leif wrote: > Francois Bissey wrote: >> I would have to check my discussions with Paul but I think that’s why there >> is a difference for the ecm doctest (with Paul saying that he likes the >> current >> gmp behaviour). > > We fixed FLINT w.r.t. this (invertability modulo +/-1) at least, such > that FLINT > 2.4.5 copes with both GMP 5.x and 6.x, so IMHO yes. > (Haven't yet looked at the ECM / libecm extension module doctest though, > so I don't know if it's that, too.)
Oh (now I did), that's really a nice one. I'd say it's an upstream bug that whether ECM considers 1 a factor of 1 depends on the underlying library (GMP 5.x or MPIR vs. GMP 6.x). But the last release of GMP-ECM (6.4.4) dates back nearly three and a half years... (There was some activity to make a 7.x release years ago, but probably developers unexpectedly left such that development got stuck.) > MPIR will "catch up" with the next major release, some time in the future. So I'll have to make the doctest depend not only on the MP library used (GMP or MPIR), but also its version... Or just change the test with GMP upon upgrading the optional package to 6.1.1. And again change it for MPIR as well once we upgrade to some 2.8 or later. (As mentioned, a new major release of MPIR doesn't seem to happen soon.) -leif >>> On 2/08/2016, at 21:35, Jean-Pierre Flori <jpfl...@gmail.com> wrote: >>> >>> Also note that GMP changed its behavior for some gcd or modular >>> reduction computations or something like that btw the 5.x and 6.x >>> series, MPIR did not mimick this behavior yet. -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at https://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
