Re: 7.1 pg_dump fails for user-defined types (release stopper?)

At 14:55 30/03/01 -0500, Tom Lane wrote:
>
>A more promising idea
>is to hack function creation so that the OID assigned to the function
>is lower than the OIDs assigned to any shell types created when the
>function is defined.

This seems hard; would it be better to have the CREATE TYPE use a new OID,
and fixup the refs?

>Or we could try to hack pg_dump to fix this,
>but that doesn't seem appetizing.

This *may* not be all that hard; there is a currently unused (always NULL)
parameter on the pg_dump ArchiveEntry calls intended for extra
dependencies. For UDTs, we could set the this to be the max OID that
references the type (or a list of OIDs, if we had to), then modify the
pg_restore sort code to check these values if not NULL. ie.

(TOC2 > TOC1)
iff (Max(TOC2.OID, TOC2.DEPS) > Max(TOC1.OID, TOC1.DEPS))
OR ( Max(TOC2.OID, TOC2.DEPS) = Max(TOC1.OID, TOC1.DEPS)
And TOC1.OID = Max(TOC2.DEPS)
)

Where DEPS is a list of OIDs the TOC entry depends on.

(I *think* that's right...).

Since this will only be used when the args is non-null, this code would
only be activated in the current broken case.

----------------------------------------------------------------
Philip Warner | __---_____
Albatross Consulting Pty. Ltd. |----/ - \
(A.B.N. 75 008 659 498) | /(@) ______---_
Tel: (+61) 0500 83 82 81 | _________ \
Fax: (+61) 0500 83 82 82 | ___________ |
Http://www.rhyme.com.au | / \|
| --________--
PGP key available upon request, | /
and from pgp5.ai.mit.edu:11371 |/