Changes commited to PARI GIT during the last two weeks

get_B0: replace mpadd by addii

4	4	src/modules/thue.c

fix doc of ZX_squff

7	6	doc/usersch5.tex

59- polcyclofactors(((x^6+x^3+1)*(3*x^3+1))^2) -> oo loop [#2680]

- ZX_radical is now guaranteed to be primitive, fixing the above issue
  (BD required a primitive input)

- ZX_gcd_all(A, B, &Anew) returns d = gcd(A,B) and used to set Anew to an
  integer multiple of A / d. We now set Anew to the primitive part of A / d.
  This ensures the primitivity of ZX_radical

N.B. Pre-existing minor inefficiency: ZX_* or QX_* functions will
successively call Q_primpart() on their inputs although primitivity was
already ensured by the caller.

4	0	CHANGES
4	0	COMPAT
4	4	doc/usersch5.tex
10	4	src/basemath/QX_factor.c
2	1	src/test/32/cyclo
1	0	src/test/in/cyclo

ZX_content: add test for |x_0| = 1

2	1	src/basemath/polarit2.c

improve ZX_content

16	8	src/basemath/polarit2.c

add comments

2	2	src/basemath/QX_factor.c

add/improve comments

2	2	src/basemath/QX_factor.c
2	0	src/basemath/rootpol.c

refactor computation of ch_C

14	9	src/modules/stark.c

use divrmp

1	1	src/basemath/rootpol.c

23- mpdiv is now obsolete, use divmpr or divrmp instead.

The goal of mp functions is to
- allow t_INT / t_REAL inputs with easily determined output types
  (t_REAL when one of the inputs is a t_REAL) and fast calls to
  arithmetic kernel functions compared to generic functions
- that point aside, implement "generic" arithmetic

mpdiv(t_INT, t_INT) violated the second assumption by returning a
Euclidean quotient, while all other 3 type combinations (involving at
least one t_REAL) returned the ordinary quotient.

This created subtle bugs in linear algebra functions with 'mp'
coefficients using mpxxx functions. For instance dividing a t_INT by a
t_INT pivot gave a wrong (Euclidean) quotient instead of the required
(exact rational) quotient.

2	0	CHANGES
3	0	COMPAT
15	7	doc/usersch5.tex
2	2	src/basemath/trans2.c
2	2	src/functions/symbolic_operators/div
2	2	src/functions/symbolic_operators/dive
2	0	src/headers/paridecl.h
9	3	src/kernel/none/level1.h

newblock_t: use malloc

8	1	src/language/init.c

Add support for [...| ...; a=...]

6	0	doc/develop.tex
42	0	src/basemath/alglin3.c
22	0	src/functions/symbolic_operators/compr
4	0	src/headers/paridecl.h
2	0	src/headers/paripriv.h
9	2	src/language/parse.y
2	2	src/language/parsec.h
16	0	src/language/sumiter.c
1	1	src/language/tree.h
2	1	src/test/32/ranges
1	0	src/test/in/ranges

newblock_t: allow to define non-recursive block.

A block is non-recursive if either it has non GEN component, or more generally
all its GEN component are clone.

1	0	CHANGES
23	0	doc/develop.tex
5	1	src/headers/paridecl.h
0	4	src/headers/paripriv.h
4	0	src/kernel/none/level1.h
9	16	src/language/init.c

New libpari functions F3v_dotproduct, F3v_dotsquare

1	1	CHANGES
4	0	doc/usersch5.tex
35	0	src/basemath/F3v.c
2	0	src/headers/paridecl.h

New libpari function F3m_rank

1	0	CHANGES
2	0	doc/usersch5.tex
8	1	src/basemath/F3v.c
2	1	src/basemath/alglin1.c
1	0	src/headers/paridecl.h

test-ellff: set stacksize back to 10M

2	2	src/test/32/ellff
1	1	src/test/in/ellff

kill warning

1	1	src/basemath/Zp.c

Flxq_ellcard_i: update tuning

8	2	src/basemath/FlxqE.c

58- ellcard(ellinit([ffgen([53,36]),1])) -> wrong result

1	0	CHANGES
6	2	src/test/32/ellff
9	1	src/test/in/ellff

FqX_Fq_translate: handle FpX_get_red correctly

1	1	src/basemath/polarit3.c

FqM_mul_Kronecker: handle FpX_get_red correctly

1	1	src/basemath/ZX.c

Rg_to_FpXQ: handle FpX_get_red correctly

3	3	src/basemath/polarit3.c

ZM_lll_norms: cleanup tuning for Flatter use

- move the thre/thsn arrays out of the main function
- rename and simplify get_gaussred (which doesn't return an answer
  in "qfgaussred" format but a geuine Cholesky matrix)
- avoid calling ZM_rank in marginal cases (n <= 2) where it's not needed

42	31	src/basemath/lll.c

make sure QR_init fails in all cases of precision problems

- would not be the case when k = #M
- now gsisinv can be safely removed: it will never be false when QR_init
  returns 1

5	2	src/basemath/bibli1.c
2	11	src/basemath/lll.c

max_radius: fix optimistic comments and types

- The old mpdiv call was definitely wrong (we could have S[i,j] and B[j]
t_INT), introducing a unintended 'floor'.

- But the new code introduced in 18fc92cfa was too optimistic: it is not
guaranteed that B[i] is a t_REAL for i> 1

4	4	src/basemath/bibli1.c
4	4	src/basemath/nffactor.c

Flx_Teichmuller: avoid long division

17	5	src/basemath/Zp.c

polteichmuller: require explicitly that T is irreducible.

1	0	CHANGES
1	0	src/basemath/Zp.c
4	3	src/functions/polynomials/polteichmuller
6	4	src/test/32/padic
2	2	src/test/in/padic

polteichmuller: use faster algorithm for p>7

65	1	src/basemath/Zp.c

Flxq_lroot_pre: fix c++ warning

2	2	src/basemath/Flx.c

Fix line break after type

4	2	src/basemath/ispower.c
2	1	src/basemath/mftrace.c
2	1	src/basemath/random.c
28	14	src/language/anal.c
22	11	src/language/es.c
2	2	src/language/hash.c
2	1	src/modules/algebras.c

Fix line break after static

12	6	src/basemath/F2xqE.c
4	2	src/basemath/FlxX.c
4	2	src/basemath/FpXX.c
2	1	src/basemath/ellrank.c
2	1	src/basemath/gen2.c
6	3	src/basemath/lfunutils.c
4	2	src/basemath/modsym.c
16	8	src/basemath/nflistQT.c
4	2	src/basemath/quad.c
10	5	src/basemath/trans3.c
6	6	src/kernel/gmp/mp.c
4	4	src/language/es.c
2	1	src/language/eval.c
2	1	src/language/intnum.c
2	1	src/language/parsec.h
6	3	src/language/sumiter.c
11	5	src/modules/algebras.c
2	1	src/modules/elldata.c
2	1	src/modules/thue.c

_can5_invl: fix line break

4	2	src/basemath/Zp.c

_can_invl: fix line break

6	6	src/basemath/Zp.c