Hi Russell, Thanks for all the comments. Until the DC1 testing slows down, the Committee on Software Standards and Organization (or what ever its name) is on recess. After ADASS will be a good time to convene.
The selection of eUPS vs Pacman hasn't been discussed, much less decided. Your point about co-locating the LSST pacman scripts with their package is well taken. At the moment, all pacman scripts live in DC1/middleware/pacman; they didn't migrate into the final DC1 structure. I'll fold your comments into the strawman doc so they'll be discussed. Cheers, Robyn On Wed, Oct 11, 2006 at 12:09:52PM -0700, Russell E Owen wrote: > I just read Robyn's StdAndProcedures.txt file and have a few comments. > > Overall I really like it. The svn layout looks good (now that we have > chosen our build tools, we can practically have a setup.py file for > each chunk of python code). > > Some commands and questions: > > - Please prefix *all* LSST-generated environment variables with LSST_ > -- especially the names of 3rd party packages that we use. I worry > about name collisions otherwise (especially as we accrete additional > 3rd party packages). > > - I hope that version number and dependency metadata for a package > will live with (or very near) that package. Does PACMAN allow this? I > get the impression that the info is contained in the PACMAN script > itself. Which brings up the next question... > > - Where do PACMAN scripts live? With the package? In mw/packaging/? > > - In a similar vein: where should we put building and deployment > scripts, such as setlsst.csh and setuputils.py? In mw/packaging? If > not there, then might a 3rd top level directory with a name such as > "deploy" or "install" (but please not "build") be helpful? > > - Has the decision between PACMAN and EUPS been finalized or are you > just proposing to use PACMAN? I'm not familiar enough with either to > have an opinion, I'm just curious. > > - I wrote a script that will generate the basic empty python library > tree, in which the various packages belong. I recommend we use it for > DC2 and beyond. It avoids the need for __init__.py files cluttering > up the svn tree and for a master setup.py file. It searches for > setup.py files, so it automatically adapts as packages are added. > > - I suggest we allow Python's new Easy Install > <http://peak.telecommunity.com/DevCenter/EasyInstall> as an alternate > means of installing python code. PACMAN can run it just as easily as > setup.py. If nothing else, the bundled setuptools includes some > potentially useful tools for writers of installation scripts. But it > also has some features that might simplify the PACMAN scripts, > including dependency support and the ability to switch versions. > > - How and when do you envision making the transition to this new > build system? Copying the appropriate code from the DC1 branch to the > DC2 branc and then modifying the various build files? > > - On that topic...note that the DC2 branch exists, but with the old > directory layout. If no code has been checked into it, we could > delete it and start over. > > - A comment on "6.5.2 Modifying A File". If you rename or delete a > directory, then you should run "svn update" again after you commit > the change. The old empty directory is not actually removed until you > perform the update. > > -- Russell > _______________________________________________ > LSST-data mailing list > [email protected] > http://www.lsstmail.org/mailman/listinfo/lsst-data -- Roberta (Robyn) Allsman Large Synoptic Survey Telescope phone: 520-322-8741 Data Management FAX: 520-881-2627 4703 E. Camp Lowell Dr Suite 253 email: [EMAIL PROTECTED] Tucson, AZ 85712 USA _______________________________________________ LSST-data mailing list [email protected] http://www.lsstmail.org/mailman/listinfo/lsst-data
