- I've changed pisa so that pisahelp builds using the _MONOLITHS
target rather than with an explicit ALINK.
- Had strange problem with ifd. AC_CHECK_PROG could not find tclsh
for me but STAR_CHECK_PROGS did. I've committed that change since
it seems harmless enough. (also tweaked .cvsignore)
- I think img should use STAR_CHECK_PROGS(generic) not AC_CHECK_PROG
I can change that if no-one (Peter) objects.
- I'm having trouble linking statically against MERS. As is Brad.
For some reason the msg1_gtstm and msg1_gtwsz routines can't resolve
on link. Presumably because there are circular link references between
the libraries. I've taken the bold step of getting rid of liberr
and creating a single library for adam and a single library for
standalone. This works (the tweaks are minor) so I've committed to
CVS. Trivial to reverse. Should liberr_standalone be called liberr?
dtask_link_adam needed a related tweak since it did not call
err_link_adam. I'm happy to revert if someone else works out how
to fix it (this "fix" results in an increase in size of the
shared library by 30 kB).
- I've spent some time looking into the PISAPLOT problem.
[note this is the narrative of my discovery rather than just the
answer]
The error occurs at
CALL AGI_ASSOC( )
at around line 440 (give or take my debug prints).
If I do an RDPIFD before the AGI_ASSOC the read works.
If I do the read after AGI_ASSOC I get the bizarro error message
about not being able to read the sequential file.
Simply switching to AGP/Native PGPLOT fixes the problem (seemingly
PISA only uses PGPLOT calls for plotting so we may want to just
do that anyway).
The implication is that opening the graphics device overwrites
a file unit, leading to GKS. The GKS gkmc.par file springs to mind.
Indeed, changing KMXIO from 40 to 96 (which is the old linux value)
the pisaplot problems disappear. I really don't know what KMXIO
does since it is clearly not the number of available fortran units
since I think linunx is 99 (actually, looking through the code it
seems that sometimes KMXIO+3 is used so KMXIO might be the max minus 3
which does agree with 99. How come FIO uses 40?) and the include file
used to have 1073741823 for alpha and solaris. Any ideas? Does solaris
really allow that many fortran units?
+ I've committed the GKS patch to up the number to 96
+ Can I switch pisa to native pgplot?
--
Tim Jenness
JAC software
http://www.jach.hawaii.edu/~timj
|