Changes commited to PARI GIT during the last two weeks

10- factorpadic(1+O(3)*x^2,3,2) -> SEGV [#2701]

Original analysis and patch by Johan Mercedes

1	0	CHANGES
4	4	src/basemath/polarit1.c
8	1	src/test/32/padic
5	0	src/test/in/padic

polsubcyclofast(8,4) and (16,4): wrong result

4	4	src/basemath/nflist.c
4	2	src/test/32/subcyclo
1	0	src/test/in/subcyclo

restore ndec2prec, deprecate ndec2nbits instead

1	1	CHANGES
1	1	COMPAT
32	21	doc/usersch5.tex
1	1	src/basemath/bibli2.c
1	1	src/functions/gp2c_internal/_prec
1	1	src/headers/paridecl.h
1	1	src/headers/pariinl.h
1	1	src/headers/pariold.h
2	2	src/language/default.c
2	2	src/language/es.c
1	1	src/language/eval.c

polsubcycloC4_i: remove argument fli which is always 1

12	12	src/basemath/nflist.c

groupelts_cnteq: missing vecsmall_sort [#2700]

238	237	src/modules/groupid.c

polsubcyclofast(52,4,,1) -> []

2	2	src/basemath/nflist.c
2	1	src/test/32/subcyclo
1	0	src/test/in/subcyclo

polsubcyclofast(58,7) -> []

1	0	CHANGES
1	1	src/basemath/nflist.c
2	1	src/test/32/subcyclo
1	0	src/test/in/subcyclo

move nbits2prec call to prec_fix

chk_gen_prec also needs it

2	1	src/basemath/base1.c

OCprecreal: restore compatibility by using OCbitprecreal

3	3	src/language/compile.c
4	2	src/language/eval.c
1	1	src/test/32/fingerprint

ellisomat: revert to kernel poly after a division by 0

1	0	CHANGES
110	30	src/basemath/ellisog.c
14	1	src/test/32/ellisomat
3	0	src/test/in/ellisomat

elkies98: move up

43	43	src/basemath/ellisog.c

New libpari function divisors_fact_degmax

2	0	CHANGES
4	0	doc/usersch5.tex
38	0	src/basemath/arith2.c
1	0	src/headers/paridecl.h

restore the OCbitprecreal opcode

Full backward compatibility is not possible since the OCbitprecreal and
OCprecreal opcodes now behave identically, but it didn't use to be the
case. This still improves compatibility: the compiler can no longer
generate that opcode, but the evaluator will still interpret it.

Older gp reading newly compiled code will behave differently than
current evaluator.

5	0	src/language/eval.c
2	1	src/language/opcode.h

remove the OCbitprecreal opcode and deprecate the 'b' prototype code

+ p (OCprecreal) has now the same semantics. 'b' is still accepted, for
backward compatibility but no longer documented.

+ Remove/rename the functions push_localprec -> push_localbitprec,
pop_localprec -> pop_localbitprec, get_localprec -> get_localbitprec.
They all correspond to localbitprec() uses, number of bits rather than
decimal digits.

1	1	CHANGES
12	17	doc/develop.tex
1	3	doc/usersch5.tex
4	4	src/basemath/lfunutils.c
1	1	src/functions/elliptic_curves/ellL1
1	1	src/functions/elliptic_curves/ellanalyticrank
1	1	src/functions/hypergeometric_motives/lfunhgm
1	1	src/functions/l_functions/lfun
1	1	src/functions/l_functions/lfunartin
1	1	src/functions/l_functions/lfuncheckfeq
1	1	src/functions/l_functions/lfunconductor
1	1	src/functions/l_functions/lfuncost
1	1	src/functions/l_functions/lfundiv
1	1	src/functions/l_functions/lfundual
1	1	src/functions/l_functions/lfunhardy
1	1	src/functions/l_functions/lfuninit
1	1	src/functions/l_functions/lfunlambda
1	1	src/functions/l_functions/lfunmfspec
1	1	src/functions/l_functions/lfunmul
1	1	src/functions/l_functions/lfunorderzero
1	1	src/functions/l_functions/lfunrootres
1	1	src/functions/l_functions/lfunshift
1	1	src/functions/l_functions/lfuntheta
1	1	src/functions/l_functions/lfunthetacost
1	1	src/functions/l_functions/lfunthetainit
1	1	src/functions/l_functions/lfuntwist
1	1	src/functions/l_functions/lfunzeros
1	1	src/functions/modular_forms/lfunmf
1	1	src/functions/modular_forms/mfcuspval
1	1	src/functions/modular_forms/mfeval
1	1	src/functions/modular_forms/mfmanin
1	1	src/functions/modular_forms/mfperiodpol
1	1	src/functions/modular_forms/mfsymbol
1	1	src/functions/modular_forms/mfsymboleval
1	1	src/functions/programming/getlocalbitprec
1	1	src/functions/sums/intnumromb
1	1	src/functions/sums/suminf
1	1	src/functions/transcendental/besseljzero
1	1	src/functions/transcendental/besselyzero
1	1	src/functions/transcendental/gammamellininv
1	1	src/functions/transcendental/gammamellininvinit
1	1	src/functions/transcendental/zetahurwitz
1	1	src/functions/znstar/znchargauss
1	3	src/headers/paridecl.h
1	1	src/language/anal.c
2	6	src/language/compile.c
12	24	src/language/eval.c
1	2	src/language/opcode.h
2	2	src/language/sumiter.c

