# $Id: CHANGES 10407 2008-07-08 12:07:44Z kb $
Bug numbers refer to the BTS at http://pari.math.u-bordeaux.fr/Bugs/

Done for version 2.3.4 (released 08/07/2008):
  Fixed
  Changed
    1- delete CVS.txt, add README-subversion
    2- delete TODO [ only relevant / up to date in the development branch ]
    3- factor(500501^36): impossible assignment I-->S                     [F21]
    4- quadray(-7,6) [or other "impossible conductors"] --> oo loop [#725][F22]
    5- mathnf(a,1) could have negative entries if non-trivial kernel[#741][F41]
    6- duplicate factors in nffactor (easy cases) [#761]                  [F52]
    7- wrong signs in bnfisintnorm e.g. [bnfinit(x^3+5), 5] -> x [#767]   [F55]
    8- memory leak: local(x = VECTOR/MATRIX VALUE) in user functions  
    9- acos(x^0) -> division by 0 [ instead of O(x^8), at \ps 16 ]  [2-4-2-F83]
       Analogously, acosh(x^0), asin(x^0), asinh(I+O(x)) --> division by 0
   10- zeta(0e1) --> SEGV [#627], exp(0e1*I) --> SEGV [#630]        [2-4-2-F44]
       zeta(1-2^31 or 1-2^63) -> SEGV [#730]                        [F26]
   11  bnr=bnrinit(bnfinit(y^2-229,1),5,1); rnfkummer(bnr,,3) --> oo loop [F30]
   12- missing subfields in nfsubfields [#744]                            [F43]
   13- infinite loop in nffactor [#751]                                   [F47]
   14- eint1(10,10) completely wrong                                      [F49]
   15- x/(x+2) * Mod(1,2) -> Mod(1, 2)*x/(x + 2)  [should be Mod(1,2)]    [F56]
   16- substpol(x/(y^3*x+1),y^2,1) -> x/(y^3*x + 1) [should be x/(y*x+1)] [F57]
   17- hilbert(1,1,1) --> oo loop [#770]                                  [F61]
   18- nffactor(nfinit(y),x*y)                            [F60,F61,F62,F63,F64]
       nffactor(nfinit(y),x^2*y+1)
       nffactor(nfinit(y),x^2*y)  --> errors or SEGV. [#769]
       nffactor(y^2+1, x+1/2) --> x+1/2 [ not integral ]
       nffactor(y^2+1,x*y+1) --> x*y + 1 [ not normalized, missing POLMOD ]
       nffactor(y^2+1,x^2+2*y+1) --> x^2+2*y+1 [ missing POLMOD ]
   19- p=x^4-x^3+x^2-x+1; tnf = thueinit(p); thue(tnf,1) -> missing solutions
       (could occur whenever p has no real root) [#764]                   [F67]
   20- missing GC in numtoperm()                                          [F68]
   21- harmless uninitialized memory read in mulrrz_i()                   [F69]
   22- minor inaccuracy in GMP kernel (affir)                             [F72]
   23- matsolve([1;2;3],[1,2,4]~) --> [1]~  (should raise an error)       [F76]

  Added
    1- GP function version() [ current version number ]                   [A20]

