Problems in 7.11.3  [SOLVED]

Moderator: pouillon

Locked
abokhanc
Posts: 24
Joined: Tue Jul 29, 2014 4:28 pm

Problems in 7.11.3

Post by abokhanc » Thu Nov 13, 2014 4:40 pm

Is anyone else having issues compiling src/70_gw in the 7.11.3 version?

I branched from my private and public branches, and when I compile I get the following:

Code: Select all

../../../src/70_gw/gwls_lineqsolver.F90(298): error #5082: Syntax error, found IDENTIFIER 'ENDIF' when expecting one of: ( * ) :: , <END-OF-STATEMENT> ; + . - (/ [ : ] /) ' ** / // > ...
&"        endif
----------^
../../../src/70_gw/gwls_lineqsolver.F90(298): error #6404: This name does not have a type, and must have an explicit type.   [ENDIF]
&"        endif
----------^
../../../src/70_gw/gwls_lineqsolver.F90(297): warning #6043: This Hollerith or character constant is too long and cannot be used in the current numeric context.   ['=# of valence states).']
             write(std_out,*) "              subspace (the kernel contains state i=",min_index," > ",nbandv,"=# of valence states).&
------------------------------------------------------------------------------------------------------------^
../../../src/70_gw/gwls_lineqsolver.F90(290): error #6321: An unterminated block exists.
  if (singular .and. ( (project_on_what==1 .and. (min_index > nbandv)) .or. project_on_what==0 ))  then
^
compilation aborted for ../../../src/70_gw/gwls_lineqsolver.F90 (code 1)
make[3]: *** [gwls_lineqsolver.o] Error 1
make[3]: Leaving directory `/home/stud2/7.11.3-private/build/src/70_gw'
make[2]: *** [all-recursive] Error 1
make[2]: Leaving directory `/home/stud2/7.11.3-private/build/src'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/stud2/7.11.3-private/build'
make: *** [all] Error 2


I looked over the file but I don't see anything wrong with it.

abokhanc
Posts: 24
Joined: Tue Jul 29, 2014 4:28 pm

Re: Problems in 7.11.3

Post by abokhanc » Thu Nov 13, 2014 6:36 pm

So I think it is indeed my compiler, since the same branch I'm trying to compile here, passed all the intel tests during the nightly test. It looks like the version that is installed here is 11.0.074.

Jordan
Posts: 282
Joined: Tue May 07, 2013 9:47 am

Re: Problems in 7.11.3

Post by Jordan » Fri Nov 14, 2014 9:53 am

The line is 132 charaters (max value allowed by abirules and default value use by gfortran to read lines) what might be too long for your compiler according to the options you used.
You can check that you use the correct flag to enable "infinite" line width (e.g -ffree-line-length-none with gfortran)
Or, a quick fix (maybe the best solution) is to cut the line into two so the problem does not happen in any case :-)

Cheers

Jordan

User avatar
jbeuken
Posts: 364
Joined: Tue Aug 18, 2009 9:24 pm
Contact:

Re: Problems in 7.11.3

Post by jbeuken » Fri Nov 14, 2014 9:19 pm

Hi Anton,

the ifort versions < 12.1.4 are no longer officially supported for ABINIT

for info, the following compilers are either obsolete : gcc < 4.7 , xlf < 14.0, ifort < 12.1.4
or considered too unstable/buggy for ABINIT : g95, pathscale, open64 .
This does not mean that ABINIT will not work, but there is no guarantee for ABINIT to run correctly

jmb
------
Jean-Michel Beuken
Computer Scientist

jolafc
Posts: 34
Joined: Wed Aug 26, 2009 5:37 am

Re: Problems in 7.11.3

Post by jolafc » Mon Nov 17, 2014 3:00 pm

I just modified the problematic file along with all GWLS related files so that the maximum line length is now 130 chars; which should ensure that a similar problem does not happen again. Here is my branch, if you want to try that :

[url]forge.abinit.org/abinit/jolafc/7.11.3-private/[/url]

abokhanc
Posts: 24
Joined: Tue Jul 29, 2014 4:28 pm

Re: Problems in 7.11.3

Post by abokhanc » Mon Nov 17, 2014 3:43 pm

Thanks for your replies.

