Trac #29445: Reject broken system PARI 2.11.3 and re-enable PARI libr…

…ary test in spkg-configure

There are several failing doctests depending on the PARI version of the
system, e.g.

{{{
File "src/sage/rings/number_field/number_field_ideal.py", line 1829, in
sage.rings.number_field.number_field_ideal.NumberFieldFractionalIdeal.fa
ctor
Failed example:
    fact = F.factor(2); fact
Expected:
    (Fractional ideal (2*a^2 + 1))^2 * (Fractional ideal (-2*a^2))
Got:
    (Fractional ideal (-2*a^2 - 1))^2 * (Fractional ideal (2*a^2))
}}}

see e.g.

[https://github.com/mkoeppe/sage/runs/542655863]

This is the case because debian bullseye uses pari 2.11.3 (instead of
pari 2.11.3). Many instances seem to fail, just because the choice of
generators has changed. However, there is at least one thing really
strange when upgrading to pari 2.11.3:

{{{
sage: y = var('y')
sage: pK = pari.bnfinit(y^3+3)
sage: pK.bnfisunit(pK.bnfunit()[0])
[1, Mod(1, 2)]~
}}}

URL: https://trac.sagemath.org/29445
Reported by: gh-kliem
Ticket author(s): Jonathan Kliem, Matthias Koeppe
Reviewer(s): Matthias Koeppe, Jonathan Kliem

build/pkgs/pari/spkg-configure.m4

@@ -9,7 +9,6 @@ SAGE_SPKG_CONFIGURE([pari], [
         sage_spkg_install_pari=yes
     else
       AC_MSG_RESULT([no])
-      AC_MSG_CHECKING([installing PARI/GP packages? ])
     AC_PATH_PROG([GP], [gp])
     if test x$GP = x; then dnl GP test
         AC_MSG_NOTICE([gp is not found])
@@ -74,10 +73,20 @@ SAGE_SPKG_CONFIGURE([pari], [
             AC_MSG_NOTICE([Otherwise Sage will build its own pari/GP.])
             sage_spkg_install_pari=yes
         fi
+        AC_MSG_CHECKING([whether bnfisunit bug of pari 2.11.3 is fixed])
+        bug_check=$(echo "bnf = bnfinit(y^4-y-1); bnfisunit(bnf,-y^3+2*y^2-1)" | $GP -qf 2>> config.log)
+        expected="[[0, 2, Mod(0, 2)]]~"
+        if test x"$bug_check" = x"$expected"; then
+           AC_MSG_RESULT([yes])
+        else
+           AC_MSG_RESULT([no; cannot use system pari/GP with known bug])
+           AC_MSG_NOTICE([Upgrade your system package and reconfigure.])
+           AC_MSG_NOTICE([Otherwise Sage will build its own pari/GP.])
+           sage_spkg_install_pari=yes
+        fi
     fi dnl end GP test
 
-      if test x$sage_spkg_install_pari = x; then dnl main PARI test
-        AC_MSG_RESULT([no])
+      if test x$sage_spkg_install_pari = xno; then dnl main PARI test
         AC_CHECK_HEADER([pari/pari.h], [], [sage_spkg_install_pari=yes])
         dnl matpermanent appears in pari 2.11
         AC_SEARCH_LIBS([matpermanent], [pari], [
@@ -124,9 +133,7 @@ SAGE_SPKG_CONFIGURE([pari], [
 ], [], [], [
     if test x$sage_spkg_install_pari = xyes; then
         AC_SUBST(SAGE_PARI_PREFIX, ['$SAGE_LOCAL'])
-        AC_MSG_RESULT([using Sage's pari SPKG])
     else
        AC_SUBST(SAGE_PARI_PREFIX, [''])
-       AC_MSG_RESULT([using pari/gp from the system])
     fi
 ])