Bug numbers refer to the BTS at http://pari.math.u-bordeaux.fr/Bugs/
Done for version 2.16.1 (released 04/01/2024):
  Fixed
BA   1- lfunan(lfunetaquo([12,1;13*12,1]),1) -> stack overflow [#2457]
BA   2- FpM_intersect could return a Flm [#2459]
     3- ifactor_sign: ifac_isprime could be called on input with tiny divisors
BA   4- nfcompositum could return non-monic polynomials [#2461]
BA   5- gammamellininvinit(vector(9,i,1)) -> SEGV
     6- incorrect sanity check in rnfeltdown ("not in base field") [#2464]
HC   7- precision loss in gammamellininv [#2462]
BA   8- [pthread] parselect could crash
BA   9- rare infinite loop in ellsaturation
BA  10- quadunitindex(8461,2)->1 instead of 3 [#2466]
BA  11- chinese(Mod(1,3)*(x^2+x+2),Mod(1,5)*(x-1)) -> incorrect leading term
BA  12- lindep([0,x]) -> SEGV
BA  13- [libpari] RgX_div_by_X_x: wrong variable number
    14- incorrect results in bnrisprincipal if bitprecision too low [#2474]
MS  15- hyperellratpoints(2*x^3+x^2-x,1000) -> wrong result [#2475]
BA  16- ellisogeny over a nf -> inconsistent variables error
BA  17- w=3;quadgen(5) -> print x instead of w [#2477]
BA  18- FpX_convol: wrong result for mixed degree
LGr 19- parsum was inefficient
    20- mateigen(,1): eigenvalues were given as t_VEC or t_COL. Use t_COL.
BA  21- ellcard over small finite field: potential oo-loop [#2483]
BA  22- ellisomat/nf could miss isogenies of degree dividing the conductor
    23- mfcoef(mftwist(f,D), ...) wrong result [#2484]
    24- sumeulerrat / prodeulerrat: slow if s not an integer [#2433]
    25- 'path' broken when initial or trailing ':'
    26- snfrank([1],0) -> 1 [instead of 0].
    27- snfrank(vector of polynomials, 0) -> error
    28- division of t_POLs with compatible domains Mod(1,2)/(2*x+1) -> Mod(1,2)
    29- strjoin([1]) -> memory corruption
    30- qfbclassno(684) -> oo loop [#2491]
    31- gcd(x-1,(x-1)/x) -> 1/x
PMo 32- mftonew was wrong on oldspace -> remove mfbd layer
BA  33- rnf.p did not work
    34- expm1(tiny negative t_REAL) -> stack overflow [#2494]
    35- expm1(huge t_REAL) -> junk [#2494]
BA  36- hyperellratpoints(9*x^2,3) -> SEGV
    37- lambertw for 2-adics (also fix documentation)
BA  38- plotbox, plotrbox did not clip properly
BA  39- ellheight for non-minimal curve over a number field -> wrong result
    40- for x >= 0, x^(a/b) was using (x^(1/b))^a instead of x^q*x^(r/b),
        with a = qb + r.
    41- hypergeom(t_POL parameters, z) -> type error [#2515]
    42- polrootsreal(0E10*x + 0E10) -> SEGV [#2516]
BA  43- memory leak when using member functions and ~ arguments.
    44- idealstar(nf,id,,MOD).cyc was not reduced modulo MOD
    45- changed qfbred to use standard normalization (same as qfbredsl2)

  Changed
HC   1- dirpowerssum: allow functions with t_VEC output
HC   2- dirpowerssum: add optional flag 'both' [also in library mode]
BA   3- [libpari] RgX_halfgcd now only return the matrix. Use RgX_halfgcd_all
BA   4- [libpari] ec_bmodel, ec_phi2 now take a variable number
BA   5- ellisomat: when E/nf has CM by an quadratic order included in nf,
        return the discriminant of the order instead of an error [#2231]
BA   6- ZX_compositum_disjoint: renamed to ZX_composedsum
     7- allow t_INTMOD Mod(m,N) as a Conrey label for the character m mod N
     8- nfroots: return a t_COL instead of a t_VEC (same as other 'roots' fun)
     9- qfbsolve: call qfbcornacchia when the argument is prime (and disc < 0)
BA  10- [libpari] rename qfb_apply_ZM -> qfb_ZM_apply,
                         qf_apply_RgM -> qf_RgM_apply,
                         qf_apply_ZM -> qf_ZM_apply
    11- rnfinit(nf, [T,lim]); try to fix maximal order if wrong [#2494]
BA  12- [libpari] change prec to be a number of bits, instead of lg
BA  13- [libpari] suminf_bitprec, intnumromb_bitprec: use suminf, intnumromb
        ellL1_bitprec, ellanalyticrank_bitprec: use ellL1, ellanalyticrank

