As that does not support locking.

Mark it hidden for now.

Make install just take what is there. Use os.glob instead of scons glob.

Various minor cleanups. Not really related to the name any more.

Then can save the context of created objects and use that to improve the unused warnings.

The model is for having separate read and write processes. Would have a lock attribute and a version counter. Need tool to break lock. Make checking lock exposed to the user rather than low level.

make sure that one checks if the dependency has already been added

Factory, Decorator (and const versions).

Run flow init and copy files to hooks. Add nightly branch in repo.

Four rigid bodies, interactions between A -B and C- D and the interaction graph is a clique. Something with normalize may be causing problems.

It needs to check if they are non-zero.

This branch has changes to make it clearer what the various smart pointers do. It is ready to commit to svn.

Make it look like other aspects of statistics.

It would be nice to have the output from multiple nights consolidated into plots or something too.

We don't use any of the results any more. We can skip building lots of tables in data and remove the whole function lookup table.

particularly close pairs finders and scoring/dependencies

and how they compare to handwritten scores

Restraints can be identified by the restraint they came from along with the input particles. Use the restraint name for the group (just as a check) and write the id info as attributes.

To define a session and call update when done. Add option flag to require it.

Need to update function names to show that they are probabilities. And move mover to an IMP_LIST so we can get the number and get them back.

Since the required modules are sorted arbitrarily which doesn't really work as they need to be in the right order. Need to track dependencies with external builds too.

No inheritence. Example not linked.

just support cholmod=cholmod:m

This feature contains test code that shows a bug in hdf5 with filling. Ready to commit to svn.

http://code.google.com/p/address-sanitizer/wiki/AddressSanitizer seems like a good tool to run all of IMP with, and faster than valgrind.

Add Natalia's code as test.

Like what is done with rmf files, but so it can be used with read_pdb also.

Links from a class to all functions that use it and to all examples that use it.

Can just put it in the variables file or some other file. This avoids merge conflicts and reduces the number of files that change.

A quick perusal of the docs and I still have no idea how it works.

Not clear how to do it on swig yet. Also, hide all particle member methods, and add comment to particles about using the new method.

There is an inherent circularity in the dependency computations with tests in that the dependencies for a test depend on modules that haven't been parsed yet. This is probably responsible for much of the flakiness we see. Possibilities include

• do things in two passes, first stuff that can only look backwards and then tests and running examples that can look at the whole project
• munge more

It is not clear how to do the first, we can add a pass argument in the environment...

Something broke there

update of the data set with frame names. Ready to commit to svn.

Selections, hierarchies, XYZs. May require new module. Ick.

Add it to log message if print time is turned on.

Look into data set compression, setting chunk size intelligently, make sure there is no wasted space.

Probably due to passing IMP.domino instead of just domino. Fix.

in printing error message

• Fix RMF.h to include new files.
• add enum to specify STATIC
• make decorator check cover all frames if necessary
• add test for get all values in decorators