Fix tagged pointer small integer bounds for 32/64-bit platforms - #8554

Open

Nuno Lopes (nunoplopes) with Copilot wants to merge 58 commits into

copilot/fix-tagged-pointer-bug

Commits on Feb 17, 2026

Initial plan

Copilot
authored and
Nuno Lopes (nunoplopes)
committed
d8abd04
mpz: use pointer tagging to save space (#8447)

Nuno Lopes (nunoplopes)
and
Copilot
committed
ac1bef0
Fix syntax errors in mpz.h

Copilot
and
Nuno Lopes (nunoplopes)
committed
cb3b860
Implement proper bounds checking for small integers based on platform pointer size

Copilot
and
Nuno Lopes (nunoplopes)
committed
6faaf86
Revert deallocation in set methods and restore indentation in gcd function

Copilot
and
Nuno Lopes (nunoplopes)
committed
b8ed45b
Fix undefined behavior in SMALL_INT_MAX/MIN constants by using unsigned arithmetic

Copilot
and
Nuno Lopes (nunoplopes)
committed
4dc74e2
Add assertion to set() method and simplify get_sign_cell by removing redundant SMALL_INT_MIN case

Copilot
and
Nuno Lopes (nunoplopes)
committed
dcfeb55
Remove nonsensical assertion and revert to simpler shift for SMALL_INT constants

Copilot
and
Nuno Lopes (nunoplopes)
committed
5edef9e
Fix uses of value() that should be value64() for 64-bit small integers

Copilot
and
Nuno Lopes (nunoplopes)
committed
76dde47
Fix more uses of value() that should use value64() for 64-bit correctness

Copilot
and
Nuno Lopes (nunoplopes)
committed
a5e1fa9
Fix machine_div2k to not truncate when result doesn't fit in small (add assertion instead)

Copilot
and
Nuno Lopes (nunoplopes)
committed
10c82a3
Refactor decompose function and consolidate value()/value64() into single value() returning int64_t

Copilot
and
Nuno Lopes (nunoplopes)
committed
9bf691f
Fix remaining int truncation issues in is_power_of_two, power_of_two_multiple, and log2 for 64-bit platforms

Copilot
and
Nuno Lopes (nunoplopes)
committed
cd699ca
Fix undefined behavior when negating SMALL_INT_MIN in power_of_two_multiple

Copilot
and
Nuno Lopes (nunoplopes)
committed
4941ef1
Fix SMALL_BITS constant reference to use direct calculation instead

Copilot
and
Nuno Lopes (nunoplopes)
committed
227e48c
Add out-of-line definitions for static constexpr members to fix ODR violations

Copilot
and
Nuno Lopes (nunoplopes)
committed
45f5209
fixes
Nuno Lopes (nunoplopes)
committed
8562788
fix
Nuno Lopes (nunoplopes)
committed
10f986e
fix
Nuno Lopes (nunoplopes)
committed
67ce05c
fix
Nuno Lopes (nunoplopes)
committed
1c5cbcb
fix
Nuno Lopes (nunoplopes)
committed
9448f67
fix
Nuno Lopes (nunoplopes)
committed
88242fb
fix
Nuno Lopes (nunoplopes)
committed
5f429d1
fix
Nuno Lopes (nunoplopes)
committed
8ce362a
fix
Nuno Lopes (nunoplopes)
committed
e9556c4
fix
Nuno Lopes (nunoplopes)
committed
92158ac
fix
Nuno Lopes (nunoplopes)
committed
4cdbe22
fix
Nuno Lopes (nunoplopes)
committed
dad012a
fix
Nuno Lopes (nunoplopes)
committed
0a22332
fix
Nuno Lopes (nunoplopes)
committed
bd993ca
fix
Nuno Lopes (nunoplopes)
committed
ac748d9
fix
Nuno Lopes (nunoplopes)
committed
1a189eb
fixes
Nuno Lopes (nunoplopes)
committed
fa73827
Fix invalid zero representation in set_big_i64 and set_big_ui64

Copilot
and
Nuno Lopes (nunoplopes)
committed
04f07fe
Revert "Fix invalid zero representation in set_big_i64 and set_big_ui64"

Copilot
and
Nuno Lopes (nunoplopes)
committed
957ba91
Fix set_big_i64/set_big_ui64 zero handling: zero is always represented as small

Copilot
and
Nuno Lopes (nunoplopes)
committed
cc7902a
Revert "Fix set_big_i64/set_big_ui64 zero handling: zero is always represented as small"

Copilot
and
Nuno Lopes (nunoplopes)
committed
c01062e
fixes
Nuno Lopes (nunoplopes)
committed
4a8aeb0
fixes
Nuno Lopes (nunoplopes)
committed
6db067c
fixes
Nuno Lopes (nunoplopes)
committed
fdc5793
fixes
Nuno Lopes (nunoplopes)
committed
e6607df

Commits on Jul 15, 2026

Start resolving merge conflicts with master branch
Copilot
authored
dcf81a2
Merge master branch and resolve conflicts
Copilot
authored
468a1cb
Fix Windows ARM64 overflow detection build
Copilot
authored
685ca67
Refine MSVC ARM64 multiplication overflow fallback
Copilot
authored
c1c01b4
Potential fix for pull request finding

Nikolaj Bjorner (NikolajBjorner)
and
Copilot
authored
20bba3e
Document MSVC ARM64 magnitude extraction
Copilot
authored
4c8c511
Fix MSVC mul_overflows sign extension check
Copilot
authored
e22487e
Clarify fallback and unsupported compiler error
Copilot
authored
b778291
Handle ARM64EC in MSVC mul_overflows path
Copilot
authored
a972537
Polish MSVC overflow fallback implementation
Copilot
authored
e3cfa56
Clarify MSVC target-specific overflow paths
Copilot
authored
3b28cc9
Fix x86 Windows build by guarding _mul128 to x64 only
Copilot
authored
efd31bd
Fix mpz self-assignment in set
Copilot
authored
96128e5
Handle mpz shared-self assignment in set
Copilot
authored
2fd3a28

Commits on Aug 22, 2026