Horton

Re: pg_upgrade fails to detect unsupported arrays and ranges

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Daniel Gustafsson <daniel@yesql.se>
Cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Tomas Vondra <tomas.vondra@enterprisedb.com>
Date: 2021-04-28T20:47:51Z
Lists: pgsql-hackers 
Daniel Gustafsson <daniel@yesql.se> writes:
>> On 28 Apr 2021, at 17:09, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> +		pg_fatal("Your installation contains system-defined composite type(s) in user tables.\n"
>> +				 "These type OIDs are not stable across PostgreSQL versions,\n"
>> +				 "so this cluster cannot currently be upgraded.  You can\n"
>> +				 "remove the problem tables and restart the upgrade.\n"
>> +				 "A list of the problem columns is in the file:\n"

> Would it be helpful to inform the user that they can alter/drop just the
> problematic columns as a potentially less scary alternative to dropping the
> entire table?

This wording is copied-and-pasted from the other similar tests.  I agree
that it's advocating a solution that might be overkill, but if we change
it we should also change the existing messages.  I don't mind doing
that in HEAD; less sure about the back branches, as (I think) these
are translatable strings.

Thoughts?

>> -		 * The type of interest might be wrapped in a domain, array,
>> +		 * The types of interest might be wrapped in a domain, array,

> Shouldn't this be "type(s)” as in the other changes here?

Fair enough.

			regards, tom lane

Commits

  1. Improve wording of some pg_upgrade failure reports.

  2. Fix some more omissions in pg_upgrade's tests for non-upgradable types.

  3. Handle arrays and ranges in pg_upgrade's test for non-upgradable types.