[daip] Re: install problem with AIPS 31DEC02

[Eric Greisen]

haller at aoc.nrao.edu writes:

 >      The reason for this E-mail is that I have just installed the
 > 3-27-02 version of AIPS on my laptop and encountered an error which
 > I traced to a small error in the INSTEP2 script.  There are
 > numerous places where the script attempts to use include files
 > from 31DEC01/INC/    (note the year).  A typical failure looked like
 > this in the INSTEP2.LOG file:
 > 
 > PP        : Preprocess /opt/aips/31DEC02/APL/DEV/UNIX/ZABOR2.C
 > PP        : into       /opt/aips/31DEC02/LINUX/PREP/ZABOR2.c
 > AIPSCC    : Date       Fri Apr  5 10:36:18 EST 2002
 > AIPSCC    : Interpret  AIPSCC  \
 > AIPSCC    :            /opt/aips/31DEC02/LINUX/PREP/ZABOR2.c
 > AIPSCC    : as         LIST=FALSE PURGE=TRUE
 > AIPSCC    : plus       /usr/bin/gcc -c -O3 -fomit-frame-pointer -funroll-loops -I/opt/aips/31DEC01/INC -D_FILE_OFFSET_BIT
 > S=64 -DHAVE_LINUX_GLIBC -DHAVE_LINUX_GLIBC \
 > AIPSCC    :            /opt/aips/31DEC02/LINUX/PREP/ZABOR2.c
 > /opt/aips/31DEC02/LINUX/PREP/ZABOR2.c:7: AIPSPROTO.H: No such file or directory
 > AIPSCC    : Compile of /opt/aips/31DEC02/LINUX/PREP/ZABOR2.c
 > AIPSCC    : ends with fatal error(s)!
 > 
 > the offending code appears to be: -I/opt/aips/31DEC01/INC
 > 
 > For sites that have 31DEC01 sitting around, this will not generate an error UNLESS
 > one of the *.H files was changed.  That might introduce very subtle bugs.
 > I had moved my DEC01 stuff to another partition (as part of a OS upgrade) which is
 > why I stumbled on this problem.  My kludge was to just make a symbolic link:
 > 31DEC01/INC -> 31DEC02/INC

     The error probably comes from your running INSTEP2 without
setting up all the logicals properly.  The actual code is -I$INC - so
you were running INSTEP2 on (one hopes) 31DEC02 but with all essential
logicals still pointing at 31DEC01.  The install.pl proc handles this,
but if you do it by hand the instructions in the installation
documents tell you to . LOGIN.SH or source LOGIN.CSH and then say
$CDTST so that all logicals point correctly.

Never having attempted this I do not know what may go wrong.  We have
not had this report from anyone else.  The logical is not likely to
be set to anything unless a previous version is present and *mostly*
the files do not change.  I would think all the Fortran pre-processing
would also have trouble.

 > 
 > .................................................
 > 
 > Termcap Vs ncurses issue: 
 >       At least when using SuSE (versions 7.1 & 7.3), INSTEP4 was
 > happy when I used the entry (in $SYSLOCAL/LIBR.DAT):
 > 
 > ........ preexisting code ................................
 > # the -ltermcap may need to be replaced with two lines having
 > # -lreadline:0: etc
 > # -lncurses:0: etc
 > 
 > $LIBR/AIPSUB/SUBLIB:0:$AIPPGM
 > $LIBR/APLLINUX/SUBLIB:0:$AIPPGM
 > $LIBR/YSUB/SUBLIB:0:$AIPPGM
 > $LIBR/YSS/SUBLIB:0:$AIPPGM
 > $LIBR/APLSUB/SUBLIB:0:$AIPPGM
 > $LIBR/APLLINUX/SUBLIB:0:$AIPPGM
 > $LIBR/APLSUB/SUBLIB:0:$AIPPGM
 > $LIBR/APLLINUX/SUBLIB:0:$AIPPGM
 > $LIBR/GNU/libreadline.a:0:$AIPPGM
 > ..................end preexisting code ...........................
 > -lncurses:0:$AIPPGM                     <- replaced line
 > 
 > I discovered (the hardway) that addition of the line "-lreadline:0:$APSPGM"
 > broke the script.
 >       I hope these comments help!   Have a good week end.


The lreadline used to be required but we have Suze systems to test
things on.

Eric Greisen