Обсуждение: Novice lo_import problem (7.0.3)

Поиск
Список
Период
Сортировка

Novice lo_import problem (7.0.3)

От
Rhys Jones
Дата:
Hello,

I hope this isn't too obvious a question.

I'm currently using Postgres 7.0.3, having installed the postgresql.org
i386 RPMs for RedHat 6.x on an RH 6.1 machine. I'm attempting to create a
database where one of the fields contains a large object read from the
filesystem.

But on piping the following batch file into psql:

CREATE TABLE FileSystem ( NAME varchar(80), URL varchar(200), LASTUPDATE date, LENGTH numeric(10), DATA oid);
INSERT INTO FileSystem VALUES ('/work/ams/Manfac/Panels/ActionRepOpenJobs_0.pnl', NULL, '03-09-2000 2:24:00', 573,
lo_import('/work/ams/Manfac/Panels/ActionRepOpenJobs_0.pnl'));
[then other similar INSERT statements: batch file's about 1850 lines long]

the backend always, but always, crashes on the ninth INSERT with a 'my
bits moved right off the end of the world' error. I've tried jigging the
batch file around, using different filenames, no joy.

All files for the lo_import up to and including the ninth INSERT are
< 2048 bytes in size, so (admittedly without knowledge of the postgresql
code) an over-run seems unlikely. Any hints or ideas?

Many thanks,
Rhys Jones, Swansea
--
http://www.sucs.org/~rhys/


Re: Novice lo_import problem (7.0.3)

От
Tom Lane
Дата:
Rhys Jones <ohnoitsbroken@postwales.com> writes:
> the backend always, but always, crashes on the ninth INSERT with a 'my
> bits moved right off the end of the world' error.

If you can create a reproducible test case that generates that error,
I would *love* to see it.  I think there is some corner-case bug in
the btree index routines, but finding it without a test case has been
impossible ...

            regards, tom lane