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Done for version 2.3.3 (released 22/12/2007):
  Fixed
    1- remove hack in internal function 'readbin' that used the clone bit as
    a special marker hence returned an invalid object, which could be
    propagated to higher level public functions.                          [F4]
BA  2- bitnegimply(1,2^65) returned a wrong result [#560]                 [F6]
    3- nfeltreduce(nf,t_POLMOD,id) didn't work [#558]                     [F7]
    4- segv in polrootspadic / repeated roots in FpX_root (only in deg 2) [F9]
    5- numbpart inaccurate (and slow). E.g numbpart(52602) off by 1 [#557][F12]
    6- remove type assumptions in mulcxI, mulcxmI: unsafe [#516]]         [F14]
    7- (y/x)/(x^0/y) not correctly simplified                             [F16]
    8- sumpos(x=1,1/x^2,1) had the wrong sign [#587]                      [F21]
    9- default(compatible,3); default(compatible,1) --> case no longer    [F47]
       taken into account (I = i) [#629]       
   10- possible oo loop in _isprincipal [ precision was supposed to increase,
       but could in fact remain the same ]                                [F81]
   11- agm(-1, 1+1e-55) --> oo loop, agm(-1,1) --> wrong result           [F27]
   12- read("a b.gz")   \\ filename contains a space                      [F34]
       gzip: ./a.gz: No such file or directory
   13- quadgen([]) --> incorrect object [#606]                            [F31]
   14- hyperu(0,1,1) --> oo loop [#608]                                   [F32]
   15- fix headers so that ulong is always defined                        [F33]
   16- gclone didn't work on t_INTs with lg > lgefint                     [F37]
   17- exp(2^200*I) --> catastrophic cancellation [cos/sin were OK] [#631][F45]
   18- [Configure] don't rely on $CC exit status, check whether a non-0 size
       executable is produced [ problems with tcc ]                       [F55]
   19- wrong Prototype for subgrouplist [ extra 'p' ]                     [F56]
   20- nf.codiff was only correct up to multiplication by some rational number
       [cf #510 & 2.4.1-F7]                                               [F58]
   21- typo in bnfisintnorm(): missed some solutions  [ couldn't find a unit
       of norm -1 even though one exists ]                                [F74]
   22- dilog(-1) [ more generally polylog of < 0 t_REALs ] should have
       0 imaginary part                                                   [F84]
   23- ix86, x86_64: missing earlyclobber constraint for addllx, subllx,  [F87]
       divll
   24- deriv(x/(x+1),y) --> invalid t_RFRAC with exact 0 numerator [#687] [F90]
   25- factornf(x^5+(-a+2)*x^4-a*x^3+(3*a-6)*x^2+(5*a-10)*x+(2*a-5), a^2-5)
       --> SEGV  [ not squarefree -> denominators creeping in ] [#708]   [F110]
BA 26- fix compilation problem with g++-4.2 and GMP                       [F94]
BA 27- missing GP2C descriptions for Pol and Polrev                      [F114]
   28- typo in FpM_gauss_pivot: FpM_rank, FpM_image, FpM_suppl, FpM_indexrank
       much slower than they should be. Same problem in FqM_gauss_pivot. [F113]

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Done for version 2.3.2 (released 28/03/2006):
[last column crossreferences current development release 2.4.1]
  Fixed
    1- [Cygwin] missing -L... -lgmp when compiling with gmp.              [F2]
    2- ispower(522^3) -> 0 [ looked like a 7th power to is_357_power(), which
       then forgot to test for cubes ] [#506]                             [F3]
    3- when nf.disc < 0, nf.diff was an incorrect PARI ideal [#510]       [F6]
    4- nf.codiff was only correct up to multiplication by some rational   [F7]
       number (a divisor of nf.disc) [#510]                              
    5- inaccuracy (>= 2ulp) in [cached] log(2) [#498]                     [F8]
    6- exp, sinh, asinh, tanh, atanh were inaccurate near 0               [F9]
    7- [GMP kernel] forvec(x=[[-1,0]],print(x)) --> error [#509]          [F10]
       [ 'resetloop' failed when passing through '0' ]
    8- nfbasistoalg(nfinit(y),x) created an invalid t_POLMOD              [F11]
    9- incorrect result in ZX_resultant (accuracy loss computing bound)   [F12]
   10- [Configure] gcc-specific flags were used on linux/freebsd/cygwin,
       even when __gnuc__ was unset                                       [F14]
   11- factor( pure power FqX ) --> SEGV                                  [F15]
   12- [GMP kernel] polrootsmod(f, 4) --> wrong result [ low level t_INT  [F16]
       manipulation not using the int_* macros ]
   13- polrootspadic(f, 2, r) --> some roots would be found twice         [F17]
       [ due to FpX_roots(f, 4) called ]  [#521]
   14- ??sumalt doesn't compile: in GPHELP, treat \ref in verbatim        [F18]
       mode[#518]
   15- matinverseimage returned [;] when no pre-image exists. Conform to  [F20]
       the docs: "an empty vector or matrix", depending on the input types.
   16- 3.5 % 2 --> error [ should be 0.5 ]                                [F22]
   17- sin(1/10^100) --> 0e-28 [ also affected cos,tan,cotan ]            [F23]
   18- check that k >= 0 in thetanullk [#531]                             [F26]
   19- isprime(-2,1) returned 1                                           [F27]
   20- Fix 'Not enough precision in thue' error                           [F28]
BA 21- [OS X] Fix kernel detection on x86_64-darwin                       [F29]
BA 22- [Configure] spectacular failure to recognize gcc under some locales[F34]
   23- polredabs(x^8+2*x^6-5*x^4+78*x^2+9) was incorrect [ missed
       x^8+6*x^6-x^4+54*x^2+25 due to incorrect "skipfirst" ]             [F35]
   24- typo in resmod2n (specific to GMP kernel) [#546]                   [F36]
   25- nfmodprinit could create FpX's which were not reduced mod p        [F40]
   26- O(x^3)^(1/2) was O(x^2) instead of O(x)                            [F41]
   27- substpol(x^-2+O(x^-1),x^2,x) --> error [#555]                      [F43]