  Added
     1- [libpari] nfeltembed_i
BA   2- [libpari] Flx_extresultant_pre, FlxqX_resultant_pre
BA   3- [libpari] gc_stoi, gc_utoi, gc_utoipos
BA   4- [libpari] F2xX_shift
BA   5- [libpari] Flx_Fl_mul_pre
BA   6- [libpari] Flx_halfgcd_all, Flx_halfgcd_all_pre, FpX_halfgcd_all,
        FlxqX_halfgcd_all, FlxqX_halfgcd_all_pre, FpXQX_halfgcd_all,
        F2xqX_halfgcd_all, FFX_halfgcd_all, RgX_halfgcd_all
     7- [libpari] modRr_i
BA   8- [libpari] random_Fq
     9- [libpari] maxprimelim
BA  10- [libpari] FlxqX_div_by_X_x, FlxqX_div_by_X_x_pre
BA  11- [libpari] Fl_2gener_pre_i, Fp_2gener_i, Fl2_sqrt_pre, Flxq_sqrt_pre
BA  12- [libpari] Fq_elljissupersingular
BA  13- [libpari] FpXQ_ellcard_supersingular, Fq_ellcard_supersingular
BA  14- [libpari] Fp_ellj_to_a4a6
BA  15- [libpari] ec_half_deriv_2divpol
BA  16- New GP functions polisclass, elliscm
BA  17- [libpari] elljsupersingular_FpXQ
BA  18- New GP function ellsupersingularj
BA  19- [libpari] Flx_composedsum, FlxV_composedsum
                  FpX_composedsum, FpXV_composedsum FlxqX_composedsum
BA  20- [libpari] Flx_convol, Flx_composedprod, FpX_composedprod
BA  21- [libpari] RgV_F2v_extract_shallow
    22- [libpari] Fp_double
    23- [libpari] convenience function veclast
    24- allow snfrank(D,0) for snfrank(D)
    25- [libpari] dbg_fill_stack() [#2361]
    26- allow erfc(t_SER)
    27- allow hypergeom(N,D,t_PADIC)
BA  28- n#: primorial of n
    29- allow bernpol(k, a) [ = B_k(a) ]
BA  30- allow Qfb(a*x^2+b*x+c) and Qfb([a,b0;b1,c])
BA  31- New GP function parforstep
BA  32- New GP function polfromroots
BA  33- digits: support for negative bases but positive digits.
AP  34- alginit: new flag value for lazy factorization
BA  35- [libpari] F2xqXV_prod, F2xqV_roots_to_pol, FFV_roots_to_pol
BA  36- [libpari] RgM_div
BA  37- [libpari] ZV_max_expi, ZM_max_expi
BA  38- [libpari] ZM_merge_factor
BA  39- [libpari] RgM_Cholesky
BA  40- [libpari] hash_keys_GEN
    41- support kB or KB, MB, GB and TB as units for memory sizes (2^10,
        2^20, 2^30 and 2^40 bytes)
BA  42- [libpari] lg2prec, prec2lg
BA  43- New GP function ellisisom
BA  44- allow qfsolve([Q,F]) where F factorization of disc(Q)
BA  45- New GP function plotell

