Return to PROBLEMS CVS log

Up to [CCP4] / ccp4

possible clash with `tplot'

                                    -*-mode:indented-text; fill-column:78-*-

This is a partial list of known problems with the current CCP4 code.  They
are either problematical to fix or have not been fixed yet for lack of time.

* Carriage control on the printer stream is inconsistent and there's no
  guarantee what the preconnected streams are.  The library should make
  guarantees about the preconnexion of streams 5 and 6 and whether or not
  carraigecontrol is honoured on stream 6, but it's not clear that this is
  possible. 

* Portability: various code uses somewhat non-portable features like
  INTEGER*2, long names (even as library externals :-().  Doubtless there's
  more which hasn't come to light yet.

* PLTDEV: scaling reportedly wrong HP output and reported to throw blank pages
  and hang HPGL plotter.

* PLUTO has many potential problems, including bad control flow (into loops)
  in routine circo -- NPO is the replacement

* The manual *still* needs more work.  See ccp4/manual/README about possible
  problems with the PostScript etc.

* Documentation for the libraries is still inadequate in some cases -- consult
  the library source or a program doing something similar if necessary.

* Some man pages assume fixed-pitch fonts and don't format properly with
  troff.

* `man' works differently on different systems.  It may tell you to need
  "Doumentor's Workbench" on SGI systems, for instance.  See comments in
  $CINCL/ccp4.setup-dist regarding MANPATH etc. or avoid man and just read the
  .doc files in an editor.

* AMORE may need to use enormous amounts of memory and probably won't work for
  large cases on many systems (including most VAXen?).  It also contains
  illegal Fortran and thus probably needs to be run with optimisation turned
  off (apart from peephole-type things, probably).

* Some of the programs may use too much memory to link on small VMS systems --
  no workaround known without changing their dimensions.

* Configure may fail without error messages if the shell runs out of space
  wherever it puts temporary files (e.g. /tmp).  It will also not stop if
  components fail to avoid lossage in the broken shell on one system that
  should remain nameless -- check for error messages as it runs.

* Some of the example unix sh scripts may not run on system with an old
  /bin/sh e.g. Ultrix.  You could install the GNU shell, `bash', and change
  /bin/sh in the scripts to the bash pathname.  I've also seen hints that
  there's a /bin/sh5 which should do.

* SCALA uses lots of memory.  You can reduce MAXPMR and MAXPAR in
  src/scala/parameter.h to save space, especially if running in rotavata mode.

* RSPS is reported to crash IRIX f77 3.5.1 (even without optimisation), but I
  can't reproduce this, so suspect it depends on available VM or somesuch.
  Presumably the remedy is to complain to SGI (but please report a workaround
  if you find one).

* VAX C: apparently stdlib.h is missing in some (older?) installations of VAX
  C.  The code should work without the #include if you comment it out.

* XLOGGRAPH isn't available for VMS.  The Athena widgets it uses don't appear
  to be distributed with DECwindows and there may be other problems.  If
  anyone with VMS Xaw can make it work, please send the recipe.

* ccp4.setup-bash doesn't set aliases up if using ksh rather than bash

* There have been several complaints about messages on SGIs such as:
    internal error writing to /tmp/ctmca0076S: Error 0
  Clear out your /tmp or read about TMPDIR in man f77

* Messages on Irix systems like:
    Object file format error in: /usr/local/ccp4/lib/libccp4.a(CCPONL.o): 
    cannot mix Elf and COFF objects.
  arise from trying to mix Irix4 and Irix5 objects.  Re-compile everything
  under Irix5.

* compar is known to have runtime-undefined variables which haven't been
  tracked down yet.  (Won't cause trouble on most systems with default
  compilation options.)

* Bugs in VMS DEC Fortran have broken MTZ handling at least with version
  v6.1-68 with optimsiation on; v6.2 is OK, as is 6.1 with /NOOPT.

* The lib/src/pack_c.c compiled with the --with-mosflm configuration option
  requires ANSI C, i.e. won't work in vanilla SunOS4, at least.  Could be
  fixed if anyone needs it.

* There are unresolved problems compiling under Irix6, which we don't have.
  There are reports that binaries from Irix5 mis-behave under Irix6.  These
  don't appear to be our fault and if you want them resolving you'll have to
  help us, probably by exercising your SGI support contract.

* The `proclean' component of Procheck doesn't compile on (some?) ESV systems,
  apparently due to a compiler bug.  The `f2c' compiler reportedly works if
  you can't get a fix for the MIPS compiler.  Beware also that there is a
  system `tplot' command which could clash with one of the prochek components.