19990827: decoding grib by XCD (cont.)

>From: "Jennie L. Moody" <address@hidden>
>Organization: UVa
>Keywords: 199907281535.JAA02200 McIDAS GRIB DMGRID

Jennie,

I just ran a quick test of decoding one of the grib files from
/incoming/data/grib back here at Unidata.  The file I used was:

us008_gf078_97091200_YxAAx

Just for future reference, what I did was:

o FTPed us008_gf078_97091200_YxAAx from aeolus to my mcidas76/workdata
  directory
o set file REDIRECTions for GRID5* and HRS.SPL to be in mcidas76/workdata:

  <start a McIDAS-X session>
  REDIRECT ADD GRID5* ".
  REDIRECT ADD HRS.SPL ".

o made a copy of xcd_run in the mcidas76/workdata directory

  cd mcidsa76/workdata
  cp ../bin/xcd_run .
  <edit ./xcd_run to setup MCHOME, etc.>

o set the pointer in GRIBDEC.PRO to the beginning of information:

  LWU POKE GRIBDEC.PRO 4096 0

o catted us008_gf078_97091200_YxAAx into ./xcd_run:

  cat us008_gf078_97091200_YxAAx | ./xcd_run HRS

o started the GRID decoder:

  DECINFO SET DMGRID A
  DMGRID

o watched as a GID file was created by the decoder:

  DMAP GRID5
  PERM      SIZE LAST CHANGED FILENAME DIRECTORY
  ---- --------- ------------ -------- ---------
  -rw-    711448 Aug 27 15:09 GRID5105 .
  711448 bytes in 1 files

  Did a listing of the GRID file:

  IGG LIST GRIDF=5105
   Grid file:  5105 **ALL  00Z MRF 0  HR<=VT<=24 HR   **
      #  YYDDD HHMMSS NAME  LEVEL  SRC   VT  NR  NC   LLNW    ROWINC  COLINC
   ----- ----- ------ ---- ------- ----  --  --  -- --------  ------  ------
       1 97255      0 P    MSL     MRF    0  65  65   -PS-
       2 97255      0 Z    1000 MB MRF    0  65  65   -PS-
       3 97255      0 Z    850  MB MRF    0  65  65   -PS-
       4 97255      0 Z    700  MB MRF    0  65  65   -PS-
       5 97255      0 Z    500  MB MRF    0  65  65   -PS-
       6 97255      0 Z    300  MB MRF    0  65  65   -PS-
       7 97255      0 Z    250  MB MRF    0  65  65   -PS-
       8 97255      0 T    1000 MB MRF    0  65  65   -PS-

   ...

So, the procedure works as it should here on our SPARC Ultra running
McIDAS-X 7.60.

To investigate why things are not working for you, I did basically the
same thing as above on aeolus:

o login as 'mcidas'

o cd to workdata

o set DISPLAY back to Unidata

o ran a McIDAS-X session

o deleted HRS.SPL:

  LWU DEL HRS.SPL

o set the pointer in GRIBDEC.PRO

  LWU POKE GRIBDEC.PRO 4096 0

o changed to the user ldmadmin and:

  cd /home/mcidas/workdata
  chmod 664 XCD_START.LOG
  cd /incoming/data/grib
  chmod 644 *
  exit

o verified the settings in xcd_run (the copy in /home/mcidas/bin

o catted the same input grid into xcd_run:

  cat /incoming/data/grib/us008_gf078_97091200_YxAAx | xcd_run HRS

o change the REDIRECTion for GRID5* to be /home/mcidas/workdata:

  REDIRECT ADD GRID5 ".

o started the GRID decoder back in the McIDAS-X session:

  DMGRID

o ran top to see dmgrid.k working

DMGRID ran for awhile but produced nothing.  Rats.

Now, the question is why this does not run on your machine.  It must be
due to a grib format change that your 7.1 code does not reflect.  Time
to switch to attempting the decoding on windfall (or load 7.6 on aeolus).
I actually tried bringing a current version of dmgrid.pgm to aeolus,
but I couldn't get it to link.  There were new routines added to XCD
that the 7.1 distribution doesn't have.  I would pursue this, but
it seems like it might be a waste of time given how old 7.1 is and
how slow aeolus is.

Before logging off, I changed the REDIRECTion for GRID5* back to 
/incoming/data/mcidas/xcd.

Tom

>From address@hidden  Sat Aug 28 09:02:32 1999

re: weird directory structure on aeolus
>Strange, I think that you would have to work harder to unpack
>the distribution under something other than the home directory
>for 'mcidas'.

re: where to put the compressed tar file
They do.  This way the compressed tar file ends up in /home/mcidas.

re: unpacking under /home/mcidas/mcidas7.6
Even though the instructions did not say to do this.  Tch, Tch :-)

re: defining McINST_ROOT to be /home/mcidas/mcidas76
Sounds reasonable.  In my other email, I referred to a 760 directory,
but I made the commnet that the name of this installation directory
didn't really matter.  On our systems, I refer to the new installation
directories using something like mcidas76.  You could use 760 or
anything else EXCEPT mcidas7.6!  This is where a number of people get
tripped up.  Please don't make this mistake!!

>this is the directory
>I will link against to set the runtime designations....

Right.

>Anyway, I ran mcinstall (which _really_ should be called
>mcunpack, since it upacks the distribution, but doesn't 
>install anything, right?),

Yes, I have realized that for quite some time.

>and it did unpack the
>distribution, into (I'm sure you've seen this coming):
>/home/mcidas/mcidas7.6/mcidas7.6
>
>sigh. well...it seemed logical to me at the time.  

Backing out of this is very simple.  all you have to do is:

cd mcidas7.6
mv mcidas7.6 ..

>Let me know if I should fix or live with this ugly setup.

Fix it.  It doesn't really make any difference (as illustrated by
the working installation of McIDAS 7.1 on aeolus), but it will make
following the online instructions easier later.  Why invite more
consternation for a time when you have stopped thinking about these
things when the fix is so easy.

>From address@hidden  Fri Aug 27 15:18:08 1999

Just so you are not confused (like me), I just read the long message
you sent clarifying the setup of dual versions to work with runtime,
etc.  I hope its obvious that the message I sent you (as user mcidas on
windfall, meant to use my own id, oops) regarding the weird directory
structure I now have sitting on windfall in /home/mcidas, was sent
before I saw that you had made some recommendations on how to proceed.

So, I was going with a McINST_ROOT of mcidas76, rather than 760, etc.
I can work through this, and I am assuming there is no lasting problem
from having the source buried in:  /home/mcidas/mcidas7.6/mcidas7.6
(unless you tell me otherwise).

Now I really do have to go (just paused to read your long (and helpful)
message before walking out the door).  Whew, TGIF.

Jennie

>From address@hidden  Fri Aug 27 19:00:18 1999

Thanks for investigating the grib decoding by pulling a file over to
your machine.  In a way I am relieved to know that its software (and
not me) that seems to be the problem.

Given how slow aeolus is, I have no interest in upgrading to 7.6 on
that machine.  (thats why I made the joke about going off to smoke a
pack of cigarettes....thats how slowwwww it is- guess I should have
said something more palatable like drink a pint :-).  _

I guess its time to figure out how to run a decoder on windfall that
won't interfere with the regular real-time decoding of products.

Next week, I guess.

I appreciate your getting to the bottom of it...

Jennie