  Removed
     1- [graphic] support for Qt4

Done for version 2.16.0 (released 25/02/2023):
  Fixed
BA   1- ploth long help was misformatted
     2- polredbest(y^4 - y^3 - 1/4, 1) -> Error [#2414]
     3- mathnf over polynomial rings: wrong when rank not maximal [#2415]
BA   4- seralgdep(Ser([0,1,0,u]*Mod(1,(u^2+u+1)*Mod(1,2)),y),2,2) -> SEGV
     5- polroots(4*x^3+4*x^2+28*x+144) -> overflow [#2418]
     6- conversion to float of t_QUAD with t_FRAC components [#2419]
     7- bnfinit(x^3-810693844986180*x-30702939461011257135364,1).fu
        -> wrong result or SEGV [#2420]
     8- [1,2]*[3,4] -> strange error message about qfbcomp
     9- obscure oo loop in solve when solution ~ 0 [#2412]
    10- polcyclo(huge n, t_PADIC of positive valuation) -> overflow
    11- charpoly over R or C was unstable [#2406]
    12- polredbest: rare error "inconsistent concatenation" [#2423]
    13- hypergeom([0,0],[1],2) -> SEGV [#2425]
    14- fromdigitsu(v, B) actually computed fromdigitsu(v, |B|)
    15- powers(t_INT, ...) was very slow
    16- factor(prod(i=1,20,nextprime(2^i)), 131) would include 131 instead of
        stopping at 67
    17- sumeulerrat(x^2/(1+x),2) -> FPE [#2433]
    18- lambertw(10^1000) -> overflow [#2429]
    19- rare infinite loop in idealtwoelt [#2437]
BA  20- nfisisom(16*x^4+1,16*x^4+1) -> error [#2432]
    21- forprimestep(p=a,b,q,) overflow when a > primelimit and q large [#2439]
BA  22- lfun([bnrinit(bnfinit(y^2+1),12),[[2,0],[6,0]]],2) -> error
    23- lfunthetacost(polcyclo(43)): silent overflow
BA  24- ellcard(ellinit([1,1]*ffgen([41,16])))-> impossible inverse [#2441]
    25- wrong result (missing content) in idealred(nf, [id, elt])
    26- det(integer matrix) -> inconsistent exact division [#2445]
BA  27- fft([],[]) -> SEGV, fft([1],[1]) -> SEGV
BA  28- nfrootsof1(x^0) -> oo-loop
BA  29- hgmalpha did not check its input
BA  30- hyperellchangecurve([],[]) -> SEGV
BA  31- hyperellchangecurve(x^0,[1,matid(2),0]) -> SEGV
BA  32- rnfeltnorm(x,x) -> SEGV
BA  33- polylogmult(1,[],-1) -> SEGV
    34- ellheight(non-integral E) -> crash [#2448]
    35- content(1,y) -> pol_1(y) instead of t_INT 1
    36- s=taylor(1/((x-1)*(y-1)),x); taylor(s,y) -> invalid object [#2452]
    37- avoid crashing when sumnumsidi diverges (0 bits of accuracy) [#2454]

  Changed
     1- binomial(negative t_INT, t_INT): use Loeb's definition
     2- normalize t_RFRAC so that leading coeff of denoimator is "positive";
        compare: 1/(1-x) and 1/(x-1)
     3- t_SER: valser, evalvalser, setvalser need to be used instead of
               valp, evalvalp, setvalp
     4- mathess over R or C: use maximal pivot instead of first non-zero
     5- default 'parisize' is now 8M on both 32bit and 64bit architectures
     6- default 'realprecision' is now 38 on both 32bit and 64bit architectures
     7- default 'realbitprecision' is now 128 on both 32bit and 64bit
        architectures
     8- default 'primelimit' is now 2^20 [instead of 500000]
     9- allow fromdigits(t_COL or t_VECSMALL, B)
    10- [libpari] cgetalloc arguments swapped to mirror cgetg's definition
    11- simplified lfunabelianrelinit (absolute equation is now optional)
    12- quadclassnoF_fact and uquadclassnoF_fact now return h(D)/h(d),
        like quadclassnoF.

  Added
PMo  1- new syntax for forvec loops
BA   2- [GP] genus2igusa
     3- printf numerical conversions: support complex numbers [#2431]
     4- libpari vecsmall_counting_indexsort, vecsmall_counting_sort
        vecsmall_counting_uniq
BA   5- [riscv64] add riscv64 assembly kernel
BA   6- lfuncreate([bnr, subg]): Dedekind zeta for the corresponding Abelian
        extension.

  Removed
BA   1- [GP] lfunabelianrelinit. Use lfunfinit([bnr, subgroup]) instead.
