- serious issue on 64bit mmalloc() ?  [aug 2008]
  this occurs when you use shared libraries  in lnux and 
  mem.h is not in the main program, s-data.addr is then < 0
  
  Proposed solution(s):
  - increase MAXBUF so that we don't need mmalloc()
  - ensure we use static libraries
  - put mem.h in the program main (horrible, but works)
  - compile miriad with 8byte integers (gfortran now supports that,
        but g77 is unlikely to ever support it)

  the last step has not been well tested, but i'll give that a shot          
  shortly. For this 'intf2c' needs the -x flag and maybe the -c flag.
  gfortran uses the -fdefault-integer-8 flag.
   
  also check:  intel compilerL  ifort -i8 
               gfortran -fdefault-integer-8


- WSRT code merge:
  - ratty and intf2c to handle the new PTRDIFF_T type in for/f2c files
    and auto-detect *4 vs. *8
  - fftw (CUDA)
  - memory based scratch files
  - OpenMP 

- MIR4 vs. MIR5: consider Sault's proposal to handle long integers.
  **mostly implemented **

- integer*8 support; ratty flag to go back to the old situation if
  the fortran compiler does not support integer*8
  [this will support scratchfiles > 8GB]
  This could now be solved with the -i8 type flags to compilers,
  but need some code overhaul - see configure for detection/setting

- write out proper long integers in itemize
  question is how fortran can do this?

- 64bit machines might need a few manual patches, mostly because our software
  does not everywhere know that /usr/X11R6/lib should be /usr/X11R6/lib64:
	- in pgplot/sys_linux make sure /usr/X11R6/lib64 is used in
	  g77_gcc.conf. Normally the install.miriad file now takes care
	  of this.
	- in wip, readline is not found, so make a symlink:

		ln -s /usr/lib64/libreadline.a $MIRLIB

          this is also mostly solved by pgm.wip now
  Also on 64bit: gidriv.f, ppdriv.f and wddriv.f need to
  edit an INTEGER to INTEGER*8 declaration where PIXMAP is declared.
  Use the fix64bit.csh script in borrow/pgplot/drivers and re-install
  pgplot

- borrow/readline does not seem to work on macosx. the default system
  version

- no more maxdim.h and maxdimc.h in $MIRINC, since they are 
  generated via configure (planned for 4.1.x?)

- check if character*N variables (N some number) work ok with the 
  miriad f2c interface. Sault reverted the bug i found in uvdat.for.

- a conversion routine from MIR4 to MIR3 format (int8 -> int) if possible
  this program/script should check if all the int8 variables are short
  enough that they can be cut into an int4:
	nwcorr
	ncorr
	vislen	(points to visdata, which does not need to be changed)
	
	? flags		?? could this ever be > 2GB ???
	? wflags

- programs that read ascii files, tables and such should *as a rule*
  skip lines that start with the (universal unix) comment symbol (#)
	** partially done, see tabio.c **

- mir-config that can spit out the cflags, libs etc...
  for example , in the current release uvio  is not linked with LFS,
  unlike the rest. so uvio will hang....

- use scropen8,scrclose8,scread8,scrwrite8 in addition to scropen stuff
  since we 

- miriad name collisions:
	solaris:/usr/ucb/implot

- pgplot:  gfortran cannot compile the GIF drivers for earlier versions
	   of gfortran (4.0 and 4.1 series). For 4.2 and up this is ok
	   and the default in $MIR/src/sys/bin/lib.pgplot is to enable gif

- darwin:
  some don't seem to see -D on darwin, darwin_x86 ??  removed from bug.c

- macosx: linking errors restFP/saveFP
	http://nxg.me.uk/note/2004/restFP/
  or
	http://purl.org/nxg/note/restfp
