Summary: | sci-libs/scipy-0.17.0 fails to build after latest cblas update | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Mateusz Lenik <mlen> |
Component: | Current packages | Assignee: | Gentoo Science Related Packages <sci> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | audvare, dschridde+gentoobugs, harrisl, malocascio, prefix, tetromino |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- |
Description
Mateusz Lenik
2016-09-30 18:55:46 UTC
Getting the same error here with 0.16.1. Putting libraries = cblas in the various site.cfg files makes the compile go further but then it fails to link the first so file. # eselect cblas set 1 Fixed the issue for me (In reply to Harris Landgarten from comment #3) > # eselect cblas set 1 > > Fixed the issue for me Unfortunately not for me. # for i in blas cblas lapack; do eselect "$i" show; done lib64: reference lib64: reference lib64: reference The eselect symlink to the cblas.pc file was incorrect in the lastest cblas version. Can you please try again with cblas-reference version 20151113-r2? commit ec07bd6704edd93082d57c41c7edf461e46cbe58 Author: Matthias Maier <tamiko@gentoo.org> Date: Sun Oct 9 10:22:44 2016 -0500 sci-libs/cblas-reference: fix eselect symlnk to pc file Package-Manager: portage-2.3.0 (In reply to Matthias Maier from comment #5) > The eselect symlink to the cblas.pc file was incorrect in the lastest cblas > version. Can you please try again with cblas-reference version 20151113-r2? > > > commit ec07bd6704edd93082d57c41c7edf461e46cbe58 > Author: Matthias Maier <tamiko@gentoo.org> > Date: Sun Oct 9 10:22:44 2016 -0500 > > sci-libs/cblas-reference: fix eselect symlnk to pc file > > Package-Manager: portage-2.3.0 Had to get this manually (mirror slowly updating?), but it does make everything work. Works for me as well (In reply to Matthias Maier from comment #5) > The eselect symlink to the cblas.pc file was incorrect in the lastest cblas > version. Can you please try again with cblas-reference version 20151113-r2? It's completely broken on prefix. Look at the 20151113-r2 ebuild: mkdir -p "${ED}/usr/$(get_libdir)/blas/reference" || die mv "${ED}/usr/$(get_libdir)"/lib* "${ED}/usr/include"/cblas* \ "${ED}/usr/$(get_libdir)/pkgconfig"/* \ "${ED}/usr/$(get_libdir)/blas/reference" || die rmdir "${ED}/usr/$(get_libdir)/pkgconfig" || die rmdir "${ED}/usr/include" || die rm -rf "${ED}"/usr/lib/ || die See what happens when $(get_libdir) == "lib" (which is the case on prefix)? All the files you are installing in "${ED}/usr/$(get_libdir)/blas/reference will get wiped out by the "rm -rf" :)) (In reply to Alexandre Rostovtsev from comment #8) > (In reply to Matthias Maier from comment #5) > > The eselect symlink to the cblas.pc file was incorrect in the lastest cblas > > version. Can you please try again with cblas-reference version 20151113-r2? > > It's completely broken on prefix. Look at the 20151113-r2 ebuild: > > mkdir -p "${ED}/usr/$(get_libdir)/blas/reference" || die > mv "${ED}/usr/$(get_libdir)"/lib* "${ED}/usr/include"/cblas* \ > "${ED}/usr/$(get_libdir)/pkgconfig"/* \ > "${ED}/usr/$(get_libdir)/blas/reference" || die > > rmdir "${ED}/usr/$(get_libdir)/pkgconfig" || die > rmdir "${ED}/usr/include" || die > rm -rf "${ED}"/usr/lib/ || die > > See what happens when $(get_libdir) == "lib" (which is the case on prefix)? That's correct. Not only on prefix but also for i686, arm, etc. > All the files you are installing in "${ED}/usr/$(get_libdir)/blas/reference > will get wiped out by the "rm -rf" :)) Yup. Wow. I cleaned up sci-libs/blas-reference but apparently forgot cblas. commit aca8ce1ed089c80e40dbb6aa922ef70e45ac2eb9 Author: Matthias Maier <tamiko@gentoo.org> Date: Tue Oct 25 20:27:46 2016 -0500 sci-libs/cblas-reference: do not remove /usr/lib, bug #596036, bug #597288 Non non-amd64 systems "$(get_libdir)" might actually be "lib". So do not blindly delete this folder. Package-Manager: portage-2.3.0 *** Bug 546858 has been marked as a duplicate of this bug. *** |