Canonicalize commutative operations before simplification

[twizmwazin]

Summary

Sorts the arguments of commutative operations by hash in simplify() so that expressions differing only in operand order canonicalize to the same AST:

• x + y and y + x produce the same expression
• And(a, b) and And(b, a) produce the same expression

This improves expression sharing and simplification hit rates. Ported from the feat/better-canonicalization branch (just the commutative-canonicalization change — independent of #725).

Change

if op in operations.commutative_operations:
    args = tuple(sorted(args, key=lambda x: x.hash()))

plus the from claripy import operations import.

Testing

Verified canonicalization behavior locally (x + y is y + x, And(x==1, y==2) is And(y==2, x==1)). Running the full angr test suite against this branch is in progress; will follow up with results.

🤖 Generated with Claude Code

[twizmwazin]

⚠️ angr test results: this change causes 12 regressions

Ran the full angr test suite against this branch and triaged each failure against master (baseline):

12 tests pass on baseline but fail with this change — caused by the commutative arg reordering:

• tests/analyses/cfg/test_jumptables.py::TestJumpTableResolver::test_amd64_win_jumptable_with_secondary_redirections
• tests/analyses/cfg/test_jumptables.py::TestJumpTableResolver::test_s390x_cfgswitches
• tests/analyses/cfg/test_cfgfast.py::TestCfgfast::test_cfg_switches_s390x
• tests/analyses/test_calling_convention_analysis.py::...::test_cdecl_nonconsecutive_stack_args_2
• tests/analyses/test_calling_convention_analysis.py::...::test_cdecl_nonconsecutive_stack_args_3
• tests/engines/test_unicorn.py::TestUnicorn::test_similarity_fauxware
• tests/exploration_techniques/test_oppologist.py::TestOppologist::test_fauxware_oppologist
• tests/exploration_techniques/test_stochastic.py::TestRun::test_stochastic
• tests/procedures/libc/test_string.py::...::test_strstr_sym_haystack_sym_needle
• tests/sim/test_fauxware.py::TestFauxware::test_fastmem
• tests/test_cli.py::TestCommandLineInterface::test_base_addr_dec
• tests/test_cli.py::TestCommandLineInterface::test_decompiling

2 pre-existing flakies (fail on baseline too, float-precision asserts, unrelated): test_manyfloatsum_symbolic_i386, test_manyfloatsum_symbolic_x86_64.

Several angr analyses (jumptable resolution, calling-convention recovery, decompilation) pattern-match on expression argument order, so reordering commutative operands during simplification breaks them. This likely needs corresponding fixes in angr before it can land — flagging rather than merging as-is.

[angr-bot]

Corpus decompilation diffs can be found at angr/dec-snapshots@master...angr/claripy_726