Bug #30309 mysql_truncate() does not inform engine that the recreated table is temporary
Submitted: 8 Aug 2007 14:39
Reporter: Guilhem Bichot Email Updates:
Status: Verified Impact on me:
None 
Category:MySQL Server: DDL Severity:S3 (Non-critical)
Version: OS:Linux
Assigned to: CPU Architecture:Any
Triage: Triaged: D3 (Medium)

[8 Aug 2007 14:39] Guilhem Bichot
Description:
When a table is first created, with CREATE TEMPORARY TABLE, the engine is
passed the flag HA_LEX_CREATE_TMP_TABLE.
If this table is later truncated with TRUNCATE TABLE, the engine is not passed this flag (this flag is not filled by update_create_info_from_table(), which would have a hard time filling it as ha_create_table() passes to this function a TABLE_SHARE which is built with init_tmp_table_share() and thus has
INTERNAL_TMP_TABLE in TABLE_SHARE::tmp_table).
So, the engine's create() is not passed HA_LEX_CREATE_TMP_TABLE when in TRUNCATE. This is a potential issue, as we can see that ha_innodb::create() and ha_myisam::create() test this flag.
No visible issue has been identified so far, though.

How to repeat:
no testcase.
[8 Aug 2007 14:49] Guilhem Bichot
note that when this is fixed, mi_create() will have to be fixed too, as it does
  if (flags & HA_CREATE_TMP_TABLE)
  {
    options|= HA_OPTION_TMP_TABLE;
    create_mode|= O_EXCL | O_NOFOLLOW;
  }
and thus assumes that TRUNCATE cannot pass HA_CREATE_TMP_TABLE.
We might need a flag "this is TRUNCATE, not CREATE" in HA_CREATE_INFO to solve this cleanly
[12 Oct 2007 17:19] Konstantin Osipov
Guilhem, what do you suggest is done for this bug report?
I welcome you to submit the patch.
[15 Oct 2007 9:14] Guilhem Bichot
Hi Konstantin. I don't think you need to work on this now, as no concrete issue with some engine has been identified. Maybe "to be fixed later"?
[27 Oct 2009 17:32] Valeriy Kravchuk
I think Bug #48350 can be related.