pari-2.0.14 bugs

[Igor Schein <igor@txc.com>]

Hi,

here's my current list of pari bugs.  Some of them were found after
the release:

1) quadclassunit(-148-16!)			\\ Infinite loop
2) nfgaloisconj(x^3+2,4)			\\ Infinite loop
3) poldiscreduced(polzagier(4,0))			\\ Error
4) bnfclassunit(x^2-x-1)				\\ Error
5) bnfinit(polzagier(4,0),3)				\\ SEGV
6) nf=nfinit(x^2+1);idealred(nf,idealfactor(nf,2))	\\ SEGV

Bugs 2 and 6 are bad input bugs, hence of low priority.

And here're a few things related to help:

1) ?bnfinit mentions non-existing flags 4,5 and 6
2) cross-reference for quadclassunit() in the manual refers to page
63, but the function is in fact described on page 64.  There might be
other *drifted* references.
3) I think it's a good idea to mention a version of readline in start
up message.  For example,

    GP/PARI CALCULATOR Version 2.0.14 (alpha)
  UltraSparc (MicroSparc kernel) 32-bit version
 (readline v4.0 enabled, extended help available)


This can be achieved by running a small program which would print out
variable rl_library_version ( as Gerhard pointed out ), or say that
the version is 1.0 or earlier ( as Karim pointed out ).

4) I discovered a weird completion bug on Linux and Cygwin.  When I
hit <TAB> to complete the 10 character long function
(e.g. vecextract(), bnfcertify() etc. ), it either adds '^y' character
in the and, or segfaults.  Sparc doesn't exhibit this bug.  Neither
Karim nor Gerhard were able to reproduce.  I wonder if anyone else
can.  The easiest thing to try to reproduce is to type 'bnfce' and
then hit <TAB> 

Thanks

Igor