[Bug gdb/34416] Incorrect loss of sign in 64-bit multiply (e.g.: DW_OP_mul in DWARF evaluation)

https://sourceware.org/bugzilla/show_bug.cgi?id=34416

--- Comment #4 from Tom Tromey <tromey at sourceware dot org> ---
Seems to be from

commit 303a881f8789733248f27af0c872d356a34be009
Author: Tom Tromey <[email protected]>
Date:   Wed Mar 1 15:13:21 2023 -0700

    Use gdb_gmp for scalar arithmetic

In particular this runs afoul of this code in gdb_mpz::read

  if (!unsigned_p)
    {
      /* The value was imported as if it was a positive value,
         as mpz_import does not handle signs. If the original value
         was in fact negative, we need to adjust VAL accordingly.  */
      gdb_mpz max;

      mpz_ui_pow_ui (max.m_val, 2, buf.size () * HOST_CHAR_BIT - 1);
      if (mpz_cmp (m_val, max.m_val) >= 0)
        mpz_submul_ui (m_val, max.m_val, 2);
    }

-- 
You are receiving this mail because:
You are on the CC list for the bug.