isql leaks requests?

[Dimitry Sibiryakov]

Hello All,

in ISQL code I see quite a few calls to compileRequest() but no calls to
IRequest->free(). How is it supposed to work?

--
WBR, SD.

[Alex Peshkoff]

All are gpre-generated, yes?
Stored in static variables  - like:

static Firebird::IRequest*
   fb_1121 = 0;     /* request handle */

Released after attaching another database:

if (fbStatus->getErrors()[1] == isc_bad_req_handle) {
fb_1121->release(); fb_1121 = NULL; }

Or when exiting isql.

[Dimitry Sibiryakov]

'Alex Peshkoff' via firebird-devel wrote 14.12.2022 9:55:
> Released after attaching another database:
>
> if (fbStatus->getErrors()[1] == isc_bad_req_handle) { fb_1121->release();
> fb_1121 = NULL; }
>
> Or when exiting isql.

It means that result set and locks are kept indefinitely. Such a waste of
resources...

--
WBR, SD.

[Alex Peshkoff]

That's how it used to work since at least IB6.0.1.
Agree that with server-side statements cache it has absolutely no sense.

[Dimitry Sibiryakov]

'Alex Peshkoff' via firebird-devel wrote 14.12.2022 14:24:
>>   It means that result set and locks are kept indefinitely. Such a waste of
>> resources...
>>
>
> That's how it used to work since at least IB6.0.1.
> Agree that with server-side statements cache it has absolutely no sense.

It also could explain garbage temporary files that gathering in temp
directories: without proper release there is no execution point where they can
be deleted.

--
WBR, SD.

[Dimitry Sibiryakov]

'Alex Peshkoff' via firebird-devel wrote 14.12.2022 14:24:
> That's how it used to work since at least IB6.0.1.

BTW, is it normal that single "SHOW TABLE T" in isql allocates 13 requests or
should I look for a bug somewhere at my side?

--
WBR, SD.

[Alex Peshkoff]

Depends upon table. If it has various constraints - 13 is ok.

[Adriano dos Santos Fernandes]

One of reasons that I like Oracle approach where metadata extraction is
done through a server-side reusable package (DBMS_METADATA).


Adriano