The .gitignores contain the autotools files at the location wehre I finally want
them when I'm doen with the autmomake conversion.

Move test for enidanness into autoconf.
Previously, a Makefile would compile and run a special program that
tested system characteristics nad wrote an output file, which in
turn would be inlcuded as a header file.
This task is better left to autoconf.

…slower).

…stead

This proved more difficult than expected; in particular to pick
the right inputs for libmarx.a (which is far more than the *lis file states).
Because everything depends on everything (the directory structure on configure.ac,
the configure.ac on automake, automake on my ability ot formulate the right rules, ...)
this is not a small atomic commit but rather a "dumb everything in one buckt".

At this point automake + autoconf + autoheader work and compile
(I did not test if marx runs, yet!).
There is no use of libtool yet.

Also, change their installation location from /lib/marx to /bin.
This puts all binary executables in one place.

This file has not changed for 10 years and it's not big. It is much
simpler to include the binary in the distribution tarball.
(It's been in the repro for a long time anyway).

This neds seceral new dependencies, which I've just put in the Makefiles by hand for experimenting now:
  -L/melkor/d1/guenther/soft/simput/lib/ -lsimput -lhdio in marx/srclib and marx/scr
and CFLAGS and LD_LIBRARY_FLAGS to point to the include and ./.so files.

…RX variables

MARX PSF wrong for off-axis sources

SIMPUT format source

Ensuring the constant ARF for SIMPUT stays constant.

closes Chandra-MARX#17

An email from Vinay Kashyap summarizes the reason for this change:
We discovered a problem with how HRC raytraces of point sources are
handled.  See attached figure (figure removed because data is not public)
 of radial surface brightness
profiles of data from a suspected point source vs different methods of
raytraces.  Green is bare HRMA, sans detector; blue is data; black is
current default ChaRT+MARX; and red is ChaRT+MARX with different HRC-I
blur parameters (see below).  The curves are normalized to a peak of
1.

As you can see, the default raytrace predicts a significantly larger
wing than is observed.  We believe we have tracked this down to a
difference in how MARX renormalizes the amplitude of the 2D Lorentzian
compared to the Sherpa model function.

MARX uses a 2 Gaussian + 1 Lorentzian function to model the HRC
detector blur.  Diab tested cases where he manually set the amplitudes
of the second Gaussian and the Lorentzian to zero, and gets a
reasonably good match to the data --
HRC-I-BlurG1FWHM=12.44
HRC-I-BlurG2Amp=0
HRC-I-BlurL1Amp=0
The same changes would also apply to HRC-S-*.

Change defaults in marx.par for HRC PSF

This is required to dmmerge marx2fits output to e.g. ACIS stowed background files.
The exampel I used ot test this earlier used he background file
acis7sD2009-09-21bkgrndN0001.fits
where the N0001 in the file name is a version number. That older version uses the
 >i2 type and I had made marx to match that. In the current CALDB, there is a newer version of this file
acis7sD2009-09-21bkgrndN0002.fits
which is version N0002, which uses >i4 data datatype for PHA and PI columns.
My scripts did not catch that change, because I had the filename of the datafile hardcoded.
This commit switches marx2fits to use >i4 to match the event files (and background files)
that the current version of CIAO produces.

Change data type of PI and PHA columns written to match CIAO

Alan

User-compiled sources (``SourceType="USER"``) have a function called
``user_create_ray`` which is called with a pointer to ray properties
(direction, time, and energy). All ray properties used to be undefined before
calling this function and the user function could return ``-1`` for energy and
time to have |marx| assign energy and time. Now, ``energy`` is initialized
*before* ``user_create_ray`` is called, to allow user code that assigns
locations based on the ray energy.  With this change, marx.par must contain a
valid specification for the source spectrum, since an energy will be drawn,
even if the user source later assigns a different value. For
backwards-compatibility, if ``user_create_ray`` sets the energy to a negative
value, |marx| will assign the energy drawn before to the variable again.

closes Chandra-MARX#38

Assign energy before calling user_create_ray

This model was developed by @andrea-belfiore for one specific application.
It's included in MARX not necessarily because we expect other users to need
exactly this model, but like the other examples, because it is a good case
to see how different, realistic types of user model can be implemented.

Thanks to @andrea-belfiore for sharing!

Add contributed user model

Move test for enidanness into autoconf.
Previously, a Makefile would compile and run a special program that
tested system characteristics nad wrote an output file, which in
turn would be inlcuded as a header file.
This task is better left to autoconf.

…slower).

…stead

This proved more difficult than expected; in particular to pick
the right inputs for libmarx.a (which is far more than the *lis file states).
Because everything depends on everything (the directory structure on configure.ac,
the configure.ac on automake, automake on my ability ot formulate the right rules, ...)
this is not a small atomic commit but rather a "dumb everything in one buckt".

At this point automake + autoconf + autoheader work and compile
(I did not test if marx runs, yet!).
There is no use of libtool yet.

Also, change their installation location from /lib/marx to /bin.
This puts all binary executables in one place.

This file has not changed for 10 years and it's not big. It is much
simpler to include the binary in the distribution tarball.
(It's been in the repro for a long time anyway).