Now here is another problem. During the nightly build, I got warnings by ROBODOK for not having "!!***" at the end of my routines. I took them out on purpose because it seems that ABILINT doesn't like them:

Code: Select all

Fatal Error: 
'!!***'
--> No detected routine!
This part of code can not be parsed as Fortran file:
Analysis Error in ./src/84_fold2Bloch/getargs.F90


I can't find anything different from other F90 files, like in CUT3D, so I attached getargs.F90 so maybe one of you can tell me what I'm doing wrong.

Thank you very much,

Anton
Attachments
getargs.F90
(4.45 KiB) Downloaded 377 times

Jordan
Posts: 282
Joined: Tue May 07, 2013 9:47 am

Re: Problems in 7.11.3

Post by Jordan » Wed Nov 19, 2014 11:46 am

Hi,

I don't understand what is wrong or where the problem is. I added the !!*** to the end of each of you files of your last revision. abilint does not complain and the make robodoc works fine as well. If you don't put this end marker, the robodoc built will fail for sure.

BTW, I think you know but you have a lot of "allocate", "deallocate" and "write(*,*)" statements which are not allowed in abinit. Your inca testbot will fail untill you correct that (Check also you dummy variables and unused variables - also forbidden).

Cheers

Jordan

abokhanc
Posts: 24
Joined: Tue Jul 29, 2014 4:28 pm

Re: Problems in 7.11.3

Post by abokhanc » Wed Nov 19, 2014 7:34 pm

Yes, I changed all the things that failed Abi Rules and Robodoc (added !!***). However, I just pushed the new revision, and it failed the pre-build tests:

Code: Select all

(view as text)
Project ABINIT (76 directories, 1292 files)
Add the libraries...(76 directories, 1292 files)done.
Analyze all the project...
Error[1]: Use own version of 'mpif.h'
Error[2]: No implicit statement in gwls_ComplementSpacePolarizability (./src/70_gw/gwls_ComplementSpacePolarizability.F90)
Error[3]: The include file 'fexcp.h' does not exist in the project!

Fatal Error:
'!!***'
--> No detected routine!
This part of code can not be parsed as Fortran file:
Analysis Error in ./src/84_fold2Bloch/getargs.F90

With the option --nofatal, abilint will always generate the files of interfaces (already processed).


So I am very confused why this happens if when you added !!*** everything went ok.

Jordan
Posts: 282
Joined: Tue May 07, 2013 9:47 am

Re: Problems in 7.11.3  [SOLVED]

Post by Jordan » Thu Nov 20, 2014 10:33 am

Well something is bizarre.
I merge with your last commit and I got the same issue. It seems to come from the fact you appended "getargs" after end subroutine.
If I remove everything after end subroutine and add again getargs then it works.

So I would say that you use a text editor that added some binary mistake that makes the script failed. But I guess you used the same text editor for all your files ... so I'm not sure to understand. I compared the files in hexa and I don't see any difference.

What editor do you use ? are you on mac/win/linux ? The newline character can be different like \r or \n or even \r\n.

The easiest fix is then to remove the line you think is correct and type it again. Make sure you are in UTF-8 and you should be safe.

The attached file is a "working" one.

Jordan
Attachments
getargs.F90
(4.46 KiB) Downloaded 378 times

abokhanc
Posts: 24
Joined: Tue Jul 29, 2014 4:28 pm

Re: Problems in 7.11.3

Post by abokhanc » Thu Nov 20, 2014 3:58 pm

You were right. I used the editor that comes with WinSCP (windows) that's when it didn't work, but when I changed the ending in Vim (linux) it compiled just fine. I never though that using a wrong editor could affect the results.

Thank you

Locked