MDEV-39194 SUM(b'1100') and SUM(DISTINCT b'1100') return different numeric results for the same BIT literal (e.g. 12 vs 9)

[DerZc]

SUM(DISTINCT b'1100') returns 9 instead of 12. Bit and hybrid hex literals inherit the default numeric DISTINCT temporary field, which uses their byte length as a DOUBLE(M,0) display width and clips the stored value.

Bug report: https://jira.mariadb.org/browse/MDEV-39194

Changes

• Override Type_handler_hex_hybrid::make_num_distinct_aggregator_field to use unsigned integer storage for numeric DISTINCT aggregation of bit and hybrid hex literals.
• Preserve argument-based field selection for other types. In particular, decimal DISTINCT inputs retain their original scale: AVG(DISTINCT b / a) in main.distinct_notembedded continues to return 0.61110000.
• Extend the existing main.sum_distinct regression with zero, byte boundaries, hybrid hex literals, the unsigned 64-bit boundary, duplicate inputs, AVG, COUNT, and the decimal precision control.

Validation

Built against 11.4 at d10e5d726799b1cd57cc866f40aad68c720803da with GCC 13 in Release mode.

• The expanded main.sum_distinct regression fails on the unchanged base and passes with this fix.
• Reproduced the previous PR's main.distinct_notembedded failure locally: 0.61111111 instead of 0.61110000. This matches all eight failed Buildbot configurations and AppVeyor; the corrected implementation passes without changing that test's expected output.
• All 11 selected MTR tests passed: main.sum_distinct, main.distinct_notembedded, main.distinct, main.func_group, main.group_by, main.select, main.type_bit, main.type_bit_innodb, main.type_decimal, main.type_float, and main.type_hex_hybrid.
• main.sum_distinct and main.distinct_notembedded also passed with --ps-protocol and with --view-protocol, for 15 successful test executions overall.
• No test-state cleanup failures in these local runs. git diff --check passed.

The full regression suite and Windows builds were not run locally.

[CLAassistant]

All committers have signed the CLA.

[gkodinov]

Thank you for your contribution! This is a preliminary review.

To get to the final review:

• squash the two commits into one
• make sure there's no failing tests in the build bot runs.

[gkodinov]

Thanks! LGTM. Please stand by for the final review.

[raghunandanbhat]

thanks @DerZc for the fix. comments inline.
Please re-write commit message in this format-

MDEV-xxxx: <Titile>

Problem: < Short description of the issue >

Fix: <Short description of your fix>

[raghunandanbhat]

I wouldn't use the hard-coded type_handler_ulonglong here.
Any specific reason why we can't just return a new Field_longlong() from this function? (making sure it is always unsigned, ofc). for ex:

Field *
Type_handler_hex_hybrid::make_num_distinct_aggregator_field(
    MEM_ROOT *mem_root, const Item *item) const
{
  /*
    Hex and bit literals have unsigned integer values in numeric aggregates.
    Their max_length is a byte count, not a DOUBLE(M,0) display width.
  */
  return new (mem_root)
          Field_longlong(NULL, item->max_length,
                         (uchar *) (item->maybe_null() ? "" : 0),
                         item->maybe_null() ? 1 : 0, Field::NONE,
                         &item->name, 0, true /* unsigned */);
}

[raghunandanbhat]

please add end of test marker here