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Done for version 2.3.1 (released 06/10/2006):
[last column crossreferences current development release 2.4.0]
  Fixed
    1- typo in remiimul: wrong result in a "failsafe" branch (hardly ever
       taken). May affect modular powering /Fp if p >> 10^1300 [#457]
       Also affects the gmp kernel, for p >> 10^5800.                     [F1]
    2- inconsistent return type in nffactormod [#460]                     [F3]
    3- gammah(1+O(3^5)) --> incorrect type in ggamd.                      [F8]
    4- invalid read in copy_leaf [ SEGV in some libc ]                    [F9]
BA  5- ellheight short help was accidentally truncated.                   [F10]
    6- substpol(1+O(x^(2*n)),x^2,x) --> 1+O(x) instead of 1+O(x^n) [#470] [F11]
    7- [OS X] only use -no-cpp-precomp with Apple cc                      [F12]
    8- divrem(x,x,y) --> [1/y*x, 0]                                       [F13]
    9- typo in qflll: in rare cases (exact input+floating point computation+
       precision increase in last-but-1 step), the returned base change is not
       properly updated in last iteration --> basis not LLL-reduced       [F15]
   10- (1+x)/(1-x)/(1+x)^2 not simplified [#472,#473]                 [F14,F16]
   11- ispower(x^k, k) would answer 0 for some x and k in {3,5,7}  [#476] [F17]
   12- content(t_MAT with exactly 1 col) gave a wrong result              [F18]
   13- rare bug in red_montgomery (returning 0 with + sign, an incorrect object)
       [ polrootspadic(x^11+x,11,10) --> corrupts gen_0 ]                 [F20]
   14- qfbsolve(Qfb(1,2,10),5) --> [0,0; 0,0] instead of 0 [#479]         [F21]
   15- ispower(0, n, &z) would not set z                                  [F22]
   16- wrong result in conversion t_QUAD -> t_PADIC whenever disc = 1 (4) [F23]
JJ 17- gaddgsz macro was wrong [#481]                                     [F24]
   18- polsubcyclo(p^k, q) --> wrong result [ gener_Fl(p^k) can't handle
       k > 1, use gener instead ] [#480]                                  [F25]
SC 19- [GMP] mp_set_memory_functions was called with an incompatible realloc 
       function. [#484]                                                   [F26]
AM 20- [TeXmacs] typo in texmacs_output: x --> <bluex> [#491]             [F30]
BA 21- bnrstark prototype code was non-standard.                          [F34]
   22- rnfkummer(,,degree) often found too many fields [#482]             [F35]
   23- missing GC in forvec(,,2)                                          [F37]
   24- loss of accuracy in p-adic ellinit: wrong digits and spurious errors:
       E.g i = 5; ellinit([1, -1, 1, -1, -14]*(1+O(17^i)))                [F36]
   25- possible stack corruption in charpoly(,1)                          [F39]
   26- Ser(x) raised an error [#499]                                      [F40]


  Changed
    1- remove CPP from Configure tests                                    [C3]