[MITgcm-support] update breakage

Michael Schaferkotter (Contractor) Michael.Schaferkotter at nrlssc.navy.mil
Thu Aug 9 20:17:51 EDT 2007


jean-michel:

you wrote:
4) could be usefull to get a copy of your exf_readparms.f file (after cpp) 
 to see (with those line number) where the problem is.


i figured out what was causing the problem.
it seems that the cvs update is bringing down a corrupted copy of exf_readparms.F on the linux machine.
this corruption did not occur on the mac machines.

here/s the (failed) compile:
pgf77 -byteswapio -r8 -Mnodclchk -Mextend  -c exf_readparms.f
PGFTN-S-0021-Label field of continuation line is not blank (exf_readparms.f: 2848)
PGFTN-S-0038-Symbol, date_array, has not been explicitly declared (exf_readparms.f: 2850)
PGFTN-S-0038-Symbol, modelstartdate, has not been explicitly declared (exf_readparms.f: 2852)
PGFTN-S-0038-Symbol, difftime, has not been explicitly declared (exf_readparms.f: 2852)
PGFTN-S-0038-Symbol, modelstart, has not been explicitly declared (exf_readparms.f: 2854)
PGFTN-S-0021-Label field of continuation line is not blank (exf_readparms.f: 2889)
PGFTN-S-0021-Label field of continuation line is not blank (exf_readparms.f: 2893)
  0 inform,   0 warnings,   7 severes, 0 fatal for exf_readparms
PGFTN/x86 Linux/x86 6.2-3: compilation completed with severe errors
make: *** [exf_readparms.o] Error 2

in $MITGCMHOME/pkg/exf/exf_readparms.F (after the udpate) i found:


<<<<<<< exf_readparms.F
      if ( swfluxfile .NE. ' ' ) then
         call cal_FullDate  ( swfluxstartdate1   ,     swfluxstartdate2,
     &        date_array                        ,mythid )
         call cal_TimePassed(modelstartdate,date_array,difftime,mythid)
         call cal_ToSeconds ( difftime,   swfluxstartdate     ,mythid )
         swfluxstartdate    = modelstart  + swfluxstartdate
      endif

there was also a line at the bottom of the file:

<<<<<1.48


this indicated to me that the file was getting corrupted. i simply copied the 'good' version of exf_readparms.F, which
had been correctly updated via cvs. i then compared the replacement with 'cvs diff' and verified that the correct
file was now in place.

the compile then proceeded without incident.

i/ve seen this 'odd' behavior of the cvs update before. (once before i had to start over as something similar happened).

all's well that ends well.

thanks.

michael




More information about the MITgcm-support mailing list