On Sat, Nov 6, 2010 at 6:09 PM, Robert Haas <robertmhaas@gmail.com> wrote:
> If we're going to try to fix this, we probably ought to try to make
> sure that we are fixing it fairly completely. How confident are you
> that this is the only problem?
I haven't tried to isolate problems on really complicated schemas yet,
but I can tell you what I did do: I went through pg_dump and tried to
find as many parts of the code that added dropStmt to DumpableObject
nodes as possible (in the dumpFoo family of functions). ALTER seemed
like the only interesting bit so far.
But we can give the mechanic a try with, say, a PLPGSQL hack for a
little while to see if it basically gets the job done in sort-of real
life for a while. I think that's not a bad idea....I'll see what I can
do about that. (Of course, continue to share your suggestions and
revelations, I'd appreciate it)
fdr