Development Log

Highlights and Annotations of CVS Check-ins

• 2005/11/01
  • Fixed memory leak Turns out ORAFREE() followed by repeated OCIEnvCreate * leaks memory. Code has been * modified so OCIEnvCreate is called only once
  • ORACLE Binary Float and Binary Double support added
• 2004/12/01
  • Changes to SQLBindCol (was incorrectly setting an internal flag to SQL_NULL_DATA) and SQLSetStmtAttr (was NOT resetting internals when SQL_ATTR_ROWS_FETCHED_PTR indicated it should). These changes make the driver work *much* better with libodbc++, so much that a new release is warranted.
• 2004/8/6 Changes contributed by Gary Houston, ghouston@intellecthr.com
  1. SQLBindParameter.c
     got rid of useless function _SQLBindParameter
     fixed behavior of Column size for various data types to conform to ODBC standard
  2. SQLExecDirect.c
     changes to work more correctly with ood_lex_parse
  3. diagnostics.h
     add error code sqlerror_07002, wrong number of bound parameters or conversion error in binding parameter
  4. oracle_functions.c oracle_functions.h
     lots of re-working of behavior for filling ir and ar descriptors when a statement handle is reused. Fixes a crash when number of columns changes in a re-used statement handle.
• 2004/6/11 Changes contributed by Steven Reynolds , sreynolds@bradmark.com
  1. Fixes to SQLColumns.c and oracle_functions.c to clean up some size/type mismatches that resulted in core dumps.
• 2003/12/15 Changes contributed by Steven Reynolds , sreynolds@bradmark.com
  1. SQLFreeHandle.c: Change order of frees in _SQLFreeHandle(). Free oci_stmt handle first before other oci handles. Oracle OCI code was touching freed memory.
  2. SQLGetConnectAttr.c: SQLGetConnectAttr() was setting the commit mode.
  3. SQLGetDiagRec.c: Remove call to ood_clear_diag in SQLGetDiagRec(). Otherwise client code was unable to get log messges.
  4. diagnostics.c: ood_post_diag() is allocates new error nodes, but code didn't set all fields. Change malloc to calloc.
  5. oracle_functions.c: ood_driver_prepare() allocated a new oci statement with out freeing the existing one.
  6. oracle_functions.c: ocivnu_sqlslong() passed to OCI code sizeof(long), but buflen was 4. This failed on Tru64 where sizeof(long) is 8. Change to pass min of these two values. Same in ocivnu_sqlulong().
• 2003/02/11 fixed a problem with array inserts, SQLSetStmtAttr and SQLGetStmtAttr for parameters SQL_ATTR_PARAMSET_SIZE and SQL_ATTR_PARAM_BIND_TYPE are now functional
• 2003/02/10 change off-by-one error in month i/o for dates. Thanks to Gordon.Read@ranplc.co.uk
• 2003/02/03 added some data types that SQLColumns can identify
• 2003/01/30 Fixes for SQLColAttributes and underlying oracle_functions.c to return correct values for most column types when given SQL_DESC_TYPE or SQL_DESC_TYPE_NAME inputs. Fixed unit tests to reflect this new behavior
• 2003/01/18 Fixed unit tests to work on solaris, configure/make mostly works on solaris except that gmake fails until ltmain.sh is made executable with chmod+x
• 2003/01/17 Contributions by Robin Whitworth, rwhit@goldwiretech.com
  1. SQLFetch.c - There was a bug in SQLFetch() if row_status_ptr is not set.
  2. SQLPrimaryKeys.c - There was a join constraint missing in the query. Also, the results were not being ordered correctly.
  3. common.c,h; oracle_functions.c - in ood_driver_execute_describe, the address of the data_size field of an implementation row descriptor (IR_TAG) is being passed to OCIAttrGet for the OCI_ATTR_DATA_SIZE attribute. This should be a ub2, not a ub4. (Probably works OK on linux due to differing byte ordering. Fails miserably on solaris). I changed the definition in common.h & the initializations in common.c & oracle_functions.c
  4. . common.h - the macro GLOBAL_MUTEX_LOCK was wrong. It was doing an unlock! Our code uses 4 or more database threads and this caused the lexer to break.
