Upgrade optional arb spkg to 2.5.0

[fredrik.johansson]

New tarball here: ​http://sage.math.washington.edu/home/fredrik/arb-2.5.0.tar.gz

[cheuberg]

Fredrik, do you intend to prepare a patch or would you prefer that I do it?

[fredrik.johansson]

Haven't really caught up with the Sage development process, so I'd appreciate if you could do it!

[cheuberg]

I checked that the tarball provided in the description is the same as the one on the github site. The doctests in my non-merged tickets using arb (#17263 and its dependencies) pass.

---

Last 10 new commits:

​225ee88	Trac #17480: Combine description of def _add_ and cpdef _add_
​b19a161	Trac #17480: Replace "four" by "three" relevant functions
​2e57cf9	Trac #17480: Fix typo
​0549c8a	Trac #17480: Insist on subtle differences between methods
​7ebe661	Trac #17480: Language corrections
​bcceb90	element.pyx: more formatting fixes
​6fbe471	Trac #17480: Fix formatting of docstring of abs
​6ee1bb8	Merge remote-tracking branch 'trac/develop' into structure/element-doc
​340fc4f	Trac #17480: Two more ReSt fixes.
​3885e9f	Trac #17688: Upgrade arb spkg to 2.5.0

[cheuberg]

sorry, pushed wrong branch

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​0eb70b0

Trac #17688: Upgrade arb spkg to 2.5.0

[fbissey]

Marc have you checked that arb is passing its own test suite (spkg-check)? I am fairly sure it won't because there is no release of flint2 with the right function for the testsuite not to fail. You would need a git checkout or someone making a new release of flint.

I have decided to make an ebuild of arb for gentoo (in the sage-on-gentoo overlay) and I got this when running "make check"

/home/extra/portage/sci-mathematics/arb-2.5.0/temp/ccLcbPFS.o: In function `acb_modular_epsilon_arg_naive':
/home/extra/portage/sci-mathematics/arb-2.5.0/work/arb-2.5.0/acb_modular/test/t-epsilon_arg.c:52: undefined reference to `fmpq_dedekind_sum'
collect2: error: ld returned 1 exit status
../Makefile.subdirs:78: recipe for target '../build/acb_modular/test/t-epsilon_arg' failed
make[1]: *** [../build/acb_modular/test/t-epsilon_arg] Error 1

No stable release of flint has fmpq_dedekind_sum in flint 2.4.4 as shipped with sage. It is still arith_dedekind_sum. And of course that's were it stopped, there could be more.

[fbissey]

After patching it, I can confirm that it is the only failing test.

[mmezzarobba]

Replying to fbissey:

Marc have you checked that arb is passing its own test suite (spkg-check)? I am fairly sure it won't because there is no release of flint2 with the right function for the testsuite not to fail. You would need a git checkout or someone making a new release of flint.

No, I haven't. I usually don't do touch spkgs and I didn't think it was relevant. Feel free to set the ticket back to needs_work, or set its milestone to pending depending what you think is appropriate. Thanks for the comment!

[fbissey]

I feel compelled to put it back to "needs_work" which is almost certainly what Volker would do especially if he reads my comments. Te merge window for non-critical stuff in 6.5 is closed anyway so it is a matter for the next release.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​3aece2b

Trac #17688: Patch tests of arb-2.5 to be compatible with flint-2.4.4

[fbissey]

looks good to me. back to positive.