open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi all,

I noticed that time series from result files create on linux are not in a format recognized by Blue Kenue:

-001/-1/-1 -1:-1:-1.000 0.120000
...

I use the defaut date formats:
ORIGINAL DATE OF TIME = 1900;1;1
ORIGINAL HOUR OF TIME = 0;0;0

The files can be opened with fudaa and python and has the correct time records (in seconds, not formatted).

Same model run on Windows doesn't give any format problem in Blue Kenue.

I assume it has to do with linux/unix encoding but have no further idea. Does someone have a trick?

Thanks in advance!

Best regards
PL

Hi,
Can you post a small example slf file?
Cheers... Martin

Hi Martin,

Here it is.

Thanks for your help,
PL

Hi Pierre-Louis

As an idea, maybe linked to the config file (Big_endian/little endian)?

Regards

Hi Christophe,

Thanks for your remark. It looks like the guys also used big endian type in the config:
www.pdc.kth.se/software/software/TELEMAC...ow/v7p1r1/index.html

However the following tag is missing, can it be that?
-ffixed-line-length-132

I also noticed that I can't chop (runSELAFIN.py - chop) large result files created on this cluster, but small files can be chopped... I haven't investigated that further but in case this can have something to do with this missing tag I can ask the support to recompile the program and see.

Kind regards
PL

Hi

The tag --fixed-line-length-132 should be here if you plan to use long lines in your fortran files.
Telemac files are formated with short line (72 characters) so this tag is not mandatory.

What I saw an the description of the computer is your running on a Cray Machine...
I think the problem is more related to this particular computer. I know in previous version, there was specific keywords dedicated to those kind of computer...
I couldn't help you further on this problem as I never had an access to a cray machine

Regards

Hi,

Here is a description of the cluster, for your information:
www.pdc.kth.se/hpc-services/computing-systems

I was running on Hartree Centre's cluster (Daresbury, enCORE) by the past and got similar date format issues. I don't know which kind of machine it was though.

PL

Hello,

I've noticed that this is a longstanding bug, since V7p1r0 at least, I thought it was related to something missing when compiling in Linux and I recall someone mentioning that it was fixed for v7p2rx...

Regards

Hi all,

Except for the messed up date field this appears to be a bog standard Big Endian
single precision file.

The attached image shows the HEX contents of the 6 4byte date fields.
There is no way that these values represent 1990;1;1;0;0;0
I therefore conclude that there is something messed up in the code that writes the slf file header and it's a TELEMAC bug.

Cheers... Martin

Is it possible that these variables are not explicitly initialized and therefore just contain random values under some compilers?
I ask because I just came across another slf file that user
"EGleirscher" sent to me 15 April 2016 with the same problem.

The title of the post is "Blue Kenue timeline problem v7p1r0 in parallel"

This file had equally strange but different values.