[Libreoffice-bugs] [Bug 50849] SIGSEGV: ODBC to PostgreSQL, renaming column in SELECT list

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=50849

Lionel Elie Mamane <lionel at mamane.lu> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|REOPENED                    |ASSIGNED
         AssignedTo|libreoffice-bugs at lists.free |lionel at mamane.lu
                   |desktop.org                 |

--- Comment #13 from Lionel Elie Mamane <lionel at mamane.lu> 2012-07-10 14:13:13 PDT ---
(In reply to comment #11)
> (In reply to comment #10)

>> Does this suggest that it would be good to look into why LO is doing the
>> call with null table name?  Or into why LO is doing this extra call at
>> all?

> Ah yes, I had missed that; I was so focused on the CatalogName and SchemaName
> NULL vs empty string.

> Fair enough; LibO should not call it with empty string for TableName, and this
> should be fixed.

This *also* has its roots in an ODBC driver problem. The structure is as such:

For each column in the query, LibO:

 - asks the driver "which table does this column come from" by calling

   SQLColAttribute(m_aStatementHandle,
                   1,
ident,                                                         |
   |75                                          (SQLPOINTER)pName,             
                                             |
   |76                                          nRealLen,                      
                                             |
   |77                                          &nRealLen,                     
                                             |
   |78                                          NULL                           
                                             |
   |79                                          );                             
                                             |

 - store the answer in sTableName
 - calls SQLSpecialColumns(..., sTablename, ...)

> 
> > A funny thing happened on the way to the crash: In PGAPI_SpecialColumns,
> > info.c line 2783 detects the null table name, so line 2785 does
> > SC_set_error and then line 2786 returns SQL_SUCCESS.  SQL_SUCCESS sounds
> > to me like an outright lie.  If in SQLSPecialColumns I fudge the return
> > value to SQL_ERROR, then LibreOffice ...
> 
> > (*) Goes on to call SQLSpecialColumns with good catalog, schema, and
> >     table names.  This call looks very much like what I remember from
> >     the successful query.
> 
> > (*) Goes on to display the table data.
> 
> > The SQL_SUCCESS sounds to me like a candidate for reporting upstream, as
> > does the segfault itself.  LEM, do you agree?
> 
> Yes, the segfault, with the explanation about SQL_SUCCESS/SQL_ERROR should be
> reported upstream. Could you do that (and CC me)? Thanks.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.