Re: dfmgr additional ABI version fields

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Andres Freund <andres(at)anarazel(dot)de>
Cc: pgsql-hackers(at)lists(dot)postgresql(dot)org, Peter Eisentraut <peter(dot)eisentraut(at)enterprisedb(dot)com>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: dfmgr additional ABI version fields
Date: 2021-10-07 16:42:23
Message-ID: 82163.1633624943@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Andres Freund <andres(at)anarazel(dot)de> writes:
> On October 7, 2021 8:49:57 AM PDT, Tom Lane
>> I'm also kind of unclear on why we need to do anything about this
>> in the community version. If someone has forked PG and changed
>> APIs to the extent that extensions are unlikely to work, there's
>> not much stopping them from also making the two-line change
>> to fmgr.h that would be needed to guarantee that different magic
>> struct contents are needed.

> I can see two reasons. First, it'd probably allow stock pg to generate a better error message when confronted with such a module. Second, there's some value in signaling forks that they should change (or think about changing), that field.

Hmm, ok, I can buy the first of those arguments. Less sure about
the second, but the first is reason enough.

Can we make the addition be a string not a number, so that we
could include something more useful than "1234" in the error
message? Something like "Module is built for EDB v1234.56"
seems like it'd be a lot more on-point to the average user,
and it gets us out of having to design the ABI versioning scheme
that a fork should use.

regards, tom lane

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Mark Dilger 2021-10-07 16:46:39 Re: Role Self-Administration
Previous Message Robert Haas 2021-10-07 16:32:16 Re: storing an explicit nonce