An effective parallel version of close vector enumeration in lattices has finally been completed.
M. Stoll has extended some features of the Jacobian code for hyperelliptic curves so that it now applies to Jacobians over number fields. The relevant intrinsics are TorsionBound, ReducePoint and Order.
A major revision/extension of the machinery for computing various aspects of element conjugacy for the classical groups has been completed. In particular, machinery for computing conjugacy in classical groups over GF(2) is now available. Furthermore, the general intrinsic ConjugacyClasses now uses the fast algorithms for classical groups where possible. Details can be found in the V2.26-4 version of the Magma handbook. The project was undertaken by Don Tayor with major contributions by Eamonn O'Brien and his student Giovanni de Franceschi, especially for the groups in characteristic 2.
The Handbook chapter on parallelism has been significantly expanded to include many examples.
A bug has been fixed when computing a ClassGroup of a degree 2 field.
The prefactorization of discriminants when computing a MaximalOrder over the integers has been improved.
The error checking of the index to . has been improved. Reported by G. Tornaria.
A crash when computing the NumberField of a FldAb constructed using an invalid map has been improved to an error message. Reported by L. Combes.
A problem with printing polynomials over complex fields with imaginary numbers has been fixed. Reported by J. Voight.
A bug in IntegralQuarticPoints that could allow a division by zero has been fixed. Reported by S. Tengely.
A rare crash in MordellWeilShaInformation has been fixed. Reported by M. Voznyy.
An error in the maximal subgroups machinery caused MaximalSubgroups to return two identical maximal subgroups for the groups PSL(3,p^2), where p is a prime equal to or greater than 5. Reported by Andrew Sutherland and fixed by Derek Holt.
A bug which sometimes caused InvariantSesquilinearForms and ((SemiInvariantSesquilinearForms to incorrectly return an empty sequence has been fixed by Don Taylor.
The function Evaluate(f, e) for rational function field elements has been improved to avoid confusing automatic coercion of e. Reported by M. Zieve.
A crash in large matrix muliplication over GF(5) on V100 GPUs has been fixed. Reported by R. Zeier.
Changes have been made by M. Stoll to improve the dynamical handling of parameters for the intrinsic Chabauty.
A small issue in CanonicalHeight that prevented it from being used when the base field is a rational function field has been fixed by M. Stoll.
A precision problem in ReducedBasis has been solved by M. Stoll by computing certain heights directly (at no great cost due to the use of the efficient Mueller-Stoll algorithm).
M. Stoll has extended some features of the Jacobian code for hyperelliptic curves so that it now applies to Jacobians over number fields. The relevant intrinsics are TorsionBound, ReducePoint and Order.
A bug in the code for fast numerical evaluation of hypergeometric series has been reported and fixed by John Voight.
A serious error when parsing expressions at the top level involving sums of exponents has been fixed (in V2.26-3).
A crash in the distributed version of the CloseVectors function has been fixed.
New functions ShortVector, ShortestVector, CloseVector, ClosestVector to return a single non-zero short/close/shortest/closest vector in a lattice.
An effective parallel version of close vector enumeration in lattices has finally been completed.
The factorisation of a polynomial over a Laurent series field defined over a finite field has been sped up dramatically in some cases with the incorporation of a suggestion from Juergen Klueners.
The printing of elements of precision 1 fields has been fixed so that the O() term uses the correct uniformizer name instead of 0. Reported by T. Keller.