19990701: McIDAS-X 7.60 on Dec Alpha using gcc/f2c

>From: "Alliss, Randall J." <address@hidden>
>Organization: TASC
>Keywords: 199907011100.FAA08534 McIDAS-X DEC

Randy,

>As Denise mentioned we have built and compiled McIDAS7.6 on our Dec ALPHA
>using gcc2.9.1, f2c and the OS4.0f. I did have to make 1 change in
>mcimagem.c. I had to comment a call to "setpgrp" out. Actually the code
>tells you this call is not really needed.

I have been making a similar change to mcimagem.c since version 7.1.
My code modification looks like:

        if (detach)
        {
                if (setsid() == -1)
                {
                        return -1;
                }

#if 0                        /* <<<<< UPC mod 981201 >>>>> */
                if (setpgrp() == -1)
                {
#ifndef __hpux
                        return -1;
#endif /* __hpux */
                }
#endif /* __OPENNT || __osf__ */

>I can launch McIDAS and it fires up, however, any commands which try to
>access data returns an error. ie. IMGLIST, DSINFO etc. This is the case
>whether the data is local (ie. /mcidas/data) or from the ingestor. The error
>i get when issuing DSINFO I EAST ( which is GOES east ingest data)
>is the following:
>
>SERVER RETURNED ERROR NUMBER 1130458734
>ONTACT SERVER (CONNECT() FAILED)

What does the trace file look like:

DSINFO I EAST TRACE=1

To be honest, I don't know if DSINFO suppots TRACE=; if not, try getting
the trace output from IMGLIST: 'IMGLIST EAST/whatever TRACE=1.  The trace
file will be found in the mcidas/data directory of the user running
the ADDE server (either you or mcadde).

I built 7.6 (and 7.[12345] successfully on our DEC Alpha running Digital Unix.
Our current OS version is:

Digital UNIX V4.0D  (Rev. 878); Sun Feb  1 15:52:19 MST 1998

I have never seen this error.  Here are the results of a couple of
non-session McIDAS-X invocations:

dataloc.k

Group Name                    Server IP Address
--------------------         ----------------------------------------
MYDATA                       <LOCAL-DATA>
RTGRIDS                      ADDE.UNIDATA.UCAR.EDU
RTIMAGES                     ADDE.UNIDATA.UCAR.EDU
RTNIDS                       ADDE.UNIDATA.UCAR.EDU
RTNOWRAD                     ADDE.UNIDATA.UCAR.EDU
RTPTSRC                      ADDE.UNIDATA.UCAR.EDU

<LOCAL-DATA> indicates that data will be accessed from the local data directory.
DATALOC -- done

dsinfo.k IMAGE RTIMAGES

        Dataset Names of Type: IMAGE in Group: RTIMAGES

Name         NumPos   Content
------------ ------   --------------------------------------
ANTARCTIC       10    Antarctic IR Composite
EDFLOATER-I     10    Educational Floater
EDFLOATER-II    10    Educational Floater II
GE-IR           10    GOES-East North America IR
GE-IRTOPO       10    GOES-East IR/TOPO Composite
GE-VIS          10    GOES-East North America VIS
GE-VISTOPO      10    GOES-East VIS/TOPO Composite
GE-WV           10    GOES-East North America H2O
GEW-IR          10    GOES-East/West IR Composite
GEW-IRTOPO      10    GOES-East/West IR/TOPO Composite
GEW-VIS         10    GOES-East/West VIS Composite
GEW-VISTOPO     10    GOES-East/West VIS/TOPO Composite
GEW-WV          10    GOES-East/West H2O Composite
GW-IR           10    GOES-West Western US IR
GW-IRTOPO       10    GOES-West IR/TOPO Composite
GW-VIS          10    GOES-West Western US VIS
GW-VISTOPO      10    GOES-West VIS/TOPO Composite
GW-WV           10    GOES-West Western US H2O
MDR             10    Manually Digitized Radar
MDRTOPO         10    MDR/TOPO Composite
MOLL-IR         10    Mollweide Composite IR
MOLL-IRTOPO     10    Mollweide IR/TOPO Composite
MOLL-WV         10    Mollweide Composite H2O
RESFLOATER      10    Research Floater

DSINFO -- done

>I am assumming you have made some changes to the compiling process to get
>things working. Can you send those to me? I can send you in a separate email
>the makelog from our build/compile of 7.6 if you would like.

