update normaliz to 3.1.0

[mkoeppe]

Normaliz needs updating to bugfix release 2.12.2 or better

From normaliz website: Version 2.12.2: bug fix in volume computation with -v and change max_rank_submatrix back to pre2.12 version which avoids some overflows

(Upstream provides the distribution in a zip file; why does sage use a repackaged .tar.bz2?)

we still need to rename the package root directory, as our current scripts do not handle their naming scheme. Thus

tarball: ​http://users.ox.ac.uk/~coml0531/sage/normaliz-3.1.0.tar.bz2

[dimpase]

IMHO Sage's "sage -i" cannot use zip archives, it needs a tar.

[dimpase]

Replying to dimpase:

IMHO Sage's "sage -i" cannot use zip archives, it needs a tar.

zip would be yet another dependency to have; e.g. I'm not sure whether OSX without Xcode does have (un)zip installed; ditto for a typical Linux distro...

[vbraun]

Python comes with the standard zipfile module that can easily be used to unzip, e.g.: ​http://opendata.stackexchange.com/questions/1208/a-python-guide-for-open-data-file-formats/1928#1928

[dimpase]

Replying to vbraun:

Python comes with the standard zipfile module that can easily be used to unzip, e.g.: ​http://opendata.stackexchange.com/questions/1208/a-python-guide-for-open-data-file-formats/1928#1928

Do you mean to say that sage-spkg ought to get extended to support this? In #18748?

[dimpase]

oops, we now have 3.0: ​http://www.home.uni-osnabrueck.de/wbruns/normaliz/Normaliz3.0/

[dimpase]

let me try this upgrade...

[dimpase]

upstream ​says that 3.1 will be available in January.

(yes, they host the code here now: ​https://github.com/Normaliz/Normaliz/)

Am I correct that there is no exposure of normaliz in Sage, apart from normaliz lib in Singular?

[novoselt]

Replying to dimpase:

Am I correct that there is no exposure of normaliz in Sage, apart from normaliz lib in Singular?

I think so!

[dimpase]

version 3.1 is released

[dimpase]

without LD_LIBRARY_PATH, normaliz fails to find GMP/MPIR libs at runtime. Of course, without a configure script it is a pain to configure.

[vbraun]

The normaliz "buildsystem" is a joke, imho a serious red flag. Once again we see that handcrafted makefiles do not work. In this case, not passing through LDFLAGS to the linker. But really thats just the tip of the iceberg: Not able to build a shared library, not even an install target.

[vbraun]

A hackish workaround is probably to add LDFLAGS to CXXFLAGS.

[dimpase]

tested on Linux and OSX, via normaliz.lib in Singular, as was done on #13234. All seems to work.

[dimpase]

Replying to vbraun:

The normaliz "buildsystem" is a joke, imho a serious red flag.

they actually have cmake support, as I was told just now: ​https://github.com/Normaliz/Normaliz/issues/88 I haven't tried this with Sage.

[vbraun]

BTW Sage can now use zip files as "tarballs"

[dimpase]

Replying to vbraun:

BTW Sage can now use zip files as "tarballs"

I did have to re-pack anyway. Before I discovered this, I noticed a bug that I fixed. Perhaps I should have added it to this branch:

diff --git a/build/bin/sage-uncompress-spkg b/build/bin/sage-uncompress-spkg
index 0f88a21..f2d8f21 100755
--- a/build/bin/sage-uncompress-spkg
+++ b/build/bin/sage-uncompress-spkg
@@ -32,7 +32,7 @@ if __name__ == '__main__':
     if zipfile.is_zipfile(filename):
         # zip file:
         with zipfile.ZipFile(filename, 'r') as archive:
-            if len(sys.argv) == 1:
+            if len(sys.argv) == 2:
                 archive.extractall()
             else:
                 member = sys.argv[2]