Am Montag, 4. April 2005 17:36 schrieb Tom Lane:
> Janning Vygen <vygen@gmx.de> writes:
> > i have a databse in postgresql 7.4 with some pdf files in a bytea column.
> >
> > everything works fine but sometimes when i dump and restore i get
> >
> > psql:../tmp/dump.sql:704022: ERROR: invalid input syntax for type bytea
> > CONTEXT: COPY dk_dokument, line 127, column dk_content:
> > "%PDF-1.4\015%\342\343\317\323\015\01220 0 obj\015<< \015/Linearized 1
> > \015/O 22 \015/H [ 660 209 ] \..."
>
> You might look into whether your handling of the dump file is perfectly
> clean --- is there any possibility of newline transformations getting
> applied to it, for example? AFAICS the above cannot happen if the
> identical data dumped out by pg_dump is reloaded.
what i do is part of my nightly cronjob.
On my DB Server i do:
$ /usr/local/pgsql/bin/pg_dump -O myuser > dump.sql
On my Testserver i do:
$ scp -q -C myuser@dbserver:dump.sql ./dump.sql
$ cp ./dump.sql ~/dump.sql
$ psql -q -f ~/dump.sql dbname >/dev/null
I can't see any command which does newline translations or something else, but
maybe my '-C' in scp makes the difference. I had problems with pg_dump when i
used the compressed format with -Fc
But both servers should be identical. But i compared gzip, ssh, and zlib and
they are not:
dbserver$ openssh-3.5p1-42
testserver$ openssh-3.5p1-107
The rest is identical. i am wondering about this minor package number making
any difference.
Ok. It seems to be my problem and not postgresql specific. Many thanks for
setting me on the right track.
kind regards,
Janning Vygen