I just looked in my changes file to see what mods were made specifically
for DEC.  I find mods to:

makefile
mcaget.for
mcimagem.c
tclcomp.sh
mccomp.sh

The change to makefile was the addition of a VENDOR macro that is defined
to be '-vendor'.  This makes our distributions default to use of vendor
compilers instead of gcc/f2c.

The change in tclcomp.sh was to add DEC to the list of OSes that were
supported.  The change to mcaget.for is described as follows:

mcaget.for             SSEC 7.60 version with mods
                         copy FORM to CFORM and test CFORM(1:2) instead of FORM
                           this hack makes IMGDISP, etc. work on DEC OSF/1

The code mods themselves look like:

 ...
      CHARACTER*12      CID                 ! <<<<< UPC add 980420 >>>>>

This mod is for the addition of an ID= keyword for all of the IMG 
applications.  We added ID= to support the WIS NIDS ADDE servers that
I wrote.  This mod should not affect you.

 ...

C---------Determine data format
      CFORM=FORM                            ! <<<<< UPC add 961026 >>>>>
      CSTYP=' '
      stretch=0
      IF(CFORM(1:2).EQ.'I1') THEN           ! <<<<< UPC mod 961026 >>>>>
          IFORM=1
      ELSEIF(CFORM(1:1).EQ.' ') THEN        ! <<<<< UPC mod 961026 >>>>>
          IFORM=0
      ELSEIF(CFORM(1:2).EQ.'I2') THEN       ! <<<<< UPC mod 961026 >>>>>
          IFORM=2
      ELSEIF(CFORM(1:2).EQ.'I4') THEN       ! <<<<< UPC mod 961026 >>>>>
          IFORM=4
      ELSE
          IRET=-37
c         CFORM=FORM                        ! <<<<< UPC mod 961026 >>>>>
          CALL EDEST('MCAGET cant do format '//CFORM,0)
          GOTO 1000

As you can see from my modification tags, I have been making the same
change since October of 1996.

 ...
C------Default values for server keywords
      CBAND=' '
      CLMAG=' '
      CEMAG=' '
      CCCCAL=' '
      CDOC = ' '
      ctrace = 'TRACE=0'
      CT1='X'
      CT2='X'
      CT3='I'
      CDAY=' '
      CAUX=' '
      CID=' '                               ! <<<<< UPC mod 980420 >>>>>

Again, the CID mod is for support of our ID= keyword.

The change to mccomp.sh was the inclusion of the '-ieee_with_inexact'
flag in ccopts:

OSF1)

    # use the OSF1 vendor compilers
    case ${vendor:-$MCIDAS_CC} in
        ?*)
        cc=cc
        # The -DFUNCPROTO=15 is to get the X header files to include
        # their prototype information.
        # Maybe add -varargs too?
        # Using -warnprotos is good, except that some macros in
        # <ctype.h> use functions that have no prototypes.

        ccopts="-std -DFUNCPROTO=15 -ieee_with_inexact"
        ;;
    esac
    case ${vendor:-$MCIDAS_F77} in
    ?*)
        clopts=
        fc=f77
        fcopts=
        flopts=

        # If the main program is written in Fortran, we let the system
        # do its thing.  Otherwise, must tell the compiler not to use
        # its definition of main.

        case $fp_opt in
        ?*) ;;
        *)  flopts="$flopts -nofor_main"
            ;;
        esac

        fc_does_INCLUDE_search=true
        ;;
    esac
    ;;

I bet that the '-ieee_with_inexact' flag will not be recognized by gcc;
it was added for the DEC C compiler.

>Thanks for your help.

I am betting that the mods I list above do not solve your problem, but
I will be interested in finding out if they have any positive affect on
your installation.

Have you tried building using the Digital compilers?  I have not tried
building using gcc/f2c on any platform except Sun Solaris x86 and Redhat
5.2 Linux.

>regards,

If the above changes don't help, could you send along your McIDAS-X make
logfile so I can take a look?  Thanks.

Tom

>From address@hidden  Thu Jul  1 10:27:44 1999

Tom,

I will impliment your changes and remake using the vendor compilers first
then gcc/f2c if -v does not work.
thanks,

Randy