-
Notifications
You must be signed in to change notification settings - Fork 1.5k
[Consolidated] issues affecting tactic.default_tactic=smt sat.euf=true #5532
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
src/sat/smt/bv_internalize.cpp:249
|
"Failed to validate" and leak
|
heap-use-after-free at ../src/sat/smt/q_mam.cpp:2939
|
src/sat/smt/arith_solver.cpp:594
|
src/sat/sat_gc.cpp:461
|
../src/sat/sat_gc.cpp:479
|
SEGV at ./src/ast/euf/euf_enode.h:198
|
Invalid model:
|
Failed to verify:
|
SEGV ../src/sat/smt/arith_solver.cpp:626:
|
Solution unsoundness:
|
Invalid model:
|
SEGV ../src/ast/ast.h:501:
|
Invalid model (might be related to #5532 (comment)):
|
arith.reflect is really an internal parameter. I hid it. The bug was also fixed. |
Assertion violation:
|
SEGV ../src/ast/ast.h:501:
|
Invalid model (might be related to #5532 (comment)):
|
Assertion violation:
|
Assertion violation:
|
Invalid model:
|
Assertion violation:
|
Assertion violation:
|
Assertion violation:
|
Note that parallel search requires copying the solver state around. This code path is not well tested yet, but not too relevant for sequential mode. Therefore, if you start setting parallel.enable=true you are likely (as demonstrated) to produce more crashes. I might be able to investigate based on running with the parallel mode on, but setting up self-contained unit testing is going to be easier to work with to start with. |
Assertion violation:
|
disabling complete const rewriting (temporarily) as it can loop
Assertion violation:
|
Assertion violation:
|
Assertion violation:
|
Assertion violation:
|
SEGV ../src/sat/smt/dt_solver.cpp:718:
|
Assertion violation:
|
Assertion violation:
|
Refutation unsoundness:
|
Refutation unsoundness:
I will go over other above, but skip the NLSAT related bugs. The root cause is almost for sure outside of the new code and related to already filed bugs. Root causing nlsat related bugs for the new code is double overhead. |
a few of these come from the same regression introduced a few days before filing. remaining two sad faces are probably more difficult so focus is on those, but context switching with polysat branch also this week. one of the parallel.enable bugs was fixed, but others have different root causes. the nlsat (non-linear bugs) have already simpler repros in long standing open bugs independent of sat.euf. |
one of the remaining bugs exposed a subtle miss for internalization the parallel.enable bugs pointed to basic mixups in solver copy functionality. |
Hi, for the following formula,
z3 af5c6e4
The text was updated successfully, but these errors were encountered: