---------------------
HSM 0.90    6 Aug 17

* First major release


---------------------
HSM 0.91   22 Aug 17

* Added routines to assist in case setup
  - ISOMAT

* Changed applied edge moments to be conventional r x F moments.
  These differ from shell stress moments by a 90 deg rotation about n

* Made the manufactored solution for case 4 more exact by using
  quadratic von-Karman strains, replacing the linear KL strains.


---------------------
HSM 0.92    4 Sep 17

* Removed large work array allocation from all routines
  except the top-level driving program HSMRUN.  These routines
  can then be driven by applications with dynamic memory allocation.

* Crossed moment equation with normal vector n, so that its natural
  loading BCs are the conventional "r x F" moments, not the shell moments.
  This also simplifies moment-matching conditions at joint edges.


---------------------
HSM 0.93   30 Oct 17

* Added new work array ilast(kdim) to the call list of HSMSOL.
  Previously this was incorrectly allocated inside COLSET.

* Cleaned up the matrix-array size determination logic, chosen via itmax = -2

* This itmax = -2 can now be specified from the shell command, e.g.

    % hsmrun 4 1 41 41 -2

  in which case HSMRUN will only print out the minimum  nmdim  and stop.


**  Case 8 with more than one joint line still has some unresolved problem


----------------------
HSM 1.00   21 Jul 19


* Joints appear to be working well.  The wing case 8 has these.

* Data arrays have not changed, but there are a few new scalar arguments 
  passed in to HSMSOL, and a few of the other lower-level routines.

* The edge-BC parameter parb(..) array is now called pare(..), to avoid name 
  conflicts with future implementation of 1D beams.

* The lb* indexing pointers for pare(..) are now caller le* to match the 
  new name.

* HSMSOL now automatically performs membrane-only sub-iterations as needed 
  which greatly improves convergence with large deformations of very thin 
  shells. 
