This reverts commit 144739bd.

Add fixes for -Wformat-security warnings.

This doesn't allow non-string literals as formats, these may include format specifiers and can cause buffer overruns, maybe. Seen on Ubuntu.

See merge request !69

This doesn't allow non-string literals as formats, these may include format specifiers and can cause buffer overruns, maybe.

Fix in the case where the git branch and revision cannot be recovered.



See merge request !67

Dynamic unlocks

Get rid of the fixed-size unlocks array in each task, and the link-task madness that this led to.

Peter, could you give this a thorough run to make sure it doesn't break anything new? Thanks!

See merge request !66

make sure the unlock_inds are correct after scheduler_set_unlocks, just in case we cant to call it repeatedly.

use dynamic unlock pointers, i.e. the unlock_tasks array in each task is just a pointer to a pool of pointers in the scheduler. this pool needs to be sorted whenever tasks are added.

Cleaned-up runner.h to avoid include problems in the standalone tests.



See merge request !65

Parallel scheduler

Make the scheduler itself use tasks to prepare each step. This is kinda black magic, as the scheduler has to use parts of itself to get itself up and running, but it works. 

See merge request !60

OS X-flavoured build fixes

A couple fixes to avoid assuming HAVE_SETAFFINITY, and to stop the linker complaining about duplicate symbols defined in a header.

See merge request !63

Documentation

A documentation section has been created using Read The Docs and is found in "/doc/RTD". In order to create the documentation Sphinx v1.3.1 was used.
To create the html file use:

make html

which creates "index.html" in "_build/html/".

An initial design layout has been created and initial drafts of the Task-Based Parallelism and Hierarchical Cell Decomposition have been completed and need inclusion in the master branch.

See merge request !62

Agreed upon version of the density vectorization test

Wrote a unified version of the vectorization test to be used as a basis for the tests of @jwillis, @rgb and @alepper 

The test also contains a brute-force calculation to check that any approximation made even in the stock DOPAIR1() function is correct.

See merge request !64

Doesn't like short circuits hiding initialisations, not sure that ticks is handled correctly, is it supposed to be an opaque type