Arch Linux User Repository

perhaps it was a server issue, the source is available.

the build should succeed without `--enable-download`, you just lose all the other components that the build system grabs.

I have poked around building the different components independently and getting freefem's build system to recognize them. I think I've been successful with (par)metis, superlu, tetgen, and mmg3d (mostly fixing header/lib locations). I've had trouble with others, like scotch_esmumps. Freefem will recognize the presence of the scotch/mumps libs, but magically won't build the corresponding freefem solib. Someday I might bug the devs.

perhaps it was a server issue. the mmg3d source is available

The build fails because it mmg3d4.0.tgz is no longer available online:

cd ../pkg;wget --no-check-certificate http://www.math.u-bordeaux1.fr/~cdobrzyn/logiciels/download/mmg3d4.0.tgz
--2013-11-21 11:19:51-- http://www.math.u-bordeaux1.fr/~cdobrzyn/logiciels/download/mmg3d4.0.tgz
Resolving www.math.u-bordeaux1.fr (www.math.u-bordeaux1.fr)... 147.210.16.9
Connecting to www.math.u-bordeaux1.fr (www.math.u-bordeaux1.fr)|147.210.16.9|:80... connected.
HTTP request sent, awaiting response... 403 Forbidden
2013-11-21 11:19:58 ERROR 403: Forbidden.

I suggest changing `--enable-download` to `--disable-download` in build script, and rather depend on mmg3d package on AUR explicitly (if mmg3d source is made available again).

A couple other things
- `--with-mpi=openmpi` seems to fix the problem with scotch
- running the build with makeflags seems ok now

Thanks for figuring that out.
I tried deleting the `--disable-pastix` line and it actually built just fine.

Hi !

New version released. I have a solution to compile with SuiteSparse installed, here is my PKGBUILD : http://pastebin.archlinux.fr/470589. I didn't succeed to compile with pastix enabled.

Hope this helps...

It was incorrect for me to have added umfpack as a conflict, according to https://bugs.archlinux.org/task/17625

Beyond adding a note, I don't know how to ignore the incompatible stuff from suitesparse. There doesn't seem to be anything relevant in the mailing list.

I'm also inclined to add --disable-pdf because the pdf is updated much more frequently than the tarball anyways.

It was incorrect for me to have added umfpack as a conflict, according to https://bugs.archlinux.org/task/17625

Beyond adding a note, I don't know how to ignore the incompatible stuff from suitesparse. There doesn't seem to be anything relevant in the mailing list.

I'm also inclined to add --disable-pdf because the pdf is updated much more frequently than the tarball anyways.

You can modify the PKGBUILD as follow :
- If you want to signal a conflict with umfpack :
conflicts=('umfpack')
provides=('umfpack')
This solves problems when you have another program which uses umfpack.
- You can also add this, which would be better I guess (not sure about that)
conflicts=('umfpack' 'suitesparse')
provides=('umfpack' 'suitesparse=4.0.2')
- I think that you can just not mention conflict with umfpack, it works well like that.

But I didn't find how to signal a conflict for compilation step, since it's impossible to compile when SuiteSparse is already installed and conflicts has effects on installation steps but not during compilation.

Also to solve multi-threads compilations problems you should add this to the PKGBUILD :
options=('!makeflags')
which will reset all makeflags during compilation.

Hi,

Thx for this package !
I found why mrbit and others have some troubles :
- missing dep : metis
- export MAKEFLAGS="-j1"
- uninstall SuiteSparse to compile

hope this help !