Found and fixed several bugs in the MPFI interval arithmetic integration and MPFR precision handling.

Type truncation in FGenMPFComplex

• Constructor parameter was uint16_t but member is mpfr_prec_t (long). Values like --mpfrprec 100000 silently truncate to 34464 via 100000 & 0xFFFF.

// Before
explicit FGenMPFComplex(uint16_t _prec) : prec(_prec) {}
// After
explicit FGenMPFComplex(mpfr_prec_t _prec) : prec(_prec) {}

Wrong precision in digit_precision_mpfi()

• Used mpfr_init() (MPFR default 53-bit) for temporaries extracting interval endpoints. High-precision intervals lose precision, making endpoints appear identical and reporting bogus maximum digit precision. Now uses mpfr_init2() with the interval's actual precision.

Missing --mpfrprec validation

• No lower bound check. MPFR requires minimum precision of 2 bits; passing 0 or negative causes undefined behavior in mpfr_init2/mpfi_init2.

Default precision in compute_collision_prob()

• All MPFR temporaries used mpfr_init_set_*() (53-bit default) instead of explicit 256-bit precision matching the rest of the collision probability and log-printing code.

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.