Hiroshi Inoue <Inoue(at)tpf(dot)co(dot)jp> writes:
> BufferSync() after heap_insert() seems to be lost.
> Connection to a database calls GetRawDatabaseinfo()
> which doesn't look at shared buffer pool at all.
> Connection to a database immediately after the database
> was created fails under WAL e.g. in regression test.
Hmm, perhaps I misunderstood what that was for. I moved it up to
before the system('cp -r') operation in hopes of ensuring that the
source database was up-to-date. But maybe we need another one at
the bottom.
regards, tom lane