Package Details: scotch 6.0.6-1

Git Clone URL: https://aur.archlinux.org/scotch.git (read-only)
Package Base: scotch
Description: Software package and libraries for graph, mesh and hypergraph partitioning, static mapping, and sparse matrix block ordering. This is the all-inclusive version (MPI/serial/esmumps).
Upstream URL: http://www.labri.fr/perso/pelegrin/scotch/
Licenses: custom:CeCILL-C
Conflicts: ptscotch-openmpi, scotch_esmumps, scotch_esmumps5
Provides: ptscotch, ptscotch-openmpi, scotch_esmumps, scotch_ptesmumps
Submitter: None
Maintainer: ioquatix
Last Packager: ioquatix
Votes: 30
Popularity: 0.000748
First Submitted: 2006-11-07 17:51
Last Updated: 2018-07-15 06:38

Latest Comments

1 2 3 4 5 6 ... Next › Last »

ioquatix commented on 2018-07-15 06:38

Okay, the v6.0.6 release is available.

ioquatix commented on 2018-06-25 04:02

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

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

@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

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

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

@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

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?

sigvald commented on 2018-02-19 12:46

It seems that this error is because the source of the .tar.gz file used to be on an http:// adress but now it is forwarded to an https:// adress (but otherwise the same adress) and my system cannot verify the validity of the certificate. Of course, neither can I, but if you decide to "trust" the source anyway, you can download the .tar.gz-file to your clone of the scotch folder and run makepkg -sri again. Since the source file is already there it won't use curl to try to download it. That gets us one step further, but now I get this one:

mpirun -n 4 ./test_scotch_dgraph_check data/bump.grf
--------------------------------------------------------------------------
There are not enough slots available in the system to satisfy the 4 slots
that were requested by the application:
  ./test_scotch_dgraph_check

Either request fewer slots for your application, or make more slots available
for use.
--------------------------------------------------------------------------
make[2]: *** [Makefile:258: check_scotch_dgraph_check] Error 1
make[2]: Leaving directory '/home/sigvald/Downloads/scotch/src/scotch_6.0.4/src/check'
make[1]: *** [Makefile:68: ptcheck] Error 2
make[1]: Leaving directory '/home/sigvald/Downloads/scotch/src/scotch_6.0.4/src/check'
make: *** [Makefile:106: ptcheck] Error 2
==> ERROR: A failure occurred in check().
    Aborting...

I find this a bit strange since I don't normally have any problem running mpirun with 4 processes.

ioquatix commented on 2018-02-19 12:38

I get the same problem now. Is there something wrong with the SSL on that website?