOFT_SPHERE } {
>> >  LJCOS } { LENNARD_JONES_GENERIC } { MASS } { BOND_ANGLE_COSINE } {
>> >
>> > ELECTROSTATICS } { LJCOS2 } { MPI_CORE } { MORSE } { FORCE_CORE } {
>> > EXCLUSIONS } }
>>
>> This is the default set of features and is what is supposed to be the
>> outcome in both cases.
>>
>> >> code_info
>> >
>> > ESPResSo-3.1.0 { Compilation status { FFTW } { CONSTRAINTS } {
>> > ROTATIONAL_INERTIA } { BOND_ENDANGLEDIST_HARMONIC } { GAY_BERNE } {
>> > BOND_ENDANGLEDIST } { COMFIXED } { INTER_RF } { NPT } { COMFORCE } {
>> >
>> >  LB_BOUNDARIES } { HERTZIAN } { BOND_ANGLEDIST_HARMONIC } {
>> >
>> > PARTIAL_PERIODIC } { LENNARD_JONES } { LJ_ANGLE } { OVERLAPPED } {
>> > MOLFORCES } { BMHTF_NACL } { MODES } { EXTERNAL_FORCES } { TABULATED
>> >
>> >  } { BUCKINGHAM } { LANGEVIN_PER_PARTICLE } { SOFT_SPHERE } { LJCOS }
>> >  { LENNARD_JONES_GENERIC } { MASS } { SMOOTH_STEP } {
>> >
>> > BOND_ANGLE_COSINE } { VIRTUAL_SITES_RELATIVE } { ELECTROSTATICS } {
>> > LJCOS2 } { BOND_ANGLEDIST } { LB } { BOND_VIRTUAL } { NEMD } {
>> > LB_ELECTROHYDRODYNAMICS } { MORSE } { DIPOLES } { ROTATION } {
>> > COLLISION_DETECTION } { EXCLUSIONS } }
>>
>> This is not what should happen, and I'm very surprised, as it contains a
>> number of very uncommon features.
>>
>> Usually, this should only happen if out of some reason the build system
>> found a wrong myconfig.h.
>>
>> The build system looks for the myconfig.h-files in the following
>> locations, in that order:
>>
>>    $builddir/myconfig.h
>>    $srcdir/myconfig.h
>>    $srcdir/src/myconfig-default.h
>>
>> Does any of these files exist and contain the features that you saw in
>> the build?
>>
>> If not, could you please check the file $builddir/src/myconfig-final.h?
>> That is the config file that is actually used for the build, and it
>> should be a copy of the file above.
>>
>> If that doesn't help, please send me the file "config.log" that you
>> should find in the build directory.
>>
>> > The information given by two methods is different! Why does this
>> > difference produce? Will the compile methods influence the simulation
>> > results?
>>
>> If you do not specify your own myconfig.h, both ways of compiling
>> ESPResSo should in theory generate the same binary with the same default
>> features (that you can see in $srcdir/src/myconfig-default.h). In your
>> case, there seems to be some kind of problem.
>>
>> Simulations should usually not produce different results when different
>> features are turned on. In the worst case, the simulation does not run
>> at all, or it runs slower.
>>
>> > Which compile methods should I choose?
>>
>> I always suggest people to use out-of-source builds, as this allows you
>> to compile various binaries with different features and compile flags
>> from the same sources. However, if you just plan a single build as you
>> are going to use a single set of features anyway, an in-source build
>> should suffice.
>>
>> Olaf
>
>
> --
> JP Dr. Axel Arnold      Tel: +49 711 685 67609
> ICP, Universität Stuttgart      Email: address@hidden
> Pfaffenwaldring 27
> 70569 Stuttgart, Germany
>



-- 
Christoph Junghans
Web: http://www.compphys.de