Re: Detecting corrupted pages earlier

Alvaro Herrera <alvherre(at)dcc(dot)uchile(dot)cl> writes:
> Huh, and what if I accidentaly mistype the number and destroy a valid
> page? Maybe the command should only succeed if it confirms that the
> page is corrupted.

Good point ... but what if the corruption is subtle enough that the
automatic tests don't notice it? It could be that only "SELECT * FROM ..."
actually dumps core. This is an acute problem for tests that are
generic enough to be implemented in just one or a few places --- for
example, the tests I just added have no idea whether they are looking
at a heap or index page, much less what kind of index page.

> There could also be "DESTROY ALL PAGES" and/or "DESTROY n PAGES"
> commands, where the latter zeroes the first n corrupted pages in the
> table, for the case with lots of corrupted pages. Maybe a command for
> getting a list of the corrupted pages is useful?

We have "DESTROY ALL PAGES", it's called DROP TABLE. The other doesn't
appeal to me either, because it doesn't say exactly which N pages you're
willing to lose.

regards, tom lane