deprecate functions related to the old 'prec' concept

10	0	CHANGES
9	0	COMPAT
2	2	doc/usersch4.tex
3	3	doc/usersch5.tex
1	1	src/basemath/base1.c
1	1	src/basemath/bibli2.c
1	1	src/basemath/ellanal.c
1	1	src/basemath/mellininv.c
2	2	src/basemath/trans1.c
3	3	src/basemath/trans2.c
0	4	src/headers/pariinl.h
1	1	src/kernel/gmp/mp.c
4	4	src/kernel/none/mp_indep.c
6	6	src/language/default.c
3	3	src/language/es.c
3	3	src/language/eval.c
7	7	src/language/intnum.c
1	1	src/modules/stark.c

remove neutralized nbits2prec uses, rename nbits2prec_required -> nbits2prec

2	4	doc/usersch5.tex
1	1	src/basemath/Qfb.c
1	1	src/basemath/arith1.c
3	3	src/basemath/base1.c
1	1	src/basemath/base3.c
22	25	src/basemath/bern.c
2	6	src/basemath/bibli1.c
1	1	src/basemath/bnfunits.c
1	1	src/basemath/buch1.c
10	10	src/basemath/buch2.c
8	10	src/basemath/ellanal.c
3	3	src/basemath/elliptic.c
1	1	src/basemath/ellrank.c
7	7	src/basemath/elltrans.c
3	3	src/basemath/gen1.c
10	12	src/basemath/gen3.c
11	11	src/basemath/grossenchar.c
7	7	src/basemath/hgm.c
1	1	src/basemath/hyperell.c
4	4	src/basemath/hypergeom.c
6	7	src/basemath/lambert.c
32	33	src/basemath/lerch.c
148	165	src/basemath/lfun.c
16	19	src/basemath/lfunlarge.c
6	7	src/basemath/lfunquad.c
29	35	src/basemath/lfunutils.c
1	1	src/basemath/lll.c
14	15	src/basemath/mellininv.c
85	91	src/basemath/mftrace.c
3	4	src/basemath/nffactor.c
29	34	src/basemath/rootpol.c
42	43	src/basemath/trans1.c
13	13	src/basemath/trans2.c
34	46	src/basemath/trans3.c
6	7	src/basemath/zetamult.c
0	6	src/headers/paridecl.h
3	17	src/headers/pariinl.h
6	0	src/headers/pariold.h
0	2	src/kernel/none/level1.h
2	2	src/language/eval.c
33	33	src/language/intnum.c
8	9	src/language/sumiter.c
6	6	src/modules/galois.c
5	6	src/modules/part.c
1	2	src/modules/ratpoints.c
5	5	src/modules/stark.c
3	4	src/modules/thue.c
0	2	src/test/dummy.c

improve explanation of t_REAL precisions

4	2	doc/usersch2.tex
22	17	doc/usersch4.tex
36	40	doc/usersch5.tex
12	20	src/functions/transcendental/HEADER

neutralize nbits2prec

This temporary commit allows abitrary nbits in the interface of public
functions (the notion of 'prec' as a multiple of BITS_IN_LONG
disappears).

Introduce a temporary internal function nbits2prec_required for the
limited uses where we must indeed replace an nbits by the next multiple
of BITS_IN_LONG. Namely all constxxx functions caching a t_REAL
constant: the cache mechanism compares the required bit precision with
the maximal bit precision of the cached constant. So the actual bit
precision of the constant (e.g., gpi) *must* be equal to its maximal
precision.

In a followup commit, all remaining nbits2prec are removed and
nbits2prec_required is renamed nbits2prec.

The main issue with nbits2prec was (incorrect) double rounding as in

  nbits2prec(a) + nbits2extraprec(b).

instead of the correct

  nbits2prec(a + b)

It also conflicts with the general goal of choosing an arbitrary bit
accuracy b, which is important when the complexity is not polynomial in b.

After the present commit, using a + b should be fine in all cases where
nbits2prec_required was not used. Many functions still compute with the
maximal accuracy allowed by a t_REAL length (a multiple of
BITS_IN_LONG), but it is no longer an interface requirement.

2	3	src/basemath/base1.c
5	3	src/basemath/trans1.c
2	1	src/basemath/trans3.c
4	1	src/headers/pariinl.h
1	1	src/test/32/algebras
1	1	src/test/32/modfun
1	1	src/test/32/nfields

group_identify: add supprt for groups of order <=191

1	0	CHANGES
1	1	src/functions/number_fields_galois/galoisidentify
677	337	src/modules/groupid.c

doc: group_ident{,_trans}: parameter elts was removed

4	6	doc/usersch5.tex

groupelts_to_group: when not wss, return elts instead of NULL

1	0	CHANGES
1	0	COMPAT
1	2	doc/usersch5.tex
9	11	src/basemath/galconj.c
14	14	src/basemath/perm.c

New libpari function group_iswss

1	1	CHANGES
5	0	doc/usersch5.tex
14	14	src/basemath/perm.c
1	0	src/headers/paridecl.h

7- matmodhnf: incorrect handling of empty matrix [#2699]

1	0	CHANGES
1	1	src/basemath/ZV.c
6	0	src/test/32/matmodhnf
10	2	src/test/in/matmodhnf

lfunlage: add 64 bits to target bitprecision in set_q_value

lfunlarge is heuristic and has been
tested in low accuracy and high height. It loses accuracy as
bitprecision increases.
{
  Z = lfuninit(1, []);
  for(b=32, oo,
    localbitprec(b);
    e = exponent(lfunhardy(Z, 2)-lfunhardy(1,2)) + b;
    if (e> 0, print([b, e])))
}
All values are imprecise when bitprecision in [111,127], and from 146 on.
Experimentally, e tends to infinity.

This patch doesn't fix the true issue, loss of accuracy now occurs
from bitprecision = 176 on.

6	4	src/basemath/lfunlarge.c