sci-mathematics/pari ebuilds dont work on gfbsd only because the author assumes that they are only used on linux Reproducible: Always Steps to Reproduce: 1.emerge pari on gfbsd Actual Results: "cd Olinux-*" from ebuild fails Expected Results: should merge I created patches for 2.1.7-r1 and 2.3.2-r1. Both patches use code/scripts from the package to calculate the correct name of the directory to enter. I successfully tested both ebuilds on linux/x86 and fbsd/x86.
Created attachment 147008 [details, diff] pari-2.1.7-r1.ebuild.patch the code to calculate osname and arch is copied from /Configure
Created attachment 147009 [details, diff] pari-2.3.2-r1.ebuild.patch uses a script from the package to calculate osname and arch, just like ./Configure does
Created attachment 147011 [details, diff] pari-2.3.2-r1.ebuild.patch attached wrong file ...
Hi Henning, Thanks much for the note & patch and I've just fixed this in pari-2.3.2-r1 in portage cvs. If at all possible, please CC the sci team on bugs like this in the future since otherwise we'll only discover them "by accident" since we're not notified. best, Markus
I did not assign this bug to Gentoo/BSD because it is no BSD-only problem. It is a general problem which only did not show up on linux. One of the bug-wranglers got it wrong. But you found it and that does not matter anymore. I came across the problem trying to install dev-perl/math-pari which depends on pari-2.1.7. So please consider to also commit the fix for this version.
Hi Henning, Typically, I never touch stable ebuilds unless they are broken for one of the keyworded arches. Unfortunately, it looks like fbsd currently is not among them. If these packages work for you on fbsd, you should consider asking the bsd folks to keyword them and the we can actually justify making these changes. Thanks, Markus
I see why you dont want to change an ebuild that works for all officially supported arches. I must admit that i dont really know whom i should report this bug. If i report it to you you say that you dont care about fbsd, which is fine. But i also cant ask the fbsd-team to add the keyword since the keyword alone wont repair this ebuild. Before the keyword is added the ebuild must be patched in order to work on fbsd. I guess the fbsd team would have to ask the maintainer of the ebuild anyways. And i assume that you are the maintainer. I may be wrong but i think that the keyword depends on the patch. AFAIK linux and fbsd share the same portage tree. 2.1.7-r2 with only ~x86-fbsd could be a solution that works for everyone.
Hi Henning, Can we close this bug? The proposed change is part of our >=pari-2.3.2-r1 ebuilds. Thanks, Markus
The latest dev-perl/math-pari still depends on pari-2.1.7. This bug cannot be closed until no other ebuild depends on <pari-2.3.2-r1 or these ebuilds get fixed too. I did not check how many other ebuilds depend on old versions of pari, when math-pari is the only one it might be a good idea to check whether it really needs 2.1.7.
dev-perl/math-pari now depends on 2.3.4, closing.
