BUG #15627: libpq memory leak

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

BUG #15627: libpq memory leak

apt.postgresql.org Repository Update
The following bug has been logged on the website:

Bug reference:      15627
Logged by:          ice ice
Email address:      [hidden email]
PostgreSQL version: 11.1
Operating system:   Ubuntu 18.04.1 LTS
Description:        

PQisBusy has memory leak with next valgrind stack:

  1: malloc in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so
  2: PQmakeEmptyPGresult in
/opt/src/postgres/src/interfaces/libpq/fe-exec.c:146
  3: getRowDescriptions in
/opt/src/postgres/src/interfaces/libpq/fe-protocol3.c:492
  4: pqParseInput3 in
/opt/src/postgres/src/interfaces/libpq/fe-protocol3.c:293
  5: parseInput in /opt/src/postgres/src/interfaces/libpq/fe-exec.c:1747
  6: PQisBusy in /opt/src/postgres/src/interfaces/libpq/fe-exec.c:1764

The problem is that if PQisBusy doesn't revoke memory for item

/* Status for asynchronous result construction */
        PGresult   *result; /* result being constructed */

in struct pg_conn

Reply | Threaded
Open this post in threaded view
|

Re: BUG #15627: libpq memory leak

Tom Lane-2
PG Bug reporting form <[hidden email]> writes:
> PQisBusy has memory leak with next valgrind stack:

>   1: malloc in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so
>   2: PQmakeEmptyPGresult in
> /opt/src/postgres/src/interfaces/libpq/fe-exec.c:146
>   3: getRowDescriptions in
> /opt/src/postgres/src/interfaces/libpq/fe-protocol3.c:492
>   4: pqParseInput3 in
> /opt/src/postgres/src/interfaces/libpq/fe-protocol3.c:293
>   5: parseInput in /opt/src/postgres/src/interfaces/libpq/fe-exec.c:1747
>   6: PQisBusy in /opt/src/postgres/src/interfaces/libpq/fe-exec.c:1764

I don't believe this is actually a leak: what is happening is that
we're collecting the query result that will eventually be returned
by PQgetResult.  If the calling program never calls PQgetResult
after PQisBusy, then tools like valgrind might claim that's a leak,
but the claim has nothing to do with reality.

If you think there's a real problem here, please show a self-contained
test case that causes steady increase in memory consumption.

                        regards, tom lane

Reply | Threaded
Open this post in threaded view
|

Re: BUG #15627: libpq memory leak

Sergey Ivanov
Thanks Tom. My fault. Hasn't closed PGRES_TUPLES_OK last empty record in single row mode output

вс, 10 февр. 2019 г. в 19:03, Tom Lane <[hidden email]>:
PG Bug reporting form <[hidden email]> writes:
> PQisBusy has memory leak with next valgrind stack:

>   1: malloc in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so
>   2: PQmakeEmptyPGresult in
> /opt/src/postgres/src/interfaces/libpq/fe-exec.c:146
>   3: getRowDescriptions in
> /opt/src/postgres/src/interfaces/libpq/fe-protocol3.c:492
>   4: pqParseInput3 in
> /opt/src/postgres/src/interfaces/libpq/fe-protocol3.c:293
>   5: parseInput in /opt/src/postgres/src/interfaces/libpq/fe-exec.c:1747
>   6: PQisBusy in /opt/src/postgres/src/interfaces/libpq/fe-exec.c:1764

I don't believe this is actually a leak: what is happening is that
we're collecting the query result that will eventually be returned
by PQgetResult.  If the calling program never calls PQgetResult
after PQisBusy, then tools like valgrind might claim that's a leak,
but the claim has nothing to do with reality.

If you think there's a real problem here, please show a self-contained
test case that causes steady increase in memory consumption.

                        regards, tom lane


--
Kind regards,
Sergey Ivanov