Details
Major Description
I will post a reduced testcase once I have it.
The symptom of the bug is
sql/sql_bitmap.h:77: ulonglong Bitmap<width>::bit_mask(uint) const [with
|
unsigned int width = 64; ulonglong = long long unsigned int; uint = unsigned int]: Assertion `n < width' failed.
|
The stack trace:
#3 0x00007ffff5221472 in __GI___assert_fail (assertion=0x555556c6cb4a "n < width", file=0x555556c6cb08 "/home/psergey/dev-git2/ MariaDBEnterprise-10.6.14/sql/sql_bitmap.h", line=77, function=0x555556c6d380 <Bitmap<64u>::bit_mask(unsigned int) const:: __PRETTY_FUNCTION__> "ulonglong Bitmap<width>::bit_mask(uint) const [with unsigned int width = 64; ulonglong = long long unsigned int; uint = unsigned int]") at assert.c:101
|
#4 0x0000555555d94d37 in Bitmap<64u>::bit_mask (this=0x7fff341d8918, n=64) at /home/psergey/dev-git2/MariaDBEnterprise-10.6.14/sql/ sql_bitmap.h:77
|
#5 0x0000555555e41d69 in Bitmap<64u>::set_bit (this=0x7fff341d8918, n=64) at /home/psergey/dev-git2/MariaDBEnterprise-10.6.14/sql/ sql_bitmap.h:109
|
#6 0x00005555564541d2 in Item_bool_func::get_mm_parts (this=0x7fff3419e690, param=0x7ffff40f3c50, field=0x7fff3407a2d0, type=Item_func::EQ_FUNC, value=0x7fff3418edd8) at /home/psergey/dev-git2/MariaDBEnterprise-10.6.14/sql/opt_range.cc:8659
|
#7 0x0000555556453d11 in Item_equal::get_mm_tree (this=0x7fff3419e690, param=0x7ffff40f3c50, cond_ptr=0x7fff3419e7c0) at /home/psergey/ dev-git2/MariaDBEnterprise-10.6.14/sql/opt_range.cc:8568
|
#8 0x00005555564530e5 in Item_cond_and::get_mm_tree (this=0x7fff3402a9c0, param=0x7ffff40f3c50, cond_ptr=0x7fff34185438) at /home/ psergey/dev-git2/MariaDBEnterprise-10.6.14/sql/opt_range.cc:8346
|
#9 0x0000555556446c01 in calculate_cond_selectivity_for_table (thd=0x7fff34000d78, table=0x7fff340578a8, cond=0x7fff34185438) at /home/ psergey/dev-git2/MariaDBEnterprise-10.6.14/sql/opt_range.cc:3517
|
#10 0x00007fff341c0d70 in ?? ()
|
#11 0x00007fff341c0d90 in ?? ()
|
#12 0x00007fff341c0db0 in ?? ()
|
#13 0x00007fff341c0dd0 in ?? ()
|
In the testcase I'm trying, calculate_cond_selectivity_for_table() is trying to get estimates for 168 columns:
Thread 14 "mysqld" hit Breakpoint 4, create_key_parts_for_pseudo_indexes (param=0x7ffff40f3c50, used_fields=0x7fff28057a78) at /home/ psergey/dev-git2/MariaDBEnterprise-10.6.14/sql/opt_range.cc:3183
|
(gdb) p keys
|
$54 = 168
|
the range optimizer can handle at most 64 indexes.
(This is all I see so far. There may be other issues)
Attachments
Activity
Slo we do need to change create_key_parts_for_pseudo_indexes after all? Ok let me try making a patch.
igor could you please review the latest patch: https://github.com/MariaDB/server/commit/2949128be1ffd7c340676618b83145f4f3f3929b
I believe it addressed all of your input. it's pushed into bb-10.6-MDEV-33314
I don't see any problems with the patch. Yet as it's a crashing bug the patch should be applied to 10.5 as well.
Notes for the changelog:
The server could crash if a query's WHERE clause has conditions over more than 128 (Enterprise Server) or 64 (Community Server) columns such that column statistics is available for them.
igor, look:
create_key_parts_for_pseudo_indexes() uses this piece of code to check whether to create pseudo indexes for columns:
for (field_ptr= table->field; *field_ptr; field_ptr++){Field *field= *field_ptr;if (bitmap_is_set(used_fields, field->field_index) &&is_eits_usable(field))parts++;}So are you suggesting that outside of create_key_parts_for_pseudo_indexes there should be identical code that walks through the columns, checks "if_eits_usable" and collects a bitmap of first 64 columns, then calls create_key_parts_for_pseudo_indexes().
Since you request not to touch create_key_parts_for_pseudo_indexes(), it will check "if_eits_usable" for each passed column again.
Is that what you would prefer to see implemented? Please clarify.