  5. sqllex.l - the lexer was stripping newlines. This caused problems where the only whitespace between to tokens was a newline. I modified it to preserve newlines - oracle handles them fine, so no need to strip or convert them.
• 2003/01/06 fix for "constructor" of ip_tag and ap_tag, filter out trailing semicolons from sql as the minos source code is chock full of them. Added functionality to SQLGetInfo
• 2002/12/11 Automake and config cleanups contributed by Stefan.Radman@ctbto.org also a bug fix to ocivnu_sqlnts in oracle_functions.c
• 2002/11/14 Changed various unimplemented functions/attributes to return SQL_SUCCESS_WITH_INFO and a IM001 status message instead of SQL_ERROR. My reading of the documentation leads me to believe this is the correct behavior.
• 2002/08/13 Added unit-tests to read and write dates. 21 tests now.
• 2002/08/13 Fixed problem with unit-test of array inserts. All 19 unit tests report success now. The array inserts work properly and probably always have, my unit test died on an assertion after informing the driver that it was about to perform an array insert.
  Cleaned up SQLGetInfo and SQLSetStmtAttr, gettting rid of lots of ugly #ifdefs while retaining functionality they supported.
• 2002/06/26 fixed bind problem in oracle_functions.c which bound string characters with incorrect buffer length. Bound string inserts now work properly.
• 2002/06/24 Installed extensive tracing facilities. Setenv DEBUG 2 reports on all SQLxxx functions called along with input parameters and return status. Setenv DEBUG 3 reports on all OCIxxx functions called along with input parameters and return status.
• 2002/06/06 fixed LD_PRELOAD/core dump problem for oracle 8.1.6 and 8.1.7
• 2002/05/31 added reasonable comments to the unit tests fixed SQLPrepare so that it allocates space in the statement for the column ap and ir descriptors instead of waiting for SQLBindCol to do it. This defect was causing libodbc++ (and hence RDBC) to core dump.
• 2002/05/24 fixed a buffer overflow in getinfo1.c unit test program
• 2002/05/23 refinements to unit tests, detect oracle client version and warn of 8.1.6/7 bug
• 2002/05/22 more test tweaking, added a unit test to check for all the unixODBC and ORACLE environment setup that must happen prior to the c programs in the test subdirectory executing successfully.
• 2002/05/16 new unit test to exercise SQLGetInfo.c more fully, tests all valid inputs at the level of 'it accepts them and doesnt core dump'. Testing all 177 valid inputs for correct functioning will take some effort!
• 2002/05/14 bug fixes to some malloc/free code, added 'constructors' for ar and ir environment handles and asserts to try to catch struct corruption at earliest possible time.
• 2002/05/09 changed/added some unit tests
• 2002/05/02 fixed unused var warnings and flow of control problems found by insure++
• 2002/04/03 SQLGetInfo.c now correctly reports that it supports transactions more testing of transaction capability
• 2002/03/08 added commit/rollback capability through SQLEndTran, more tests for SQLColAttribute
• 2002/03/06 took out spurious warning about stubbed function in SQLRowCount.c Changed oracle_functions.c to deal with float(126) and integer (actually NUMBER(38,0) more correctly.
• 2002/03/05 added functionality to oracle_functions.c more extensive testing of SQLColAttribute and SQLDescribeCol
• 2002/02/26 more directory cleanup, remove generated files (i.e. Makefile) that automake should take care of from repository.
• 2002/02/23 added some missing cases to the switch statement of SQLColAttribute.c
• 2002/02/20 changed error reporting for stubbed out functions. Added function calls to unit tests in 'test' subdirectory.
• 2002/02/11 rearranged project into subdirectories, featuring ./src and ./test subdirectories and re-worked Makefile.am Makefile.in etc. Changed malloc/realloc/free function calls to GNU-style ORAMALLOC, ORAREALLOC, and ORAFREE macros which should prevent double-freeing, or proceeding if memory is exhausted.