Incorrect conversion from ULongLong to Num

[cbandy]

1. mysql_affected_rows returns an unsigned long long, but affected_rows uses INT2NUM
2. mysql_num_rows returns an unsigned long long, but
   • list_dbs and list_tables store it in an unsigned int
   • num_rows uses INT2NUM
3. mysql_stmt_affected_rows returns an unsigned long long, but stmt_affected_rows uses INT2NUM
4. mysql_stmt_num_rows returns an unsigned long long, but stmt_num_rows uses INT2NUM

These should use ULL2NUM, I believe, like insert_id does.

[luislavena]

@cbandy do you have any test that we can add to ensure this works as expected?

[cbandy]

I don't unfortunately. It may be possible to test both num_rows using a row generator, but affected_rows would require a table with that many rows.

Perhaps with a row generator in place, it wouldn't be hard to CREATE TABLE x AS SELECT ... FROM row_generator?

[cbandy]

I tried to write some tests, but they were too expensive. I killed the Mysql#affected_rows test at 7GiB of HDD space, and I killed the Mysql::Result#num_rows test at 6GiB of RAM and swap.

[cbandy]

What about num_rows?

[luislavena]

What about num_rows?

Kinda you forgot about that, right? was nowhere in your patch either :tongue:

Please open a separate issue and will work on that next weekend. Never encountered this issue before.

[cbandy]

I opened a new pull request and called it a "partial fix". I knew exactly what I was doing.

[luislavena]

Sorry, I'm not implying or blaming you.

Simply was not evident from the pull request or this first description.