Thread:

Hi Everybody,

I am sorry if this is an inappropriate way of finding out the information, but I suspect that there is a bug in postgres on windows, I tried searching for a similar bug on the net but only found pgsql-bugs@postgresql.org. to report bugs, before I send it as a potential bug I would like to make sure that it has not been identified already. If there is some search tool of reported bugs for postgres please let me know.

Particulars of the suspected bug:

I suspect that there is a memory leak somewhere related to handling blobs

Tested on windows 2003 server sp2 and Red Hat 4 - only happens on windows

While inserting binary data into a  blob filed postgres.exe crashes

Last statement in pg_log:

2008-09-19 22:15:16 PDT LOG: INSERT INTO db.user.table (objectid, area, perimeter, tbd90_, tbd90_id, code_90, cnty_nm, local_nm, nmcenmq, nmhanzi, a5,…shape) VALUES ( $1, $2, $3, $4, $5,…  $204 )

2008-09-19 22:15:16 PDT DETAIL:  parameters: $1 = '402', $2 = '3642082536.41162', $3 = '340956.4869', $4 = '403', $5 = '403', …, $204 = '390F0000430100000800100023000000A13C010001000000BAB2A791FB9B8316BAD

Looking in Event Viewer at the crashed process:

Faulting application postgres.exe, version 8.3.0.831, faulting module msvcr80.dll, version 8.0.50727.1433, fault address 0x0001e44a

Often when postgres crashes it is very good at recovering, in this case the left over postgres processes have to be killed manually.

We work primarily with spatial data, so the blob field is used to store either geometric shapes or raster pixels, both types of data has been tested and in both cases the failure occurs on the insert of the blob.