According to http://gmt.soest.hawaii.edu/ the current version of sci-geosciences/gmt id 5.1.0 Reproducible: Always according to http://gmt.soest.hawaii.edu/doc/5.1.0/GMT_Docs.html#switching-between-different-gmt-versions "We encourage all GMT users to start using version 5 immediately; it has been tested extensively by the GMT team and has benefitted from bug reports for the 4.5.x versions. Users who still worry about the new version breaking things may install GMT 4.5.x and 5 side by side." Maybe this is a candidate for slotting?
sci-geosciences/gmt-5.1.0 -> sci-geosciences/gmt-5.1.1
According to http://gmt.soest.hawaii.edu/ they released GMT-4.5.13 (2015-01-01).
(In reply to Thomas Beutin from comment #2) > According to http://gmt.soest.hawaii.edu/ they released GMT-4.5.13 > (2015-01-01). According to http://gmt.soest.hawaii.edu/ they released GMT-5.1.2 (2015-05-04).
I'll work on this if nobody else has done so. It won't have slotting, but at least there will be the option for GMT 5.
Created attachment 421832 [details] ebuild for 5.2.1 Proposed ebuild for latest version of GMT. Please test and provide feedback/improve!
This ebuild builds and installs fine for me. Running repoman across the ebuild identifies the following issues: > IUSE.invalid [fatal] 2 > sci-geosciences/gmt/gmt-5.2.1.ebuild: gpl > sci-geosciences/gmt/gmt-5.2.1.ebuild: lgpl These flags, which appear to restrict features within the build, will need to be added to metadata.xml. > LICENSE.invalid [fatal] 3 > sci-geosciences/gmt/gmt-5.2.1.ebuild: !gpl > sci-geosciences/gmt/gmt-5.2.1.ebuild: !lgpl > sci-geosciences/gmt/gmt-5.2.1.ebuild: Shewchuk I'm not sure how this license stanza is meant to be interpreted (specifically the "( !gpl !lgpl )") > LICENSE="LGPL-3 gpl? ( GPL-3 ) ( !gpl !lgpl ) Shewchuk" This also includes the introduction of a new license that is not present in /usr/portage/licenses. From LICENSE.TXT: > NOTE: Some of the source files in src/ have different licenses. For > instance Shewchuk's Delaunay Triangulator (triangle.c) is distributed > under a proprietary non-permissive license. Refer to the license in > every source file. Users who wish to enforce strict GPL or LGPL > conformance are advised to exclude copylefted sources in > cmake/ConfigUser.cmake.
Thanks for the feedback. (In reply to Sam Jorna (wraeth) from comment #6) > This ebuild builds and installs fine for me. > > Running repoman across the ebuild identifies the following issues: > > > IUSE.invalid [fatal] 2 > > sci-geosciences/gmt/gmt-5.2.1.ebuild: gpl > > sci-geosciences/gmt/gmt-5.2.1.ebuild: lgpl > > These flags, which appear to restrict features within the build, will need > to be added to metadata.xml. Will do. I have added them locally in order to specify that certain parts of the code are compiled. See next item... > > LICENSE.invalid [fatal] 3 > > sci-geosciences/gmt/gmt-5.2.1.ebuild: !gpl > > sci-geosciences/gmt/gmt-5.2.1.ebuild: !lgpl > > sci-geosciences/gmt/gmt-5.2.1.ebuild: Shewchuk > > I'm not sure how this license stanza is meant to be interpreted > (specifically the "( !gpl !lgpl )") > > > LICENSE="LGPL-3 gpl? ( GPL-3 ) ( !gpl !lgpl ) Shewchuk" I'll have to work on this a bit. The idea was that if you only wanted GPL or LGPL code, you would enable at most one of those USE flags. If you didn't care or agree to the below license, you can disable both. Any thoughts on a clearer presentation? > This also includes the introduction of a new license that is not present in > /usr/portage/licenses. > > From LICENSE.TXT: > > > NOTE: Some of the source files in src/ have different licenses. For > > instance Shewchuk's Delaunay Triangulator (triangle.c) is distributed > > under a proprietary non-permissive license. Refer to the license in > > every source file. Users who wish to enforce strict GPL or LGPL > > conformance are advised to exclude copylefted sources in > > cmake/ConfigUser.cmake. This would be done by setting USE="gpl" or USE="lgpl". Of course I would also add the text from these files into the "Shewchuk" license. Thanks for reminding me to check with repoman!
Created attachment 427064 [details] gmt-5.2.1.ebuild I've reworked the ebuild from attachment 421832 [details] a little with the following changes: - bumped to EAPI6 - removed multilib support (see below) - define LICENSE as I understand it's supposed to be applied - reorder RDEPEND - clean up unnecessary subshell calls - add missing '|| die's With the removal of multilib support, while the ebuild inherits and calls multilib phases, it didn't actually provide an ABI USE flag to be used, and when I tried to convert it to proper multilib it required conversion of a number of deps (sci-libs/hdf, sci-libs/hdf5, sci-libs/netcdf, sci-libs/gdal - I stopped there). It's also worth noting that this doesn't have any apparent reverse dependencies, so multilib shouldn't be required. Proper multilib support can be added back in if required, but it will need a number of other packages converted in order to do so. Otherwise, unless there are any objections (is LICENSE set out the way you intended it to be?), I think this should be able to be committed once the addition of the Shewchuk license is clarified.
It's also worth considering bug 430028 - this package downloads and makes use of GSHHG, whereas sci-geosciences/gshhs (a subset of GSHHG) already exists in the tree. Perhaps this could be used (and/or updated) instead of downloading redundant information and potentially causing file collisions.
(In reply to Sam Jorna (wraeth) from comment #8) > Created attachment 427064 [details] > gmt-5.2.1.ebuild Hi, is the any reason why this is not in the portage tree yet?
(In reply to Thomas Beutin from comment #10) > (In reply to Sam Jorna (wraeth) from comment #8) > > Created attachment 427064 [details] > > gmt-5.2.1.ebuild > > Hi, is the any reason why this is not in the portage tree yet? See comment #9...I think that bug is a blocker. I also forgot about this since then :)
According to ftp://ftp.star.nesdis.noaa.gov/pub/sod/lsa/gmt/README.GMT 5.4.2 and 4.5.16 of GMT have recently been released ($Id: README.GMT 10386 2017-06-20 03:12:50Z pwessel $)
According to ftp://ftp.star.nesdis.noaa.gov/pub/sod/lsa/gmt/README.GMT 5.4.4 and 4.5.18 of GMT have recently been released ($Id: README.GMT 10410 2018-06-22 18:51:42Z pwessel $)
While I agree that duplicate information and potential file collisions are ugly, I don't see that as a blocker which should keep ebuilds out of portage for years. It should be easy enough to declare a blocker on sci-geosciences/gshhs-data to avoid file collisions at the cost of preventing concurrent installations. Then if concurrent installs are needed by some users, perhaps they will lend a hand here.
I place ebuilds for gmt-4.5.18, gmt-5.4.4 and gmt-9999 in gsview overlay. There is also a pull request (https://github.com/gentoo/gentoo/pull/9645) to include them in the main tree.
Bumped in commit ec3bf8a6ee69aea6be83b2197915763c9962d505, thank you very much Michael.
Thanks a lot, Michael!