- IMP-dev - salilab.org

Preparation for release 1: Trying out git
by Daniel Russel 18 Apr '13

18 Apr '13

As part of the upcoming release, we would like to switch over completely to git as it simplifies branch management a lot. As preparation for that, it would be nice if people try it out in their working setup a bit. To check out imp with git do - get the code `git clone https://github.com/salilab/imp.git` - setup things `./setup_git.sh` Git may instruct you to set up your email address and name, if so, follow its instructions. That will give you a repository that has the current state of the release candidate (which, for the time being is synced with svn). You then use it just like an svn checkout for building IMP. We can talk about committing things later. When you want to update do `git pull origin master`

4 9

switching over to git for release
by Daniel Russel 17 Apr '13

17 Apr '13

IMP is looking pretty good for the release. I'd propose that we stop using the svn repository this evening and make git the official IMP repository so we have a couple of days to deal with anything unexpected on that side that might come up. As part of that, I'd like to then run the cleanup scripts tonight (to remove large files) so that tomorrow morning we have git in a nice state for moving forwards. That would mean that people who have a git repository should probably clone a new copy tomorrow (you can also do "pull -f, but just cloning a new one is easier). Make sense?

1 0

salilab on github now support private repositories
by Daniel Russel 17 Apr '13

17 Apr '13

The salilab github account now supports private repositories. People who want to standardize on using git for all their source control needs can move their private repositories over there (the easiest way to do this is probably to email me, or Ben). Note that the number of private repository slots is limited (8 left at the moment). We can always do like with did before and have an uber-repository that contains several projects, but it would be nice to avoid this.

3 2

Pre-release minor changes
by Daniel Russel 16 Apr '13

16 Apr '13

I made a couple of changes to IMP module/application structure to to try to make things a bit cleaner before the release. First, the main documentation page for modules and applications has been moved into a file called README.md in the module/applications main directory (from doc/overview.dox). This file is a markdown file that is then displayed - by github when you browse the directory - provides a natural file for people to look at if they are looking through the source - and, of course, is show by doxygen as the documentation for the module (it is now a separate page rather than being mixed in with the docs for the functions/classes) I cleaned the files up a bit, but, in general, the contents have been a bit neglected. It might be worth people looking at the files for modules they are associated with to make sure everything is how they want. These changes only effect documentation, so are unlikely to break anything with modules that are not in imp proper. Second, I renamed the "description" files to "dependencies.py" as they are python files that describe module/application dependencies. If you want backwards compatibility in a module, just rename the file and create a symlink as "description". I updated isd2 and npctransport.

1 0

Reducing the size of the IMP repository
by Daniel Russel 15 Apr '13

15 Apr '13

As part of the switchover to git from svn, I'd like to clean large, no longer interesting files from the history (eg large data file that were added and then removed). Removing those files from the history reduces the size of the repository by about 50%. Since this involves rewriting the history, it would require some action by people who have an existing copy of the git repository. If you do and have no changes of your own in it, you can - simply clone a new copy - do "git pull -f origin develop" to pull a new copy of the develop branch (similar for master) If you have branches in your clone, then the easiest way is to rebase you changes against the updated repository. I can provide instructions for this, if desired. The main effects of shrinking the repository are: - its size on disk will be reduced from ~1 gig to ~.5 gig - clone will take half as long Otherwise git is pretty good at being fast no matter what. Does this seem desirable to people? --Daniel

3 2

A proposal to move IMP to 4 or 6 month mini-releases
by Daniel Russel 12 Apr '13

12 Apr '13

<https://github.com/salilab/imp/issues/256>

3 8

Ambiguity over meaning of optimized bits on attributes in the Model
by Daniel Russel 09 Apr '13

09 Apr '13

See <https://github.com/salilab/imp/issues/255> and chime in if you have thoughts on the matter. You need a github account to do so though, nudge, nudge :-)

1 0

Simplifying IMP::domino::load_merged_assignments
by Daniel Russel 07 Apr '13

07 Apr '13

IMP::domino::load_merged_assignments has a whole bunch of extra parameters that we added as part of various (to my knowledge) failed experiments. I'd like to simplify it to IMPDOMINOEXPORT void load_merged_assignments(const Subset &first_subset, AssignmentContainer* first, const Subset &second_subset, AssignmentContainer* second, const SubsetFilterTablesTemp &filters, AssignmentContainer* ret); Let me know if any of the experiments were actually working for you and so should be kept around. Thanks.

2 3

github accounts
by Daniel Russel 05 Apr '13

05 Apr '13

People who are on imp-dev probably should have github accounts and get added to the imp github users, if you aren't already. That way you will get notification about new issues, can edit the wiki and the bug tracker and all that. Currently, there is more discussion of proposed changes to IMP in the bug tracker than on this list, which is kind of nice (as the bug tracker provides an easier to navigate record). To do this, create a github account and then email Ben or I with the account name.

2 1

if you get a build error referencing RMF/profile.h or RMF/decorators.h...
by Daniel Russel 04 Apr '13

04 Apr '13

delete src/dependency/RMF/include and rerun cmake. The RMF build scripts weren't updating RMF.h properly. It has been fixed, but you may need to do that manual cleanup. Sorry.

1 0