Re: Postgres odbc driver bug - Mailing list pgsql-interfaces

Keith, forget my earlier request about reviewing the patch.  One of our
ODBC guys (Hiroshi) says your patch fixes a different problem.

If no one objects, I will be applying yours in a day or two.  Can you
make sure my version of your patch is correct?  I will change the #if 0
and just remove that code, and remove your //chuck comments.

Thanks.


> We have found a reproducible problem with the Postgres ODBC driver.  It has
> to do with the parameter replacement feature in the driver.  Below is a
> piece of test code that shows the problem.  We were able to fix the problem
> by making a change to execute.c in the ODBC driver code (attached).
>
> odbc driver version : Found in both 6.50 and 7.01.0004.
> postgresql database version : Found with both 7.0.2 and 7.0.3 versions of
> Postgres.
> Server Operating System: Windows2000
> Client Operation System: Windows2000
>
> Explanation of the odbc driver bug:
>
>
>    /*****************************************
>     * ERROR HAPPENS HERE:
>     * If we set the buffer that we binded to column "text1" now, it will
> FAIL.
>     * The SQLExecute in doQuery() will succeed (finding 0 rows), and the
>     * SQLFetchScroll call will return SQL_NO_DATA_FOUND.
>     *
>     * This is because of a bug in the odbc driver. The call to
> SQLNumResultCols
>     * will execute the query in the driver, but the buffer that is binded is
>
>     * empty, so no rows will be found.  This of course assumes that the
> table
>     * doesn't contain any empty records for the "text1" column, which is
> correct
>     * for our test.  Then when we call SQLExcute, the driver will realize it
> has
>     * already done the query work and will not do it again, even though our
>     * buffer for parameter substitution has changed (next line of code below
>
>     * this comment).
>     *****************************************/
>
>  Zip file is attached.  The code is currently set up to fail, but I describe
> in my comments that you can uncomment the code that sets the buffer before
> SQLNumResultCols to make things will work.
>
> Thanks for your attention,
> Keith Millard
>
>
>   _____
>
>  <http://www.swifttouch.com/SwiftCardJump.asp?SwiftLink=95NN9NB1Q1FWA> Alex
> Shore..
> Senior Software Engineer..
> Pumatech..
> tel 603-888-0666 x1017..
> ..
> ashore@pumatech.com..
> www.pumatech.com..
>
>

[ Attachment, skipping... ]

[ Attachment, skipping... ]

>
> ---------------------------(end of broadcast)---------------------------
> TIP 3: if posting/reading through Usenet, please send an appropriate
> subscribe-nomail command to majordomo@postgresql.org so that your
> message can get through to the mailing list cleanly

--
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 853-3000
  +  If your life is a hard drive,     |  830 Blythe Avenue
  +  Christ can be your backup.        |  Drexel Hill, Pennsylvania 19026
*** e2    Tue May  8 12:49:51 2001
--- execute.c    Tue May  8 12:47:48 2001
***************
*** 204,209 ****
--- 204,210 ----
          it from an SQLPrepare/SQLDescribeCol type of scenario.  So
          just return success.
      */
+ #if 0 //chuck
      if ( stmt->prepare && stmt->status == STMT_PREMATURE) {
          stmt->status = STMT_FINISHED;
          if (stmt->errormsg == NULL) {
***************
*** 216,221 ****
--- 217,223 ----
              return SQL_ERROR;
          }
      }
+ #endif

      mylog("%s: clear errors...\n", func);

***************
*** 248,255 ****
      }

      /*    Check if the statement is in the correct state */
!     if ((stmt->prepare && stmt->status != STMT_READY) ||
!         (stmt->status != STMT_ALLOCATED && stmt->status != STMT_READY)) {

          stmt->errornumber = STMT_STATUS_ERROR;
          stmt->errormsg = "The handle does not point to a statement that is ready to be executed";
--- 250,259 ----
      }

      /*    Check if the statement is in the correct state */
!     //chuck if ((stmt->prepare && stmt->status != STMT_READY) ||
!        //chuck (stmt->status != STMT_ALLOCATED && stmt->status != STMT_READY)) {
!    if ((stmt->prepare && (stmt->status != STMT_READY && stmt->status != STMT_PREMATURE)) || //chuck
!       (stmt->status != STMT_ALLOCATED && stmt->status != STMT_READY && stmt->status != STMT_PREMATURE)) { //chuck

          stmt->errornumber = STMT_STATUS_ERROR;
          stmt->errormsg = "The handle does not point to a statement that is ready to be executed";
***************
*** 262,270 ****
--- 266,277 ----
      /*    The bound parameters could have possibly changed since the last execute
          of this statement?  Therefore check for params and re-copy.
      */
+     mylog("%s: Check for data at execution parameters\n", func);
+     mylog("%s: stmt->parameters_allocated = %d\n", func,stmt->parameters_allocated);
      stmt->data_at_exec = -1;
      for (i = 0; i < stmt->parameters_allocated; i++) {
          /*    Check for data at execution parameters */
+       mylog("%s: stmt->parameters[%d].data_at_exec = %d\n", func,i,stmt->parameters[i].data_at_exec);
          if ( stmt->parameters[i].data_at_exec == TRUE) {
              if (stmt->data_at_exec < 0)
                  stmt->data_at_exec = 1;