Hey Martin, On Sep 7, 2012, at 9:30 PM, Martin Desruisseaux wrote:
> Hello Chris > > Thanks for the Jenkins build fix! NP! > > > Le 08/09/12 07:51, Mattmann, Chris A (388J) a écrit : >> Actually I don't. Do they both have GPG sigs associated with them and MD5s? >> I think the Maven central requires the 1st zip file, and perhaps it >> has something to do with the naming convention for it, but I'm not sure. > > I just performed a more extensive comparisons. The two files are strictly > identical, except for one difference. The > sis-0.3-incubating-SNAPSHOT-source-release.zip file contains 3 additional > files which are not in the source code repository neither in the other ZIP > file. I presume that those three files were automatically generated: > > LICENSE > NOTICE > DEPENDENCIES > > The LICENSE and NOTICE files are strictly identical to the LICENSE.txt and > NOTICE.txt files at the root of the project directory, except for the notice > header: > > NOTICE.txt (from SVN checkout): > Apache Spatial Information System (SIS) > Copyright 2010 The Apache Software Foundation > > NOTICE (as generated in the > sis-0.3-incubating-SNAPSHOT-source-release.zip): > Apache SIS > Copyright 2010-2012 The Apache Software Foundation > > The DEPENDENCIES file does not exist in the SVN checkout. There is no GPG > signatures or MD5 checksums, but it may be (at least for GPG) because I do > not have the necessary keys. > > The ZIP files contains all of LICENSE, LICENSE.txt, NOTICE and NOTICE.txt > files. Possible approaches for avoiding this duplication could be either to > omit the LICENSE.txt and NOTICE.txt files from our code repository since they > are added automatically, or to rename them without the ".txt" extension. In > the later case the assembly plugin pickups the file as it stands in the > project repository. But if we choose the later approach, then we would also > need to remove the ".txt" extension from "README.txt", "HEADER.txt" and > "CHANGES.txt" for consistency... I wonder if there is some standardization package here or something for Maven central that is being generated. IOW, I doubt that the duplicates are there by coincidence and I also doubt we're the first people to realize it (maybe I'm wrong :) ). That being said, can folks like Kevan explain if they know why these files are present (with different or no extensions) -- is it for licensing stuff, or maybe related to Maven Central? Maybe we can reach out to the authors of the Maven apache profile or the release plugin and ask them :) > > >>> If so, then I would propose to limit ourselves to small files since those >>> data may not be the primary interest of peoples downloading the source >>> code, and in order to make debugging easier since it is harder to step >>> through a program performing million of iterations. We will nevertheless >>> need a few big test files, but I propose to put them in a separated >>> optional download. The JUnit tests using those big files would be >>> automatically skipped if those big data are not present (the exact location >>> of those optional files would be a future discussion if the principle is >>> okay). >> Yep that sounds good, +1 from me. > > Then a first question is whatever we want to rename the "sis-data" module as > "sis-test-base" or something like that. This module could contain not only > data, but also some helper classes for performing tests. We could also > question whatever we need such module at all since each module can also > package its test classes for usage by other modules... That would be fine with me. > > About large files, there is two questions: 1) where to put them on SVN and 2) > where to put them on the developer machine. For question 1, one possible > approach could be to create a > http://svn.apache.org/repos/asf/incubator/sis/data directory. +1 to create the data directory outside of trunk. Storing large data files in SVN though is probably not a great idea either, but 5mb isn't bad. Pulling it out of the code area though is good. > For question 2, I think that this is related to whatever we allow SIS to > create a ".sis" or ".apache/sis" directory in the user home directory (like > what Subversion, Maven or Jenkins do). The SIS library may benefit from such > directory for its normal operations (database storage, etc.). Test files > would just be an additional subdirectories in the ".sis" directory, where > developers can unzip test data. This might make a lot of sense, especially if we have sis config or tmp files that need to go there. +1 for a $HOME/.sis directory when the time comes. > > I guess that the directory question needs more though, so maybe we can just > move "sis/trunk/sis-data" to the above-cited "sis/data" for now and see later > for question 2? +1 to that solution, for now :) Cheers, Chris ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Chris Mattmann, Ph.D. Senior Computer Scientist NASA Jet Propulsion Laboratory Pasadena, CA 91109 USA Office: 171-266B, Mailstop: 171-246 Email: chris.a.mattm...@nasa.gov WWW: http://sunset.usc.edu/~mattmann/ ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Adjunct Assistant Professor, Computer Science Department University of Southern California, Los Angeles, CA 90089 USA ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
