Hi, Sachin! I suspect is_null() could be wrong here.
Please, add a test case where a table has a BEFORE UPDATE trigger and a long UNIQUE over a NOT NULL field. And see how cmp_binary_offset() is used in compare_record(), for example. On Oct 08, Sachin Setiya wrote: > revision-id: 56f2b3d12b6 (mariadb-10.4.11-385-g56f2b3d12b6) > parent(s): af834c218a5 > author: Sachin Setiya <[email protected]> > committer: Sachin Setiya <[email protected]> > timestamp: 2020-09-14 20:08:18 +0100 > message: > > MDEV-23264 Unique blobs allow duplicate values upon UPDATE > > Problem:- > We are able to insert duplicate value in table because cmp_binary_offset > is not able to differentiate between NULL and empty string. So > check_duplicate_long_entry_key is never called and we don't check for > duplicate. > Solution > Added a if condition with is_null() on field which can differentiate > between NULL and empty string. > > diff --git a/sql/handler.cc b/sql/handler.cc > index 40dea349272..71e6a904f23 100644 > --- a/sql/handler.cc > +++ b/sql/handler.cc > @@ -6699,8 +6699,13 @@ static int check_duplicate_long_entries_update(TABLE > *table, handler *h, uchar * > for (uint j= 0; j < key_parts; j++, keypart++) > { > field= keypart->field; > - /* Compare fields if they are different then check for duplicates*/ > - if(field->cmp_binary_offset(reclength)) > + /* > + Compare fields if they are different then check for duplicates > + cmp_binary_offset cannot differentiate between null and empty > string > + So also check for that too > + */ > + if((field->is_null(0) != field->is_null(reclength)) || > + field->cmp_binary_offset(reclength)) > { > if((error= check_duplicate_long_entry_key(table, > table->update_handler, > new_rec, i))) > Regards, Sergei VP of MariaDB Server Engineering and [email protected] _______________________________________________ Mailing list: https://launchpad.net/~maria-developers Post to : [email protected] Unsubscribe : https://launchpad.net/~maria-developers More help : https://help.launchpad.net/ListHelp
