Created attachment 405386 [details] rw-0.7.ebuild Hello, this will be a dependency of an upcoming release of SageMath. I've worked with the author to autotoolize the project and he was very helpful, so I'm willing to maintain this in the main tree. Should I add it with herd=sci? I don't care if somebody else wants to make fixes or version bumps. And if anyone wants to look over the ebuild before I commit it that would be great too =)
I am aware of this, and would you believe it I had an ebuild almost ready to go, I was just finalizing metadata.xml when my development box at the university was hit by a power cut (Friday evening 10pm). No sage-on-gentoo development until Monday. I had used autotools-utils with in source build (doesn't work out of source). I had made the executable flag default to "-" though.
(In reply to Francois Bissey from comment #1) > > I had used autotools-utils with in source build (doesn't work out of > source). I had made the executable flag default to "-" though. What did you use autotools-utils for? Does the ebuild I posted work for you?
The only box I could test your ebuild on, is my macos prefix. I may give it a try later today when I am not busy. Like I said no sage-on-gentoo until I can switch my box back on, unless someone send me clean pull requests on github. I usually use autotools eclasses when working with autotooled package. In this particular case the biggest pull to use it was to be able to add "static-libs" in IUSE. You are free to think it is over-complicated. Regarding your ebuild, it looks fine. Future work would possibly to split the shipping of examples and tie to a useflag.
(In reply to Francois Bissey from comment #3) > > I usually use autotools eclasses when working with autotooled package. In > this particular case the biggest pull to use it was to be able to add > "static-libs" in IUSE. You are free to think it is over-complicated. I wasn't suggesting that, only wondering if you thought of something that I missed. If the static libs are useful then I'll use autotools-utils. But I'm in no hurry, I can wait until your box is back up. > Regarding your ebuild, it looks fine. Future work would possibly to split > the shipping of examples and tie to a useflag. The examples are small in both number and size, so the prevailing opinion these days is to just install them unconditionally. We've been phasing out USE=bashcomp for the same reason, and to everyone's loud dismay, installing systemd unit files unconditionally as well. The one thing I do wish I'd done in Makefile.am is have the examples installed beneath the "example_graphs" directory, but that can wait until there's something else important to fix. Oh and the USE="+executable" was to match the upstream default. I argued with myself over that for a while, but settled with the rationale that I was putting it under sci-math and not sci-libs.
(In reply to Michael Orlitzky from comment #4) > (In reply to Francois Bissey from comment #3) > > > > I usually use autotools eclasses when working with autotooled package. In > > this particular case the biggest pull to use it was to be able to add > > "static-libs" in IUSE. You are free to think it is over-complicated. > > I wasn't suggesting that, only wondering if you thought of something that I > missed. If the static libs are useful then I'll use autotools-utils. But I'm > in no hurry, I can wait until your box is back up. > Not sure about the usefulness of static-libs, certainly not for sage. The static-libs use cases are not that common these days. > > > Regarding your ebuild, it looks fine. Future work would possibly to split > > the shipping of examples and tie to a useflag. > > The examples are small in both number and size, so the prevailing opinion > these days is to just install them unconditionally. We've been phasing out > USE=bashcomp for the same reason, and to everyone's loud dismay, installing > systemd unit files unconditionally as well. > OK, yes the balance point between finer control and the associated complexity is moving. Cannot say if it is good or bad. > The one thing I do wish I'd done in Makefile.am is have the examples > installed beneath the "example_graphs" directory, but that can wait until > there's something else important to fix. > > Oh and the USE="+executable" was to match the upstream default. I argued > with myself over that for a while, but settled with the rationale that I was > putting it under sci-math and not sci-libs. Good point, I was only putting it there myself because you could have an executable. Building the executable pulled more stuff than I had expected though.
I am moving your ebuild in the sage-on-gentoo tree so I can update the sage ebuild itself to build 6.8.beta5. I will move to autotools-utils on the basis that your ebuild as is, ship static libraries and I rather not do that. I would have had less problems if it didn't ship it by default. I had occasion where static libraries have been a problem. The circumstances were unusual but I still feel you should only have static libraries when you know you need/want them explicitly. I'll remove it once you have committed something to the main tree.
(In reply to Francois Bissey from comment #6) > I am moving your ebuild in the sage-on-gentoo tree so I can update the sage > ebuild itself to build 6.8.beta5. I will move to autotools-utils on the > basis that your ebuild as is, ship static libraries and I rather not do > that. Ah, using autotools-utils lets you DISABLE the static libraries. I didn't realize they were built by default. It all makes sense now. I just committed the ebuild from sage-on-gentoo with an updated metadata.xml. I added you as a maintainer and stuck the proxy-maintainers herd on there so you should be able to make fixes if I'm on vacation or something. *rw-0.7 (22 Jun 2015) 22 Jun 2015; Michael Orlitzky <mjo@gentoo.org> +metadata.xml, +rw-0.7.ebuild: New package sci-mathematics/rw to compute rank-width decompositions of graphs. Bug #552586.
