Zdenek Kotala <Zdenek.Kotala@Sun.COM> writes:
> Good catch. if we have to bump catalog version then I have there small patch
> which introduce following macro and remove PageHeaderData stucture from
> HashMetaPageData:
Seems like a bad idea --- PageGetContents is not guaranteed to deliver
a maxaligned pointer, so at least in principle this could result in
alignment violations. It would accidentally fail to fail as of CVS
HEAD, but any future rearrangement of PageHeaderData or HashMetaPageData
could create a problem.
(Possibly PageGetContents *should* guarantee a maxaligned result,
but the current coding does not.)
regards, tom lane