Thread: dump, restore, dump yields differences
As part of testing the matview issues around dumping, I ran `make installcheck-world`, ran pg_dump to dump the regression database, loaded it into a new database, dumped the newly restored database, and compared the output files from the two pg_dump runs. There were a few somewhat surprising differences, attached. Do we want to worry about these, or leave them alone? -- Kevin Grittner EnterpriseDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company
Attachment
Kevin Grittner <kgrittn@ymail.com> writes:
> As part of testing the matview issues around dumping, I ran `make
> installcheck-world`, ran pg_dump to dump the regression database,
> loaded it into a new database, dumped the newly restored database,
> and compared the output files from the two pg_dump runs.� There
> were a few somewhat surprising differences, attached.� Do we want
> to worry about these, or leave them alone?
Those are expected. You can trace the ALTER TABLE history of those
tables if you want to see why they're so odd, but basically there
are inheritance situations where it's hard to avoid this.
regards, tom lane
On Fri, Mar 22, 2013 at 6:49 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote: > Those are expected. You can trace the ALTER TABLE history of those > tables if you want to see why they're so odd, but basically there > are inheritance situations where it's hard to avoid this. Incidentally it would still be cool to have make check do this dance and intelligently compare the before and after. There have been more than one patch where you've caught omissions in pg_dump before applying. -- greg