Package Details: scotch 7.0.6-1

Git Clone URL: https://aur.archlinux.org/scotch.git (read-only, click to copy)
Package Base: scotch
Description: Software package and libraries for graph, mesh and hypergraph partitioning, static mapping, and sparse matrix block ordering
Upstream URL: https://gitlab.inria.fr/scotch/scotch
Licenses: CeCILL-C
Conflicts: ptscotch-openmpi, scotch_esmumps, scotch_esmumps5
Provides: ptscotch, ptscotch-openmpi, scotch_esmumps, scotch_ptesmumps
Submitter: None
Maintainer: ioquatix (MartinDiehl)
Last Packager: MartinDiehl
Votes: 41
Popularity: 0.78
First Submitted: 2006-11-07 17:51 (UTC)
Last Updated: 2024-12-06 16:25 (UTC)

Latest Comments

« First ‹ Previous 1 2 3 4 5 6 7 8 9 10 11 .. 16 Next › Last »

ioquatix commented on 2018-07-15 06:38 (UTC)

Okay, the v6.0.6 release is available.

ioquatix commented on 2018-06-25 04:02 (UTC)

Here is the new home for the project. https://gitlab.inria.fr/scotch/scotch

I've contacted the author again, I hope there will be a 6.0.6 release soon. Please be patient.

ioquatix commented on 2018-06-25 04:00 (UTC)

I am working with the author of this package, but I don't believe the next major release has been completed yet.

sigvald commented on 2018-06-20 20:24 (UTC)

@phleva: Indeed, "makepkg -sric --cleanbuild" (not pacman) got rid of some errors for me too.

Why is this package ouf-of-date by the way? Is it being deprecated in favor of scotc-mpich? Or is it a matter of so far unresolved issues?

ioquatix commented on 2018-02-26 01:53 (UTC)

I am communicating with the original developer and I hope there would be a good solution soon.

phleva commented on 2018-02-23 21:21 (UTC)

Just want to thank @AsmundEr and @sigvald and maybe help others along the way. Trying to build on a virtual machine with only one processor I ran into the mpirun limitation forementioned. Adding the --oversubscribe flag in the pkgbuild got rid of the error. Also, managed to clear the other errors by cleaning before and having the package localy. "pacman -sric --cleanbuild"

ioquatix commented on 2018-02-19 22:25 (UTC)

@sigvald I have passed your feedback upstream and will attempt to address your issues when I hear back.

sigvald commented on 2018-02-19 15:18 (UTC)

So I managed to install it and the mpirun problem was indeed a problem on my computer and not on the package per se. I'll give workarounds for both issues:

So for the SSL problem I don't know what's up with the SSL of that site. For those who are ready to take the chances of trusting this site, download the source file (http://gforge.inria.fr/frs/download.php/file/34618/scotch_6.0.4.tar.gz) into your scotch folder and run "makepkg -sri". Since the file is already there, it won't complain that the certificate of the site is not trusted.

Second, the testing of scotch involved running something on 4 processes while my laptop only have 2 cores. This was not a problem with my old installation of OpenMPI (perhaps an OpenMPI 3 thing?). If you can't run "mpirun -n 4 echo hello" then you have this problem too. However, adding "--oversubscribe" allows you to run 4 processes. I made makepkg add this flag to the test by adding a sed-line in the the check() function in the PKGBUILD file as follows:

check() {
  cd "${srcdir}/${pkgname}_${pkgver}/src"

  sed -i 's/mpirun/mpirun --oversubscribe/' check/Makefile

  make check LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:../../lib"
  make ptcheck LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:../../lib"
}

And that made it work. May I suggest adding that sed-line permanently to this PKGBUILD?

« First ‹ Previous 1 2 3 4 5 6 7 8 9 10 11 .. 16 Next › Last »