High priority:
==============
  
- 1 Page PDF how to use gMCP, including hint for R installation
- press e for epsilon, or 1+e+e^2
- Implement support for something like 2*\epsilon or \epsilon^2
 (without using a small value for epsilon like 10^{-6} 
  - that does work already if we ignore rounding problems).
- confidence intervals are strange...
- Edge weight from node i to node j: m[i,j] or m[j,i] - please decide.
- Flag 1+epsilon as line sum
- round adjusted p-Values up
- standard examples
- Think about "-Xrs" - see paragraph below.
- PView is not well revalidated - load bonferroni holm (3) - total alpha 0.667 is shown
  Also create node and remove node again - its panels are sometimes still visible for some time 

Medium priority:
================

- Warn user when sum of edge weights is bigger than 1
- Check mutator methods for errors (alpha > 1, edge weights > 1)
- Update graph2latex to the much better algorithm already implemented in Java.
- gMCPReport does not work for graphs without layout.
  also new BlockLayout
- call calculateWeightString() when options change
- Guide for usage under MacOS
- Test the option for rounding how many digits we want to see.
(search for DecimalFormat and replace it with getDecFormat())
- Big font sizes misplaces TextField for object name
- Ask for saving graph when quitting

Additional features / low priority:
===================================

- Do we really need PolynomF and gtools as dependencies?
- Don't create global variable like .tmpGraph etc.
- Report generation in GUI (and respect the option whether to show rejected null hypothesis)
- PNG export cuts perhaps edges. (But already improved!)
- Speed-up -> Matrices & perhaps ?ReferenceClasses ?
- Package highlight is cool - but how can I tell r-forge or CRAN to use it?
See also http://comments.gmane.org/gmane.comp.lang.r.general/200179
- Edge labels should be fixed and placed by user or free and adjust to moving nodes
- Can we add a RUnit Test for Java importing Random Graphs and then exporting the same graph?
- Icon for saving graphs? (And do we need this button?)
- option to load another R graph
- TikZ styles for rejected / rejectable nodes
- further tests

@Apple - is this a problem?

r-patched-macosx-ix86:
checking R code for possible problems ... NOTE
2010-10-02 03:00:01.838 R[97724:613] Apple AWT Java VM was loaded on first thread -- can't start AWT. 

-> reference to JGR or Deducer.

@graph-package:

How do I create a graphNEL object, where there is a node that has no out-going edges?
I thought the following will do it:

hnodes <- c("H1","H2")
edges <- vector("list", length=2)
edges[[1]] <- list(edges=c("H2"), weights=c(1.0))
edges[[2]] <- list(edges=character(0), weights=numeric(0))
# or "edges[[2]] <- list()" which gives the same result.
names(edges)<-hnodes
g <- new("graphNEL", nodes=hnodes, edgeL=edges, "directed")

But the problem is, that g@edgeData@data[[1]]$weight is now a numeric with attribute "names".
This makes problems, for example it confuses edgeWeights: 

edgeWeights(g)
$H1
H2.H1
        1 

$H2
numeric(0)

In my understanding this should be "H2" instead of "H2.H1" like it would be when H2 has any outgoing edges.

rJava: We had the following problem:

> library(rJava)
> .jinit()
> replicate(1000000,rnorm(1))
^C
Java requested System.exit(130), closing R.

We solved it by calling .jinit(parameters="-Xrs")

Does this have side effects?

From http://download.oracle.com/javase/1.3/docs/tooldocs/solaris/java.html :

-Xrs
    Reduces use of operating-system signals by the Java virtual machine (JVM). This option is available beginning with J2SE 1.3.1.
    In J2SE 1.3.0, the Shutdown Hooks facility was added to allow orderly shutdown of a Java application. The intent was to allow user cleanup code (such as closing database connections) to run at shutdown, even if the JVM terminates abruptly.
    Sun's JVM catches signals to implement shutdown hooks for abnormal JVM termination. The JVM uses SIGHUP, SIGINT, and SIGTERM to initiate the running of shutdown hooks.
    The JVM uses a similar mechanism to implement the pre-1.2 feature of dumping thread stacks for debugging purposes. Sun's JVM uses SIGQUIT to perform thread dumps.
    Applications embedding the JVM frequently need to trap signals like SIGINT or SIGTERM, which can lead to interference with the JVM's own signal handlers. To address this issue, the -Xrs command-line option has been added beginning in J2SE 1.3.1. When -Xrs is used on Sun's JVM, the signal masks for SIGINT, SIGTERM, SIGHUP, and SIGQUIT are not changed by the JVM, and signal handlers for these signals are not installed.
    There are two consequences of specifying -Xrs:
        * SIGQUIT thread dumps are not available.
        * User code is responsible for causing shutdown hooks to run, for example by calling System.exit() when the JVM is to be terminated. 

As always:
==========
- more references, since some people will actually look at this package.
- More details in manual pages.
