In some cases, Z3_algebraic_get_i() returned 0. For example, in the following
Python snippet, the last assert would fail:
import z3
x = z3.Real('x')
s = z3.Solver()
s.add( (x * x) - 2 == 0, x <= 0)
s.check()
val_x = s.model().get_interp(x)
assert val_x.index() == 1
The problem was that `algebraic_numbers::manager:👿:get_i()` did not
check whether the root index was properly initialized.
This commit fixes this issue by checking whether root index is initialized
the same way various other routines do.
Fixes issue #5807.
Signed-off-by: Jan Vrany <jan.vrany@labware.com>
* remove the bound on total iterations in simplex
Signed-off-by: Lev Nachmanson <levnach@hotmail.com>
* remove unncesseray checks in get_freedom_interval_for_column()
Signed-off-by: Lev Nachmanson <levnach@hotmail.com>
* fix the build of test-z3
Signed-off-by: Lev Nachmanson <levnach@hotmail.com>
* Revert "remove unncesseray checks in get_freedom_interval_for_column()"
This reverts commit 6770ed85e3.
* optimize get_freedom_interval_for_column() for feasible case
Signed-off-by: Lev Nachmanson <levnach@hotmail.com>
* add function lar_solver::status_feasible
Signed-off-by: Lev Nachmanson <levnach@hotmail.com>
* rename status_is_feasible() to is_feasible()
Signed-off-by: Lev Nachmanson <levnach@hotmail.com>
* fix the linux build
Signed-off-by: Lev Nachmanson <levnach@hotmail.com>
there are some different sources for the performance regression illustrated by the example. The mitigations will be enabled separately:
- m_bv_to_propagate is too expensive
- lp_bound_propagator misses equalities in two different ways:
- it resets row checks after backtracking even though they could still propagate
- it misses equalities for fixed rows when the fixed constant value does not correspond to a fixed variable.
FYI @levnach
bugs in:
- rewriting of 0-ary expressions was incomplete
- sharing annotations when a node has two theories attached it is shared
- sharing of const of an array
Remove unreadable part of pretty printer for lp solver.
strict inequality (over reals) require solving for least-upper/greatest-lower bounds that may coincide with non-strict inequalities (be epsilon stronger). Instead of using the coefficient 'a' to turn the inequality into an equality, add the slack value as a constant.
