On Tue, 6 Feb 2018 19:33:10 -0700 (MST), Nick
<[email protected]> wrote:
>> (a) an error result of some kind or (b) a corrupt database.
>
> I did not see any info about errmsg.
Your code doesn't check the returncode of the sqlite3_* calls.
>> Are your processes using the same database connection or does each one
>> have its own ?
>
> Two processes have two sqlite3_open(). So each one has its own.
>
>> Are you checking the result codes returned by all the API calls ?
>
> Yes. I use speedtest1.c as model code.
> speedtest1_exec("BEGIN");
> speedtest1_prepare("INSERT INTO t1 VALUES(?1,?2,?3); -- %d times",
> n);
> for(i=1; i<=n; i++){
> rc = sqlite3_bind_int64(g.pStmt, 1, i);
> rc = sqlite3_bind_int(g.pStmt, 2, i);
> rc = sqlite3_bind_text(g.pStmt, 3, "hello", -1, SQLITE_STATIC);
> speedtest1_run();
> }
> speedtest1_exec("COMMIT");
>
> And I have checked rc = SQLITE_OK.
I assume speedtest1_run() calls sqlite3_step(), but does it
check the return code? I suspect _step() returns SQLITE_BUSY
sometimes.
>
>> Can you reliably get less than 20000 rows ?
>
> Yes, always less than 20000.
> Process A inserts 1-10000 and process B inserts 10001-20000. I found that
> the first few rows is missing in the result. I mean there is no 10001-10xxx.
>
>> Does the problem go away if you use threadsafe = 2 ?
>
> The problem is still here.
Because the test uses two processes, every process is
single-threaded, no risc of concurrency between threads.
PRAGMA busy_timeout might be helpful.
--
Regards,
Kees Nuyt
_______________________________________________
sqlite-users mailing list
[email protected]
http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
