| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Usenet -> c.d.o.server -> Re: Oracle bug (w/ Pro*C)???
Can you unsnip the snipped portions and send table defs and version info? I agree it sounds like a bug but whos' bug?
for example, you print to tmp in get_shift_info. What is tmp? It's not an automatic stack variable is it (if it is, it shouldn't be)? Is it static? do you malloc it. What are descr, start_dt, and so on? Are they big enough? How does ingress format dates? we do it dd-mon-yy, if they did it dd/mm/yy, your strings might be the wrong size.
also, you return NULL from get_shift_info on error and copy that with a strcpy. thats sure to blow up or at least behave really funny depending on the platform.
this is pretty straightforward stuff. It works. The invalid runtime context is saying someone trashed the stack/heap, it is corrupt.
On Fri, 27 Jun 1997 15:58:00 GMT, kenf_at_usinternet.com (Ken Frank) wrote:
>Is there something unusual about the following code?
>For those not interested in wading through about 100
>lines of snipped-up code, I'll outline the problem
>first:
>
> 1) Open up a cursor (a simple query with a union all
> to a 3-table-joined query).
> 2) fetch from cursor.
> 3) after first fetch, a simple query with a join and
> the min() aggregates.
> 4) next fetch from cursor -- blows up with:
> SQL-02134 Invalid runtime context.
>
>...
>char *get_shift_info(int p_shift)
>{
>[snipped declarations]
> shift = p_shift;
> exec sql
> select min(d.description), min(s.start_dt), min(s.end_dt)
> into :descr :descr_null,
> :start_dt :start_dt_null,
> :end_dt :end_dt_null
> from shift_def d, shift s
> where s.shift_code = :shift
> and d.shift_code = s.shift_code;
>
> if (sqlca.sqlcode != 0) return NULL;
>[more snippage dealing with potential null values]
>
> /* this actually gets formatted a bit, but for simplicity... */
> sprintf(tmp, "%s (%s - %s)", descr, start_dt, end_dt);
> return tmp;
>}
>
>
>int list_emp_timesheet(char *p_emp_id, int p_shift)
>{
> [snipped declarations]
>
> exec sql declare lab_csr cursor for
> select start_time, end_time, ind_acct, 0, ' ',
> location, duration, source, ' '
> from acc_lab_chg
> where emp_id = :emp_id
> and start_time between :p_dt1 and :p_dt2
> UNION ALL
> select a.start_time, a.end_time, ' ', a.wo_no, b.job,
> b.location, a.duration, b.source, u.unit_no
> from o_labor_chg a, o_wo b, unit_main u
> where a.wo_no = b.wo_no
> and b.unit_id = u.unit_id
> and a.emp_id = :emp_id
> and a.start_time between :dt1 and :dt2
> order by 1, 4;
>
> exec sql open lab_csr;
> rows_fetched = 0;
> while (sqlca.sqlcode == 0) {
> exec sql fetch lab_csr into ...;
> if (sqlca.sqlcode != 0) break;
>
> rows_fetched++;
> if (rows_fetched == 1) {
> /*
> | there's going to be data for this employee, so we
> | need to get their "shift-description"...
> */
> strcpy(shift_descr, get_shift_info(shift));
> }
>
> printf("...omitted large printf...\n");
> }
>}
>...
>
>This thing keeps puking on the SECOND fetch! The error message I'm
>getting out of sqlca is "SQL-02134 Invalid runtime context". A
>colleague tells me that his reference manual says:
> cause: The runtime context associated with an executable
> SQL statement was not properly allocated.
> action: Rewrite you application to execute the EXEC SQL
> CONTEXT ALLOCATE statement before executing any
> SQL statements.
>Program flow is:
>1) fetch, (sqlca.sqlcode is ZERO)
>2) select-on shift_def/shift tables (sqlca.sqlcode is ZERO)
>3) fetch, (poof, somthing nasty happened to my cursor!)
>
>I am not doing any commits/rollbacks here (besides that'd be
>a different error)!
>
>This seems like a BUG to me! This code works fine as-is under Ingres,
>and under Oracle, if I simplify either of those select statements
>(the cursor's query or the shift_def/shift query), it works. Can
>anyone tell me why Oracle is having trouble with this? I've got
>a workaround (there CAN'T be more than 255 shifts anyway, so I'm
>just preloading them all into an array before anything else, but
>I just had to vent my frustrations here, as I've wasted so much
>time on it).
>
>Thanks
>
>PS - El-cheapo anti-spam measures in effect:
> my email isn't really kenf_at_usinternet.com (used to be!)
> instead, you can reply to ken AT mn.csoft.com
>
>OR, just reply to the group! I read here as often as I can.
>
>
Thomas Kyte
tkyte_at_us.oracle.com
Oracle Government
Bethesda MD
http://govt.us.oracle.com/ -- downloadable utilities
 |
 |