Updated Makefile and fixed bug where linker flags were mentioned before
object files.

free_matrix contained a bigger array than allocated

copying of material parameters at the global boundaries should be avoided
except when periodic boundaries are used left <-> right exchange should be done

release leads now master branch.
and beta leads to develop branch

(err is part of an standard library)

seperator between POS[1] and POS[2] to avoid ambiguity in output
if more than 10 processes for at least one dimension are used.

The sinus3 source wavelet was shifted in the wrong direction.

changed the eps figures to pdf (pdflatex should be used now)
Removed outdated and wrong parts in doc.
For example in section grid-dispersion a derivation at x=1/2 instead of x=0 was written.
I also cleaned up a little.

Fixed a critical bug in udate_v_interior_4.c which prevented FDORDER_TIME=4 to work properly.

Thanks Xue for reporting this issue.

To exchange the wavefield only a wrapping layer of FDORDER/2 Gridpoints is necessary.
The FDORDER/2+1 layer was probably implemented because it fixed indirectly the bug (inconsistency)
in the exchange_s/v functions.

undone last changes (nd=FDORDER/2 +1)

By using an operator of FDORDER length an addiotional frame around the subgrids with FDORDER/2 length instead of FDORDER/2+1 length is needed.
For example by using second order you will need one additional gridpoint to calculate the differences.

(can be found in file COPYING.)
+ changed link

Added the Adams-Bashforth method to the documentation.
ATTENTION: The documentation is now compatible with pdflatex, so it is strongly recommended to use pdflatex.

Added a new switch FDORDER_TIME to use a fourth order accurate temporal FD-scheme. This is realized with the Adams-Bashforth method.
Possible values for FDORDER_TIME are 2 and 4, whereas 2 refers to the classical leapfrog scheme.

TSNAP2 will be set to TIME if TSNAP>TIME. ( Time is recalculated to TIME=NT*DT in sofi3D.c and thus can differ to TIME given in .json. So even if TIME and TSNAP2 are equal in .json they can differ in the checkfd functions)

1. added derivation of the explosive source. This is necessary because the explosive source will be read in after the wavefield updates. Also the scaling of the explosive source is now corrected. 2. RSG isn't implemented in the current release